diff --git a/docs/_cpp_api/classtorch__tensorrt_1_1DataType.html b/docs/_cpp_api/classtorch__tensorrt_1_1DataType.html index b76eab3a5b..f047fbaa05 100644 --- a/docs/_cpp_api/classtorch__tensorrt_1_1DataType.html +++ b/docs/_cpp_api/classtorch__tensorrt_1_1DataType.html @@ -10,7 +10,7 @@ - Class DataType — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Class DataType — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/classtorch__tensorrt_1_1Device_1_1DeviceType.html b/docs/_cpp_api/classtorch__tensorrt_1_1Device_1_1DeviceType.html index 48e39b17d7..ded518e39f 100644 --- a/docs/_cpp_api/classtorch__tensorrt_1_1Device_1_1DeviceType.html +++ b/docs/_cpp_api/classtorch__tensorrt_1_1Device_1_1DeviceType.html @@ -10,7 +10,7 @@ - Class Device::DeviceType — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Class Device::DeviceType — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/classtorch__tensorrt_1_1TensorFormat.html b/docs/_cpp_api/classtorch__tensorrt_1_1TensorFormat.html index 22e895dd11..4030b472b1 100644 --- a/docs/_cpp_api/classtorch__tensorrt_1_1TensorFormat.html +++ b/docs/_cpp_api/classtorch__tensorrt_1_1TensorFormat.html @@ -10,7 +10,7 @@ - Class TensorFormat — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Class TensorFormat — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8CacheCalibrator.html b/docs/_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8CacheCalibrator.html index f645cc770f..8c0e1f4fc5 100644 --- a/docs/_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8CacheCalibrator.html +++ b/docs/_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8CacheCalibrator.html @@ -10,7 +10,7 @@ - Template Class Int8CacheCalibrator — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Template Class Int8CacheCalibrator — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8Calibrator.html b/docs/_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8Calibrator.html index 28c9b1f1a2..0a62d5e404 100644 --- a/docs/_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8Calibrator.html +++ b/docs/_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8Calibrator.html @@ -10,7 +10,7 @@ - Template Class Int8Calibrator — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Template Class Int8Calibrator — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/define_macros_8h_1a18d295a837ac71add5578860b55e5502.html b/docs/_cpp_api/define_macros_8h_1a18d295a837ac71add5578860b55e5502.html index 41bd8ef1e1..fdbfbb3921 100644 --- a/docs/_cpp_api/define_macros_8h_1a18d295a837ac71add5578860b55e5502.html +++ b/docs/_cpp_api/define_macros_8h_1a18d295a837ac71add5578860b55e5502.html @@ -10,7 +10,7 @@ - Define STR — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Define STR — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/define_macros_8h_1a282fd3c0b1c3a215148ae372070e1268.html b/docs/_cpp_api/define_macros_8h_1a282fd3c0b1c3a215148ae372070e1268.html index 3755971167..3e690afbc5 100644 --- a/docs/_cpp_api/define_macros_8h_1a282fd3c0b1c3a215148ae372070e1268.html +++ b/docs/_cpp_api/define_macros_8h_1a282fd3c0b1c3a215148ae372070e1268.html @@ -10,7 +10,7 @@ - Define TORCH_TENSORRT_PATCH_VERSION — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Define TORCH_TENSORRT_PATCH_VERSION — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/define_macros_8h_1a31398a6d4d27e28817afb0f0139e909e.html b/docs/_cpp_api/define_macros_8h_1a31398a6d4d27e28817afb0f0139e909e.html index 49fc577ef0..ffb008f389 100644 --- a/docs/_cpp_api/define_macros_8h_1a31398a6d4d27e28817afb0f0139e909e.html +++ b/docs/_cpp_api/define_macros_8h_1a31398a6d4d27e28817afb0f0139e909e.html @@ -10,7 +10,7 @@ - Define TORCH_TENSORRT_MAJOR_VERSION — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Define TORCH_TENSORRT_MAJOR_VERSION — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/define_macros_8h_1a35703561b26b1a9d2738ad7d58b27827.html b/docs/_cpp_api/define_macros_8h_1a35703561b26b1a9d2738ad7d58b27827.html index abd76aaf2c..d2739728ce 100644 --- a/docs/_cpp_api/define_macros_8h_1a35703561b26b1a9d2738ad7d58b27827.html +++ b/docs/_cpp_api/define_macros_8h_1a35703561b26b1a9d2738ad7d58b27827.html @@ -10,7 +10,7 @@ - Define TORCH_TENSORRT_MINOR_VERSION — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Define TORCH_TENSORRT_MINOR_VERSION — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/define_macros_8h_1abd1465eb38256d3f22cc1426b23d516b.html b/docs/_cpp_api/define_macros_8h_1abd1465eb38256d3f22cc1426b23d516b.html index f7631cb719..559019cee5 100644 --- a/docs/_cpp_api/define_macros_8h_1abd1465eb38256d3f22cc1426b23d516b.html +++ b/docs/_cpp_api/define_macros_8h_1abd1465eb38256d3f22cc1426b23d516b.html @@ -10,7 +10,7 @@ - Define TORCHTRT_API — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Define TORCHTRT_API — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/define_macros_8h_1abe87b341f562fd1cf40b7672e4d759da.html b/docs/_cpp_api/define_macros_8h_1abe87b341f562fd1cf40b7672e4d759da.html index d199183862..4ce6151964 100644 --- a/docs/_cpp_api/define_macros_8h_1abe87b341f562fd1cf40b7672e4d759da.html +++ b/docs/_cpp_api/define_macros_8h_1abe87b341f562fd1cf40b7672e4d759da.html @@ -10,7 +10,7 @@ - Define XSTR — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Define XSTR — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/define_macros_8h_1ad19939408f7be171a74a89928b36eb59.html b/docs/_cpp_api/define_macros_8h_1ad19939408f7be171a74a89928b36eb59.html index c5a3089800..78121f3b08 100644 --- a/docs/_cpp_api/define_macros_8h_1ad19939408f7be171a74a89928b36eb59.html +++ b/docs/_cpp_api/define_macros_8h_1ad19939408f7be171a74a89928b36eb59.html @@ -10,7 +10,7 @@ - Define TORCHTRT_HIDDEN — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Define TORCHTRT_HIDDEN — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/define_macros_8h_1adad592a7b1b7eed529cdf6acd584c883.html b/docs/_cpp_api/define_macros_8h_1adad592a7b1b7eed529cdf6acd584c883.html index f2b83e6210..03617b4915 100644 --- a/docs/_cpp_api/define_macros_8h_1adad592a7b1b7eed529cdf6acd584c883.html +++ b/docs/_cpp_api/define_macros_8h_1adad592a7b1b7eed529cdf6acd584c883.html @@ -10,7 +10,7 @@ - Define TORCH_TENSORRT_VERSION — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Define TORCH_TENSORRT_VERSION — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/dir_cpp.html b/docs/_cpp_api/dir_cpp.html index 79446fff44..df3b1fe568 100644 --- a/docs/_cpp_api/dir_cpp.html +++ b/docs/_cpp_api/dir_cpp.html @@ -10,7 +10,7 @@ - Directory cpp — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Directory cpp — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/dir_cpp_include.html b/docs/_cpp_api/dir_cpp_include.html index 7066725a55..0ec80ad272 100644 --- a/docs/_cpp_api/dir_cpp_include.html +++ b/docs/_cpp_api/dir_cpp_include.html @@ -10,7 +10,7 @@ - Directory include — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Directory include — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/dir_cpp_include_torch_tensorrt.html b/docs/_cpp_api/dir_cpp_include_torch_tensorrt.html index 15340833c0..4911f7506a 100644 --- a/docs/_cpp_api/dir_cpp_include_torch_tensorrt.html +++ b/docs/_cpp_api/dir_cpp_include_torch_tensorrt.html @@ -10,7 +10,7 @@ - Directory torch_tensorrt — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Directory torch_tensorrt — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/enum_namespacetorch__tensorrt_1_1logging_1a130f65408ad8cbaee060f05e8db69558.html b/docs/_cpp_api/enum_namespacetorch__tensorrt_1_1logging_1a130f65408ad8cbaee060f05e8db69558.html index a9abf4dce9..78cc2d8a6d 100644 --- a/docs/_cpp_api/enum_namespacetorch__tensorrt_1_1logging_1a130f65408ad8cbaee060f05e8db69558.html +++ b/docs/_cpp_api/enum_namespacetorch__tensorrt_1_1logging_1a130f65408ad8cbaee060f05e8db69558.html @@ -10,7 +10,7 @@ - Enum Level — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Enum Level — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/enum_namespacetorch__tensorrt_1a3fbe5d72e4fc624dbd038853079620eb.html b/docs/_cpp_api/enum_namespacetorch__tensorrt_1a3fbe5d72e4fc624dbd038853079620eb.html index 3800ca05c4..5d8ec3e872 100644 --- a/docs/_cpp_api/enum_namespacetorch__tensorrt_1a3fbe5d72e4fc624dbd038853079620eb.html +++ b/docs/_cpp_api/enum_namespacetorch__tensorrt_1a3fbe5d72e4fc624dbd038853079620eb.html @@ -10,7 +10,7 @@ - Enum EngineCapability — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Enum EngineCapability — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/file_cpp_include_torch_tensorrt_logging.h.html b/docs/_cpp_api/file_cpp_include_torch_tensorrt_logging.h.html index f4a6419a7b..478f53f248 100644 --- a/docs/_cpp_api/file_cpp_include_torch_tensorrt_logging.h.html +++ b/docs/_cpp_api/file_cpp_include_torch_tensorrt_logging.h.html @@ -10,7 +10,7 @@ - File logging.h — Torch-TensorRT v2.6.0.dev0+e43833d documentation + File logging.h — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/file_cpp_include_torch_tensorrt_macros.h.html b/docs/_cpp_api/file_cpp_include_torch_tensorrt_macros.h.html index 108b32292d..2eaaa7c3ab 100644 --- a/docs/_cpp_api/file_cpp_include_torch_tensorrt_macros.h.html +++ b/docs/_cpp_api/file_cpp_include_torch_tensorrt_macros.h.html @@ -10,7 +10,7 @@ - File macros.h — Torch-TensorRT v2.6.0.dev0+e43833d documentation + File macros.h — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/file_cpp_include_torch_tensorrt_ptq.h.html b/docs/_cpp_api/file_cpp_include_torch_tensorrt_ptq.h.html index 76d79ebb3e..4b5ca33ec3 100644 --- a/docs/_cpp_api/file_cpp_include_torch_tensorrt_ptq.h.html +++ b/docs/_cpp_api/file_cpp_include_torch_tensorrt_ptq.h.html @@ -10,7 +10,7 @@ - File ptq.h — Torch-TensorRT v2.6.0.dev0+e43833d documentation + File ptq.h — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/file_cpp_include_torch_tensorrt_torch_tensorrt.h.html b/docs/_cpp_api/file_cpp_include_torch_tensorrt_torch_tensorrt.h.html index 8910e03326..0edfe70851 100644 --- a/docs/_cpp_api/file_cpp_include_torch_tensorrt_torch_tensorrt.h.html +++ b/docs/_cpp_api/file_cpp_include_torch_tensorrt_torch_tensorrt.h.html @@ -10,7 +10,7 @@ - File torch_tensorrt.h — Torch-TensorRT v2.6.0.dev0+e43833d documentation + File torch_tensorrt.h — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0593f776f469c20469e2f729fc7861a3.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0593f776f469c20469e2f729fc7861a3.html index 5a348677ac..eaf67a354f 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0593f776f469c20469e2f729fc7861a3.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0593f776f469c20469e2f729fc7861a3.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::logging::get_logging_prefix — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::logging::get_logging_prefix — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0c012cb374addd90eb1f42eaec570650.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0c012cb374addd90eb1f42eaec570650.html index d6fc69343c..240e6730e9 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0c012cb374addd90eb1f42eaec570650.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0c012cb374addd90eb1f42eaec570650.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::logging::get_reportable_log_level — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::logging::get_reportable_log_level — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a56e110feaaba2c3fd44bd201fd21a76a.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a56e110feaaba2c3fd44bd201fd21a76a.html index dc01d74788..a0854c288f 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a56e110feaaba2c3fd44bd201fd21a76a.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a56e110feaaba2c3fd44bd201fd21a76a.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::logging::get_is_colored_output_on — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::logging::get_is_colored_output_on — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a7cb50492421ea9de4e3db895819df6f2.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a7cb50492421ea9de4e3db895819df6f2.html index 51a8b2bb4c..411ce88204 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a7cb50492421ea9de4e3db895819df6f2.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a7cb50492421ea9de4e3db895819df6f2.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::logging::set_reportable_log_level — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::logging::set_reportable_log_level — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ac46ac0901cb97e3ae6e93b45f24e90b8.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ac46ac0901cb97e3ae6e93b45f24e90b8.html index 4a2f59ac76..08133f3107 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ac46ac0901cb97e3ae6e93b45f24e90b8.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ac46ac0901cb97e3ae6e93b45f24e90b8.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::logging::log — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::logging::log — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ad2efd47b6c3689e58ccc595680579ae5.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ad2efd47b6c3689e58ccc595680579ae5.html index 6a20840a43..6146261138 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ad2efd47b6c3689e58ccc595680579ae5.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ad2efd47b6c3689e58ccc595680579ae5.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::logging::set_is_colored_output_on — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::logging::set_is_colored_output_on — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1af8f3443813315af7901903d25dd495cc.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1af8f3443813315af7901903d25dd495cc.html index 0bfd489dc9..497c2ed47d 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1af8f3443813315af7901903d25dd495cc.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1logging_1af8f3443813315af7901903d25dd495cc.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::logging::set_logging_prefix — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::logging::set_logging_prefix — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a226e3c83379d1012cde8578c1c86b16c.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a226e3c83379d1012cde8578c1c86b16c.html index 6ec16d2c93..5ae1726094 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a226e3c83379d1012cde8578c1c86b16c.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a226e3c83379d1012cde8578c1c86b16c.html @@ -10,7 +10,7 @@ - Template Function torch_tensorrt::ptq::make_int8_cache_calibrator — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Template Function torch_tensorrt::ptq::make_int8_cache_calibrator — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a6186e305f47c1d94b6130ef6c7f7e178.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a6186e305f47c1d94b6130ef6c7f7e178.html index b80c731c8e..9ecfec7bd7 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a6186e305f47c1d94b6130ef6c7f7e178.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a6186e305f47c1d94b6130ef6c7f7e178.html @@ -10,7 +10,7 @@ - Template Function torch_tensorrt::ptq::make_int8_calibrator — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Template Function torch_tensorrt::ptq::make_int8_calibrator — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a5b405fd3bf3c8fc2e2a54cbbab979797.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a5b405fd3bf3c8fc2e2a54cbbab979797.html index 077677fa69..eaf89c6f8b 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a5b405fd3bf3c8fc2e2a54cbbab979797.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a5b405fd3bf3c8fc2e2a54cbbab979797.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::torchscript::check_method_operator_support — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::torchscript::check_method_operator_support — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a6e19490a08fb1553c9dd347a5ae79db9.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a6e19490a08fb1553c9dd347a5ae79db9.html index cead3b7625..96a3d84cc6 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a6e19490a08fb1553c9dd347a5ae79db9.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a6e19490a08fb1553c9dd347a5ae79db9.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::torchscript::compile — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::torchscript::compile — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a81f9783517335dda877d8cfcf38987c9.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a81f9783517335dda877d8cfcf38987c9.html index c2f5b8678b..81e6d25a5d 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a81f9783517335dda877d8cfcf38987c9.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a81f9783517335dda877d8cfcf38987c9.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::torchscript::embed_engine_in_new_module — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::torchscript::embed_engine_in_new_module — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1ae8d56472106eeef37fbe51ff7f40c9b2.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1ae8d56472106eeef37fbe51ff7f40c9b2.html index d0a89efed3..3c85de164e 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1ae8d56472106eeef37fbe51ff7f40c9b2.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1ae8d56472106eeef37fbe51ff7f40c9b2.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::torchscript::convert_method_to_trt_engine — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::torchscript::convert_method_to_trt_engine — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1ac4ab8313ae72c2c899ea31548b528528.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1ac4ab8313ae72c2c899ea31548b528528.html index 1b7a7b9118..77322f2281 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1ac4ab8313ae72c2c899ea31548b528528.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1ac4ab8313ae72c2c899ea31548b528528.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::get_build_info — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::get_build_info — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1ad1acd06eaeaffbbcf6e7ebf426891384.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1ad1acd06eaeaffbbcf6e7ebf426891384.html index f67f92e42a..bd77e00bc9 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1ad1acd06eaeaffbbcf6e7ebf426891384.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1ad1acd06eaeaffbbcf6e7ebf426891384.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::set_device — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::set_device — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/function_namespacetorch__tensorrt_1ad6a4ee8ca6c8f6e5519eb1128ec7f4a1.html b/docs/_cpp_api/function_namespacetorch__tensorrt_1ad6a4ee8ca6c8f6e5519eb1128ec7f4a1.html index 67846ff766..274a92efaf 100644 --- a/docs/_cpp_api/function_namespacetorch__tensorrt_1ad6a4ee8ca6c8f6e5519eb1128ec7f4a1.html +++ b/docs/_cpp_api/function_namespacetorch__tensorrt_1ad6a4ee8ca6c8f6e5519eb1128ec7f4a1.html @@ -10,7 +10,7 @@ - Function torch_tensorrt::dump_build_info — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Function torch_tensorrt::dump_build_info — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/namespace_torch_tensorrt.html b/docs/_cpp_api/namespace_torch_tensorrt.html index 65b1ad027d..6ef62c95e6 100644 --- a/docs/_cpp_api/namespace_torch_tensorrt.html +++ b/docs/_cpp_api/namespace_torch_tensorrt.html @@ -10,7 +10,7 @@ - Namespace torch_tensorrt — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Namespace torch_tensorrt — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/namespace_torch_tensorrt__logging.html b/docs/_cpp_api/namespace_torch_tensorrt__logging.html index ed75bf2bd2..6bd9323560 100644 --- a/docs/_cpp_api/namespace_torch_tensorrt__logging.html +++ b/docs/_cpp_api/namespace_torch_tensorrt__logging.html @@ -10,7 +10,7 @@ - Namespace torch_tensorrt::logging — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Namespace torch_tensorrt::logging — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/namespace_torch_tensorrt__ptq.html b/docs/_cpp_api/namespace_torch_tensorrt__ptq.html index db58997f97..58d0e8e483 100644 --- a/docs/_cpp_api/namespace_torch_tensorrt__ptq.html +++ b/docs/_cpp_api/namespace_torch_tensorrt__ptq.html @@ -10,7 +10,7 @@ - Namespace torch_tensorrt::ptq — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Namespace torch_tensorrt::ptq — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/namespace_torch_tensorrt__torchscript.html b/docs/_cpp_api/namespace_torch_tensorrt__torchscript.html index c6b56ec85b..b976714a85 100644 --- a/docs/_cpp_api/namespace_torch_tensorrt__torchscript.html +++ b/docs/_cpp_api/namespace_torch_tensorrt__torchscript.html @@ -10,7 +10,7 @@ - Namespace torch_tensorrt::torchscript — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Namespace torch_tensorrt::torchscript — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_logging.h.html b/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_logging.h.html index 2fce602d8c..581f57f64d 100644 --- a/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_logging.h.html +++ b/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_logging.h.html @@ -10,7 +10,7 @@ - Program Listing for File logging.h — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Program Listing for File logging.h — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_macros.h.html b/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_macros.h.html index c64c990492..be14302ed5 100644 --- a/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_macros.h.html +++ b/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_macros.h.html @@ -10,7 +10,7 @@ - Program Listing for File macros.h — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Program Listing for File macros.h — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_ptq.h.html b/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_ptq.h.html index 1bb3c1b5bb..56ea8c366e 100644 --- a/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_ptq.h.html +++ b/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_ptq.h.html @@ -10,7 +10,7 @@ - Program Listing for File ptq.h — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Program Listing for File ptq.h — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_torch_tensorrt.h.html b/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_torch_tensorrt.h.html index fb28c8198e..ab035fc564 100644 --- a/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_torch_tensorrt.h.html +++ b/docs/_cpp_api/program_listing_file_cpp_include_torch_tensorrt_torch_tensorrt.h.html @@ -10,7 +10,7 @@ - Program Listing for File torch_tensorrt.h — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Program Listing for File torch_tensorrt.h — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/structtorch__tensorrt_1_1Device.html b/docs/_cpp_api/structtorch__tensorrt_1_1Device.html index e297639774..71141f6171 100644 --- a/docs/_cpp_api/structtorch__tensorrt_1_1Device.html +++ b/docs/_cpp_api/structtorch__tensorrt_1_1Device.html @@ -10,7 +10,7 @@ - Struct Device — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Struct Device — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/structtorch__tensorrt_1_1GraphInputs.html b/docs/_cpp_api/structtorch__tensorrt_1_1GraphInputs.html index 313faeaf19..27da5a7106 100644 --- a/docs/_cpp_api/structtorch__tensorrt_1_1GraphInputs.html +++ b/docs/_cpp_api/structtorch__tensorrt_1_1GraphInputs.html @@ -10,7 +10,7 @@ - Struct GraphInputs — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Struct GraphInputs — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/structtorch__tensorrt_1_1Input.html b/docs/_cpp_api/structtorch__tensorrt_1_1Input.html index d5343e42f9..5c609bfea5 100644 --- a/docs/_cpp_api/structtorch__tensorrt_1_1Input.html +++ b/docs/_cpp_api/structtorch__tensorrt_1_1Input.html @@ -10,7 +10,7 @@ - Struct Input — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Struct Input — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/structtorch__tensorrt_1_1torchscript_1_1CompileSpec.html b/docs/_cpp_api/structtorch__tensorrt_1_1torchscript_1_1CompileSpec.html index 1f4f736bd9..932bdff992 100644 --- a/docs/_cpp_api/structtorch__tensorrt_1_1torchscript_1_1CompileSpec.html +++ b/docs/_cpp_api/structtorch__tensorrt_1_1torchscript_1_1CompileSpec.html @@ -10,7 +10,7 @@ - Struct CompileSpec — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Struct CompileSpec — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
diff --git a/docs/_cpp_api/torch_tensort_cpp.html b/docs/_cpp_api/torch_tensort_cpp.html index 12f88d1852..bfbc3e76ef 100644 --- a/docs/_cpp_api/torch_tensort_cpp.html +++ b/docs/_cpp_api/torch_tensort_cpp.html @@ -10,7 +10,7 @@ - Torch-TensorRT C++ API — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Torch-TensorRT C++ API — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
- v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
@@ -474,7 +474,7 @@

Class Hierarchy
  • diff --git a/docs/_cpp_api/unabridged_orphan.html b/docs/_cpp_api/unabridged_orphan.html index b16fbb10b7..328ff75536 100644 --- a/docs/_cpp_api/unabridged_orphan.html +++ b/docs/_cpp_api/unabridged_orphan.html @@ -10,7 +10,7 @@ - Full API — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Full API — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_downloads/2c4fd8e65aa979aa6a0402a43ff9b15e/cross_runtime_compilation_for_windows.py b/docs/_downloads/2c4fd8e65aa979aa6a0402a43ff9b15e/cross_runtime_compilation_for_windows.py new file mode 100644 index 0000000000..184470ffa0 --- /dev/null +++ b/docs/_downloads/2c4fd8e65aa979aa6a0402a43ff9b15e/cross_runtime_compilation_for_windows.py @@ -0,0 +1,82 @@ +""" +.. _resnet_cross_runtime_compilation_for_windows_example: + +cross runtime compilation limitations: +The cross compile and saved model can only be loaded in Windows, it can no longer be loaded in Linux +The cross compile and saved model can only be loaded in the same Compute Capability as the Linux which it was cross compiled +(for example, if the model was cross compiled in Linux with GeForceRTX 4080 which has Compute Capability of 8.9, +It cannot be loaded in Windows with GeForceRTX 3080 which has Compute Capability of 8.6) + +Cross runtime compilation for windows example +====================================================== + +Compile and save the Resnet Model using Torch-TensorRT in Linux: + +python examples/dynamo/cross_runtime_compilation_for_windows.py --path trt_resnet.ep + +Load the Resnet Model saved in Windows: + +python examples/dynamo/cross_runtime_compilation_for_windows.py --path trt_resnet.ep --load True + +""" + +# %% +# Imports and Model Definition +# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +import argparse +import platform + +import torch +import torch_tensorrt as torchtrt +import torchvision.models as models + +PARSER = argparse.ArgumentParser( + description="Cross runtime comilation for windows example: Resnet Model" +) +PARSER.add_argument( + "--load", default=False, type=bool, required=False, help="Load the model in Windows" +) +PARSER.add_argument( + "--path", + type=str, + required=True, + help="Path to the saved model file", +) + +args = PARSER.parse_args() +torch.manual_seed(0) +model = models.resnet18().eval().cuda() +input = torch.rand((1, 3, 224, 224)).to("cuda") +inputs = [input] + +# %% +# According to the argument, it is either cross compile and save resnet model for windows in Linux +# or load the saved resnet model in Windows +# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +if args.load: + # load the saved model in Windows + if platform.system() != "Windows" or platform.machine() != "AMD64": + raise ValueError( + "cross runtime compiled model for windows can only be loaded in Windows system" + ) + loaded_model = torchtrt.load_cross_compiled_exported_program(args.path).module() + print(f"model has been successfully loaded from ${args.path}") + # inference + trt_output = loaded_model(input) + print(f"inference result: {trt_output}") +else: + if platform.system() != "Linux" or platform.architecture()[0] != "64bit": + raise ValueError( + "cross runtime compiled model for windows can only be compiled in Linux system" + ) + compile_spec = { + "debug": True, + "min_block_size": 1, + } + torchtrt.cross_compile_for_windows( + model, file_path=args.path, inputs=inputs, **compile_spec + ) + print( + f"model has been successfully cross compiled and saved in Linux to {args.path}" + ) diff --git a/docs/_downloads/669d3d90aba7fad1bec8bbd852aa9cbc/cross_runtime_compilation_for_windows.ipynb b/docs/_downloads/669d3d90aba7fad1bec8bbd852aa9cbc/cross_runtime_compilation_for_windows.ipynb new file mode 100644 index 0000000000..40beac7f4c --- /dev/null +++ b/docs/_downloads/669d3d90aba7fad1bec8bbd852aa9cbc/cross_runtime_compilation_for_windows.ipynb @@ -0,0 +1,68 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "\n\ncross runtime compilation limitations:\nThe cross compile and saved model can only be loaded in Windows, it can no longer be loaded in Linux\nThe cross compile and saved model can only be loaded in the same Compute Capability as the Linux which it was cross compiled\n(for example, if the model was cross compiled in Linux with GeForceRTX 4080 which has Compute Capability of 8.9,\nIt cannot be loaded in Windows with GeForceRTX 3080 which has Compute Capability of 8.6)\n\n# Cross runtime compilation for windows example\n\nCompile and save the Resnet Model using Torch-TensorRT in Linux:\n\npython examples/dynamo/cross_runtime_compilation_for_windows.py --path trt_resnet.ep\n\nLoad the Resnet Model saved in Windows:\n\npython examples/dynamo/cross_runtime_compilation_for_windows.py --path trt_resnet.ep --load True\n" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Imports and Model Definition\n\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "collapsed": false + }, + "outputs": [], + "source": [ + "import argparse\nimport platform\n\nimport torch\nimport torch_tensorrt as torchtrt\nimport torchvision.models as models\n\nPARSER = argparse.ArgumentParser(\n description=\"Cross runtime comilation for windows example: Resnet Model\"\n)\nPARSER.add_argument(\n \"--load\", default=False, type=bool, required=False, help=\"Load the model in Windows\"\n)\nPARSER.add_argument(\n \"--path\",\n type=str,\n required=True,\n help=\"Path to the saved model file\",\n)\n\nargs = PARSER.parse_args()\ntorch.manual_seed(0)\nmodel = models.resnet18().eval().cuda()\ninput = torch.rand((1, 3, 224, 224)).to(\"cuda\")\ninputs = [input]" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "According to the argument, it is either cross compile and save resnet model for windows in Linux\nor load the saved resnet model in Windows\n^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "collapsed": false + }, + "outputs": [], + "source": [ + "if args.load:\n # load the saved model in Windows\n if platform.system() != \"Windows\" or platform.machine() != \"AMD64\":\n raise ValueError(\n \"cross runtime compiled model for windows can only be loaded in Windows system\"\n )\n loaded_model = torchtrt.load_cross_compiled_exported_program(args.path).module()\n print(f\"model has been successfully loaded from ${args.path}\")\n # inference\n trt_output = loaded_model(input)\n print(f\"inference result: {trt_output}\")\nelse:\n if platform.system() != \"Linux\" or platform.architecture()[0] != \"64bit\":\n raise ValueError(\n \"cross runtime compiled model for windows can only be compiled in Linux system\"\n )\n compile_spec = {\n \"debug\": True,\n \"min_block_size\": 1,\n }\n torchtrt.cross_compile_for_windows(\n model, file_path=args.path, inputs=inputs, **compile_spec\n )\n print(\n f\"model has been successfully cross compiled and saved in Linux to {args.path}\"\n )" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 0 +} \ No newline at end of file diff --git a/docs/_downloads/6a6052d9668b2cb8332d349d328e21c1/_rendered_examples_jupyter.zip b/docs/_downloads/6a6052d9668b2cb8332d349d328e21c1/_rendered_examples_jupyter.zip index 22665d54ac..22872ae7ba 100644 Binary files a/docs/_downloads/6a6052d9668b2cb8332d349d328e21c1/_rendered_examples_jupyter.zip and b/docs/_downloads/6a6052d9668b2cb8332d349d328e21c1/_rendered_examples_jupyter.zip differ diff --git a/docs/_downloads/798cda8f83bd9f5e2cc93f329a04332c/_rendered_examples_python.zip b/docs/_downloads/798cda8f83bd9f5e2cc93f329a04332c/_rendered_examples_python.zip index 77ffc7c772..90e4179d15 100644 Binary files a/docs/_downloads/798cda8f83bd9f5e2cc93f329a04332c/_rendered_examples_python.zip and b/docs/_downloads/798cda8f83bd9f5e2cc93f329a04332c/_rendered_examples_python.zip differ diff --git a/docs/_images/sphx_glr_cross_runtime_compilation_for_windows_thumb.png b/docs/_images/sphx_glr_cross_runtime_compilation_for_windows_thumb.png new file mode 100644 index 0000000000..8a5fed589d Binary files /dev/null and b/docs/_images/sphx_glr_cross_runtime_compilation_for_windows_thumb.png differ diff --git a/docs/_modules/index.html b/docs/_modules/index.html index 7ee5938125..58b232a8ce 100644 --- a/docs/_modules/index.html +++ b/docs/_modules/index.html @@ -9,7 +9,7 @@ - Overview: module code — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Overview: module code — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/_Device.html b/docs/_modules/torch_tensorrt/_Device.html index dfd918ebc8..14aa64cfdd 100644 --- a/docs/_modules/torch_tensorrt/_Device.html +++ b/docs/_modules/torch_tensorrt/_Device.html @@ -9,7 +9,7 @@ - torch_tensorrt._Device — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt._Device — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/_Input.html b/docs/_modules/torch_tensorrt/_Input.html index 29fdc60b97..a46342ec15 100644 --- a/docs/_modules/torch_tensorrt/_Input.html +++ b/docs/_modules/torch_tensorrt/_Input.html @@ -9,7 +9,7 @@ - torch_tensorrt._Input — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt._Input — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/_compile.html b/docs/_modules/torch_tensorrt/_compile.html index 4023a6a720..4950fbd17a 100644 --- a/docs/_modules/torch_tensorrt/_compile.html +++ b/docs/_modules/torch_tensorrt/_compile.html @@ -9,7 +9,7 @@ - torch_tensorrt._compile — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt._compile — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    @@ -467,6 +467,7 @@

    Source code for torch_tensorrt._compile

     
     import collections.abc
     import logging
    +import platform
     from enum import Enum
     from typing import Any, Callable, List, Optional, Sequence, Set
     
    @@ -494,11 +495,27 @@ 

    Source code for torch_tensorrt._compile

         from torch_tensorrt.dynamo._compiler import (
             convert_exported_program_to_serialized_trt_engine as dynamo_convert_exported_program_to_serialized_trt_engine,
         )
    +    from torch_tensorrt.dynamo._compiler import (
    +        cross_compile_for_windows as dynamo_cross_compile_for_windows,
    +    )
    +    from torch_tensorrt.dynamo._compiler import (
    +        load_cross_compiled_exported_program as dynamo_load_cross_compiled_exported_program,
    +    )
    +    from torch_tensorrt.dynamo._compiler import (
    +        save_cross_compiled_exported_program as dynamo_save_cross_compiled_exported_program,
    +    )
         from torch_tensorrt.dynamo._tracer import trace as dynamo_trace
     
     logger = logging.getLogger(__name__)
     
    -__all__ = ["compile", "convert_method_to_trt_engine", "save", "load"]
    +__all__ = [
    +    "compile",
    +    "cross_compile_for_windows",
    +    "load_cross_compiled_exported_program",
    +    "convert_method_to_trt_engine",
    +    "save",
    +    "load",
    +]
     
     
     def _non_fx_input_interface(
    @@ -746,6 +763,105 @@ 

    Source code for torch_tensorrt._compile

             raise RuntimeError("Module is an unknown format or the ir requested is unknown")
    +def cross_compile_for_windows( + module: torch.nn.Module, + file_path: str, + inputs: Optional[Sequence[Input | torch.Tensor]] = None, + arg_inputs: Optional[Sequence[Sequence[Any]]] = None, + kwarg_inputs: Optional[dict[Any, Any]] = None, + enabled_precisions: Optional[Set[torch.dtype | dtype]] = None, + **kwargs: Any, +) -> None: + """Compile a PyTorch module using TensorRT in Linux for Inference in Windows + + Takes an existing PyTorch module and a set of settings to configure the compiler + and it will convert methods to AOT graphs which call equivalent TensorRT serialized + engine info into the disk in the specified file_path user provided. + It will then allow user to load the deserialized model from the disk in Windows. + Note: the model cross compiled for windows in Linux environmen can only be loaded + in Windows. + + Argument: + module (torch.nn.Module): Source module + file_path (str): the file path to store the serialized module into the disk + + Keyword Arguments: + inputs (List[Union(torch_tensorrt.Input, torch.Tensor)]): **Required** List of specifications of input shape, dtype and memory layout for inputs to the module. This argument is required. Input Sizes can be specified as torch sizes, tuples or lists. dtypes can be specified using + torch datatypes or torch_tensorrt datatypes and you can use either torch devices or the torch_tensorrt device type enum + to select device type. :: + + inputs=[ + torch_tensorrt.Input((1, 3, 224, 224)), # Static NCHW input shape for input #1 + torch_tensorrt.Input( + min_shape=(1, 224, 224, 3), + opt_shape=(1, 512, 512, 3), + max_shape=(1, 1024, 1024, 3), + dtype=torch.int32 + format=torch.channel_last + ), # Dynamic input shape for input #2 + torch.randn((1, 3, 224, 244)) # Use an example tensor and let torch_tensorrt infer settings + ] + arg_inputs (Tuple[Any, ...]): Same as inputs. Alias for better understanding with kwarg_inputs. + kwarg_inputs (dict[Any, ...]): Optional, kwarg inputs to the module forward function. + enabled_precision (Set(Union(torch.dtype, torch_tensorrt.dtype))): The set of datatypes that TensorRT can use when selecting kernels + **kwargs: Additional settings for the specific requested strategy (See submodules for more info) + + """ + + if platform.system() != "Linux" or platform.architecture()[0] != "64bit": + raise RuntimeError( + f"Cross compile for windows is only supported on x86-64 Linux architecture, current platform: {platform.system()=}, {platform.architecture()[0]=}" + ) + + if not file_path: + raise ValueError("File path cannot be empty. Please provide a valid file path") + + enabled_precisions_set: Set[dtype | torch.dtype] = ( + enabled_precisions + if enabled_precisions is not None + else _defaults.ENABLED_PRECISIONS + ) + + # Prepare torch and torchtrt inputs + if not arg_inputs and not inputs: + raise AssertionError("'arg_inputs' and 'inputs' should not both be None.") + + elif arg_inputs and inputs: + raise AssertionError( + "'arg_inputs' and 'inputs' should not be used at the same time." + ) + + arg_inputs = inputs or arg_inputs + + if kwarg_inputs is None: + kwarg_inputs = {} + + from torch_tensorrt.dynamo.utils import prepare_inputs + + if not isinstance(arg_inputs, collections.abc.Sequence): + arg_inputs = [arg_inputs] # type: ignore + + # Export the module + torchtrt_arg_inputs = prepare_inputs(arg_inputs) + torchtrt_kwarg_inputs = prepare_inputs(kwarg_inputs) + + exp_program = dynamo_trace( + module, torchtrt_arg_inputs, kwarg_inputs=torchtrt_kwarg_inputs, **kwargs + ) + logger.debug("successfully exported the module") + + # Compile and save the module + trt_gm = dynamo_cross_compile_for_windows( + exp_program, + arg_inputs=torchtrt_arg_inputs, + enabled_precisions=enabled_precisions_set, + **kwargs, + ) + + dynamo_save_cross_compiled_exported_program(trt_gm, file_path) + logger.debug("successfully compiled and saved the module for windows") + + def torch_compile(module: torch.nn.Module, **kwargs: Any) -> Any: """ Returns a boxed model which is the output of torch.compile. @@ -871,6 +987,19 @@

    Source code for torch_tensorrt._compile

             raise RuntimeError("Module is an unknown format or the ir requested is unknown")
    +def load_cross_compiled_exported_program(file_path: str = "") -> Any: + """ + Load an ExportedProgram file in Windows which was previously cross compiled in Linux + + Arguments: + file_path (str): Path to file on the disk + + Raises: + ValueError: If the api is not called in windows or there is no file or the file is not a valid ExportedProgram file + """ + return dynamo_load_cross_compiled_exported_program(file_path) + +
    [docs]def load(file_path: str = "") -> Any: """ Load either a Torchscript model or ExportedProgram. diff --git a/docs/_modules/torch_tensorrt/_enums.html b/docs/_modules/torch_tensorrt/_enums.html index e8e2c98601..d3e1b49a85 100644 --- a/docs/_modules/torch_tensorrt/_enums.html +++ b/docs/_modules/torch_tensorrt/_enums.html @@ -9,7 +9,7 @@ - torch_tensorrt._enums — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt._enums — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/dynamo/_compiler.html b/docs/_modules/torch_tensorrt/dynamo/_compiler.html index 5733e0d8bd..31bf9c7620 100644 --- a/docs/_modules/torch_tensorrt/dynamo/_compiler.html +++ b/docs/_modules/torch_tensorrt/dynamo/_compiler.html @@ -9,7 +9,7 @@ - torch_tensorrt.dynamo._compiler — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.dynamo._compiler — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    @@ -467,6 +467,7 @@

    Source code for torch_tensorrt.dynamo._compiler

    < import collections.abc import logging +import platform import warnings from typing import Any, Collection, List, Optional, Sequence, Set, Tuple, Union @@ -484,6 +485,7 @@

    Source code for torch_tensorrt.dynamo._compiler

    < parse_non_trt_nodes, ) from torch_tensorrt.dynamo._engine_cache import BaseEngineCache, DiskEngineCache +from torch_tensorrt.dynamo._exporter import replace_execute_engine_no_op_node from torch_tensorrt.dynamo.conversion import ( CompilationSettings, UnsupportedOperatorException, @@ -512,6 +514,285 @@

    Source code for torch_tensorrt.dynamo._compiler

    < logger = logging.getLogger(__name__) +def cross_compile_for_windows( + exported_program: ExportedProgram, + inputs: Optional[Sequence[Sequence[Any]]] = None, + *, + arg_inputs: Optional[Sequence[Sequence[Any]]] = None, + kwarg_inputs: Optional[dict[Any, Any]] = None, + device: Optional[Union[Device, torch.device, str]] = _defaults.DEVICE, + disable_tf32: bool = _defaults.DISABLE_TF32, + assume_dynamic_shape_support: bool = _defaults.ASSUME_DYNAMIC_SHAPE_SUPPORT, + sparse_weights: bool = _defaults.SPARSE_WEIGHTS, + enabled_precisions: Union[ + Set[Union[torch.dtype, dtype]], Tuple[Union[torch.dtype, dtype]] + ] = _defaults.ENABLED_PRECISIONS, + engine_capability: EngineCapability = _defaults.ENGINE_CAPABILITY, + make_refittable: bool = _defaults.MAKE_REFITTABLE, + debug: bool = _defaults.DEBUG, + num_avg_timing_iters: int = _defaults.NUM_AVG_TIMING_ITERS, + workspace_size: int = _defaults.WORKSPACE_SIZE, + dla_sram_size: int = _defaults.DLA_SRAM_SIZE, + dla_local_dram_size: int = _defaults.DLA_LOCAL_DRAM_SIZE, + dla_global_dram_size: int = _defaults.DLA_GLOBAL_DRAM_SIZE, + truncate_double: bool = _defaults.TRUNCATE_DOUBLE, + require_full_compilation: bool = _defaults.REQUIRE_FULL_COMPILATION, + min_block_size: int = _defaults.MIN_BLOCK_SIZE, + torch_executed_ops: Optional[Collection[Target]] = None, + torch_executed_modules: Optional[List[str]] = None, + pass_through_build_failures: bool = _defaults.PASS_THROUGH_BUILD_FAILURES, + max_aux_streams: Optional[int] = _defaults.MAX_AUX_STREAMS, + version_compatible: bool = _defaults.VERSION_COMPATIBLE, + optimization_level: Optional[int] = _defaults.OPTIMIZATION_LEVEL, + use_python_runtime: bool = _defaults.USE_PYTHON_RUNTIME, + use_fast_partitioner: bool = _defaults.USE_FAST_PARTITIONER, + enable_experimental_decompositions: bool = _defaults.ENABLE_EXPERIMENTAL_DECOMPOSITIONS, + dryrun: bool = _defaults.DRYRUN, + hardware_compatible: bool = _defaults.HARDWARE_COMPATIBLE, + timing_cache_path: str = _defaults.TIMING_CACHE_PATH, + lazy_engine_init: bool = _defaults.LAZY_ENGINE_INIT, + cache_built_engines: bool = _defaults.CACHE_BUILT_ENGINES, + reuse_cached_engines: bool = _defaults.REUSE_CACHED_ENGINES, + engine_cache_dir: str = _defaults.ENGINE_CACHE_DIR, + engine_cache_size: int = _defaults.ENGINE_CACHE_SIZE, + custom_engine_cache: Optional[BaseEngineCache] = _defaults.CUSTOM_ENGINE_CACHE, + use_explicit_typing: bool = _defaults.USE_EXPLICIT_TYPING, + use_fp32_acc: bool = _defaults.USE_FP32_ACC, + enable_weight_streaming: bool = _defaults.ENABLE_WEIGHT_STREAMING, + **kwargs: Any, +) -> torch.fx.GraphModule: + """Compile an ExportedProgram module using TensorRT in Linux for Inference in Windows + + Takes an exported program and a set of settings to configure the compiler + and it will convert methods to AOT graphs which call equivalent TensorRT engines + + Arguments: + exported_program (torch.export.ExportedProgram): Source module, running torch.export on a ``torch.nn.Module`` + inputs (Tuple[Any, ...]): List of specifications of input shape, dtype and memory layout for inputs to the module. This argument is required. Input Sizes can be specified as torch sizes, tuples or lists. dtypes can be specified using + torch datatypes or torch_tensorrt datatypes and you can use either torch devices or the torch_tensorrt device type enum + to select device type. + + .. code-block:: py + + inputs=[ + torch_tensorrt.Input((1, 3, 224, 224)), # Static NCHW input shape for input #1 + torch_tensorrt.Input( + min_shape=(1, 224, 224, 3), + opt_shape=(1, 512, 512, 3), + max_shape=(1, 1024, 1024, 3), + dtype=torch.int32 + format=torch.channel_last + ), # Dynamic input shape for input #2 + torch.randn((1, 3, 224, 244)) # Use an example tensor and let torch_tensorrt infer settings + ] + + Keyword Arguments: + arg_inputs (Tuple[Any, ...]): Same as inputs. Alias for better understanding with kwarg_inputs. + kwarg_inputs (dict[Any, ...]): Optional, kwarg inputs to the module forward function. + device (Union(torch_tensorrt.Device, torch.device, dict)): Target device for TensorRT engines to run on :: + + device=torch_tensorrt.Device("dla:1", allow_gpu_fallback=True) + + disable_tf32 (bool): Force FP32 layers to use traditional as FP32 format vs the default behavior of rounding the inputs to 10-bit mantissas before multiplying, but accumulates the sum using 23-bit mantissas + assume_dynamic_shape_support (bool): Setting this to true enables the converters work for both dynamic and static shapes. Default: False + sparse_weights (bool): Enable sparsity for convolution and fully connected layers. + enabled_precision (Set(Union(torch.dtype, torch_tensorrt.dtype))): The set of datatypes that TensorRT can use when selecting kernels + refit (bool): Enable refitting + debug (bool): Enable debuggable engine + capability (torch_tensorrt.EngineCapability): Restrict kernel selection to safe gpu kernels or safe dla kernels + num_avg_timing_iters (int): Number of averaging timing iterations used to select kernels + workspace_size (int): Maximum size of workspace given to TensorRT + dla_sram_size (int): Fast software managed RAM used by DLA to communicate within a layer. + dla_local_dram_size (int): Host RAM used by DLA to share intermediate tensor data across operations + dla_global_dram_size (int): Host RAM used by DLA to store weights and metadata for execution + truncate_double (bool): Truncate weights provided in double (float64) to float32 + calibrator (Union(torch_tensorrt._C.IInt8Calibrator, tensorrt.IInt8Calibrator)): Calibrator object which will provide data to the PTQ system for INT8 Calibration + require_full_compilation (bool): Require modules to be compiled end to end or return an error as opposed to returning a hybrid graph where operations that cannot be run in TensorRT are run in PyTorch + min_block_size (int): The minimum number of contiguous TensorRT convertible operations in order to run a set of operations in TensorRT + torch_executed_ops (Collection[Target]): Set of aten operators that must be run in PyTorch. An error will be thrown if this set is not empty but ``require_full_compilation`` is True + torch_executed_modules (List[str]): List of modules that must be run in PyTorch. An error will be thrown if this list is not empty but ``require_full_compilation`` is True + pass_through_build_failures (bool): Error out if there are issues during compilation (only applicable to torch.compile workflows) + max_aux_stream (Optional[int]): Maximum streams in the engine + version_compatible (bool): Build the TensorRT engines compatible with future versions of TensorRT (Restrict to lean runtime operators to provide version forward compatibility for the engines) + optimization_level: (Optional[int]): Setting a higher optimization level allows TensorRT to spend longer engine building time searching for more optimization options. The resulting engine may have better performance compared to an engine built with a lower optimization level. The default optimization level is 3. Valid values include integers from 0 to the maximum optimization level, which is currently 5. Setting it to be greater than the maximum level results in identical behavior to the maximum level. + use_python_runtime: (bool): Return a graph using a pure Python runtime, reduces options for serialization + use_fast_partitioner: (bool): Use the adjacency based partitioning scheme instead of the global partitioner. Adjacency partitioning is faster but may not be optimal. Use the global paritioner (``False``) if looking for best performance + enable_experimental_decompositions (bool): Use the full set of operator decompositions. These decompositions may not be tested but serve to make the graph easier to convert to TensorRT, potentially increasing the amount of graphs run in TensorRT. + dryrun (bool): Toggle for "Dryrun" mode, running everything except conversion to TRT and logging outputs + hardware_compatible (bool): Build the TensorRT engines compatible with GPU architectures other than that of the GPU on which the engine was built (currently works for NVIDIA Ampere and newer) + timing_cache_path (str): Path to the timing cache if it exists (or) where it will be saved after compilation + lazy_engine_init (bool): Defer setting up engines until the compilation of all engines is complete. Can allow larger models with multiple graph breaks to compile but can lead to oversubscription of GPU memory at runtime. + cache_built_engines (bool): Whether to save the compiled TRT engines to storage + reuse_cached_engines (bool): Whether to load the compiled TRT engines from storage + engine_cache_dir (Optional[str]): Directory to store the cached TRT engines + engine_cache_size (Optional[int]): Maximum hard-disk space (bytes) to use for the engine cache, default is 1GB. If the cache exceeds this size, the oldest engines will be removed by default + custom_engine_cache (Optional[BaseEngineCache]): Engine cache instance to use for saving and loading engines. Users can provide their own engine cache by inheriting from BaseEngineCache. If used, engine_cache_dir and engine_cache_size will be ignored. + use_explicit_typing (bool): This flag enables strong typing in TensorRT compilation which respects the precisions set in the Pytorch model. This is useful when users have mixed precision graphs. + use_fp32_acc (bool): This option inserts cast to FP32 nodes around matmul layers and TensorRT ensures the accumulation of matmul happens in FP32. Use this only when FP16 precision is configured in enabled_precisions. + enable_weight_streaming (bool): Enable weight streaming. + **kwargs: Any, + Returns: + torch.fx.GraphModule: Compiled FX Module, when run it will execute via TensorRT + + """ + if platform.system() != "Linux" or platform.architecture()[0] != "64bit": + raise RuntimeError( + f"Cross compile for windows is only supported on x86-64 Linux architecture, current platform: {platform.system()=}, {platform.architecture()[0]=}" + ) + + if debug: + set_log_level(logger.parent, logging.DEBUG) + + if "truncate_long_and_double" in kwargs.keys(): + if truncate_double is not _defaults.TRUNCATE_DOUBLE: + raise ValueError( + 'Provided configuration for "truncate_double" and deprecated API "truncate_long_and_double", please only use "truncate_double"' + ) + else: + truncate_double = kwargs["truncate_long_and_double"] + warnings.warn( + 'Compiler option "truncate_long_and_double" is deprecated in favor of "truncate_double" as int64 is now natively supported, this option will be removed in the next version', + DeprecationWarning, + stacklevel=2, + ) + + if "refit" in kwargs.keys(): + warnings.warn( + "Refit is deprecated. Please use make_refittable=True if you want to enable refitting of the engine.", + DeprecationWarning, + stacklevel=2, + ) + if make_refittable: + raise ValueError("Use flag make_refittable only. Flag refit is deprecated.") + else: + make_refittable = kwargs["refit"] + + engine_capability = EngineCapability._from(engine_capability) + + if torch_executed_modules is not None and torch_executed_modules: + logger.warning( + f"Detected torch_executed_modules was non-empty: {torch_executed_modules}" + "\nThis feature is unimplemented in Torch-TRT Dynamo currently." + ) + + if use_explicit_typing: + if len(enabled_precisions) != 1 or not any( + x in enabled_precisions for x in {torch.float32, dtype.f32} + ): + raise AssertionError( + f"When use_explicit_typing is enabled, only torch.float32 is allowed in the enabled_precisions but found {enabled_precisions}" + ) + + if use_fp32_acc: + logger.debug( + "FP32 accumulation for matmul layers is enabled. This option should only be enabled if the model already has FP16 weights and has no effect if it has FP32 weights. \ + This flag inserts casts around matmul layers and ensures TensorRT executes the matmul layers in FP16 with FP32 accumulation." + ) + + if enable_weight_streaming and not use_explicit_typing: + raise AssertionError( + "When enable_weight_streaming is enabled, it requires use_explicit_typing to be set to True" + ) + # Aliasing inputs to arg_inputs for better understanding + if not arg_inputs and not inputs: + raise AssertionError("'arg_inputs' and 'inputs' should not both be None.") + + elif arg_inputs and inputs: + raise AssertionError( + "'arg_inputs' and 'inputs' should not be used at the same time." + ) + + arg_inputs = inputs or arg_inputs + + if kwarg_inputs is None: + kwarg_inputs = {} + + if not isinstance(arg_inputs, collections.abc.Sequence): + arg_inputs = [arg_inputs] # type: ignore + + # Prepare torch_trt inputs + trt_arg_inputs: Sequence[Input] = prepare_inputs(arg_inputs) + trt_kwarg_inputs: Optional[dict[Any, Any]] = prepare_inputs(kwarg_inputs) + device = to_torch_tensorrt_device(device) + enabled_precisions = {dtype._from(p) for p in enabled_precisions} + + compilation_options = { + "enabled_precisions": ( + enabled_precisions if enabled_precisions else _defaults.ENABLED_PRECISIONS + ), + "debug": debug, + "device": device, + "assume_dynamic_shape_support": assume_dynamic_shape_support, + "workspace_size": workspace_size, + "min_block_size": min_block_size, + "torch_executed_ops": ( + torch_executed_ops if torch_executed_ops is not None else set() + ), + "pass_through_build_failures": pass_through_build_failures, + "max_aux_streams": max_aux_streams, + "version_compatible": version_compatible, + "optimization_level": optimization_level, + "use_python_runtime": False, + "truncate_double": truncate_double, + "use_fast_partitioner": use_fast_partitioner, + "num_avg_timing_iters": num_avg_timing_iters, + "enable_experimental_decompositions": enable_experimental_decompositions, + "require_full_compilation": require_full_compilation, + "disable_tf32": disable_tf32, + "sparse_weights": sparse_weights, + "make_refittable": make_refittable, + "engine_capability": engine_capability, + "dla_sram_size": dla_sram_size, + "dla_local_dram_size": dla_local_dram_size, + "dla_global_dram_size": dla_global_dram_size, + "dryrun": dryrun, + "hardware_compatible": hardware_compatible, + "timing_cache_path": timing_cache_path, + "lazy_engine_init": lazy_engine_init, + "cache_built_engines": cache_built_engines, + "reuse_cached_engines": reuse_cached_engines, + "enable_cross_compile_for_windows": True, + "enable_weight_streaming": enable_weight_streaming, + } + + # disable the following settings is not supported for cross compilation for windows feature + unsupported_settings = ( + "use_python_runtime", + "lazy_engine_init", + "cache_built_engines", + "reuse_cached_engines", + ) + # disable these settings if anything is turned on + for key, value in compilation_options.items(): + if key in unsupported_settings and value: + compilation_options[key] = False + logger.warning( + f"arg: {key} is not supported for cross compilation for windows feature, hence it is disabled." + ) + + settings = CompilationSettings(**compilation_options) + logger.info("Compilation Settings: %s\n", settings) + exported_program = pre_export_lowering(exported_program, settings) + exported_program = exported_program.run_decompositions( + get_decompositions(enable_experimental_decompositions) + ) + + gm = exported_program.module() + logger.debug("Input graph: " + str(gm.graph)) + + # Apply lowering on the graph module + gm = post_lowering(gm, settings) + logger.debug("Lowered Input graph: " + str(gm.graph)) + + trt_gm = compile_module( + gm, + trt_arg_inputs, + trt_kwarg_inputs, + settings, + ) + return trt_gm + +
    [docs]def compile( exported_program: ExportedProgram, inputs: Optional[Sequence[Sequence[Any]]] = None, @@ -661,6 +942,14 @@

    Source code for torch_tensorrt.dynamo._compiler

    < else: make_refittable = kwargs["refit"] + if ( + "enable_cross_compile_for_windows" in kwargs.keys() + and kwargs["enable_cross_compile_for_windows"] + ): + raise ValueError( + "Please use cross_compile_for_windows() api if you want to cross compile the module in Linux for inferencing in Windows." + ) + engine_capability = EngineCapability._from(engine_capability) if torch_executed_modules is not None and torch_executed_modules: @@ -763,6 +1052,7 @@

    Source code for torch_tensorrt.dynamo._compiler

    < "reuse_cached_engines": reuse_cached_engines, "use_explicit_typing": use_explicit_typing, "use_fp32_acc": use_fp32_acc, + "enable_cross_compile_for_windows": False, "enable_weight_streaming": enable_weight_streaming, } @@ -998,7 +1288,7 @@

    Source code for torch_tensorrt.dynamo._compiler

    < # Replace all FX Modules with TRT Modules for name, trt_module in trt_modules.items(): setattr(partitioned_module, name, trt_module) - if settings.lazy_engine_init: + if settings.lazy_engine_init and not settings.enable_cross_compile_for_windows: getattr(partitioned_module, name).setup_engine() # Reset settings object to user specification after fallback to global partitioning mode @@ -1226,6 +1516,61 @@

    Source code for torch_tensorrt.dynamo._compiler

    < serialized_engine: bytes = interpreter_result.serialized_engine return serialized_engine + + +def save_cross_compiled_exported_program( + gm: torch.fx.GraphModule, + file_path: str, +) -> None: + """ + Save cross compiled exported program to disk. + + Arguments: + module (torch.fx.GraphModule): Cross compiled Torch-TensorRT module + file_path (str): the file path where the exported program will be saved to disk + """ + if not file_path: + raise ValueError("File path cannot be empty. Please provide a valid file path") + + from torch_tensorrt.dynamo._exporter import export + + exp_program = export(gm, cross_compile_flag=True) + torch.export.save(exp_program, file_path) + logger.debug(f"successfully saved the module for windows at {file_path}") + + +def load_cross_compiled_exported_program(file_path: str = "") -> Any: + """ + Load an ExportedProgram file in Windows which was previously cross compiled in Linux + + Arguments: + file_path (str): Path to file on the disk + + Raises: + ValueError: If the api is not called in windows or there is no file or the file is a valid ExportedProgram file + """ + if not file_path: + raise ValueError("File path cannot be empty. Please provide a valid file path") + + if platform.system() != "Windows" or platform.machine() != "AMD64": + raise ValueError( + "cross runtime compiled model for windows can only be loaded in Windows system" + ) + + try: + logger.debug(f"Loading the provided file {file_path} using torch.export.load()") + # TODO: think about how to handle the torch.jit.load route? + exp_program = torch.export.load(file_path) + except Exception as e: + logger.info( + f"Loading the provided file {file_path} via torch.export.load() failed with the following error: {e}", + exc_info=True, + ) + raise ValueError( + f"cross_load the file {file_path} doesn't correspond to a valid ExportedProgram. Please verify the file path." + ) + + return replace_execute_engine_no_op_node(exp_program)
    diff --git a/docs/_modules/torch_tensorrt/dynamo/_exporter.html b/docs/_modules/torch_tensorrt/dynamo/_exporter.html index 5b87343c06..3c8b03d093 100644 --- a/docs/_modules/torch_tensorrt/dynamo/_exporter.html +++ b/docs/_modules/torch_tensorrt/dynamo/_exporter.html @@ -9,7 +9,7 @@ - torch_tensorrt.dynamo._exporter — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.dynamo._exporter — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    @@ -463,9 +463,10 @@

    Source code for torch_tensorrt.dynamo._exporter

    -import copy
    +import base64
    +import copy
     import operator
    -from typing import Any, Dict, Sequence, Tuple, cast
    +from typing import Any, Dict, Optional, Sequence, Tuple, cast
     
     import torch
     from torch._guards import detect_fake_mode
    @@ -481,24 +482,28 @@ 

    Source code for torch_tensorrt.dynamo._exporter

    < OutputSpec, TensorArgument, ) +from torch_tensorrt.dynamo.runtime._TorchTensorRTModule import ENGINE_IDX, NAME_IDX
    [docs]def export( gm: torch.fx.GraphModule, + cross_compile_flag: Optional[bool] = False, ) -> ExportedProgram: """Export the result of TensorRT compilation into the desired output format. Arguments: gm (torch.fx.GraphModule): Compiled Torch-TensorRT module, generated by ``torch_tensorrt.dynamo.compile`` inputs (torch.Tensor): Torch input tensors + cross_compile_flag (bool): Flag to indicated whether it is cross_compilation enabled or not """ - patched_module = transform(gm) + patched_module = transform(gm, cross_compile_flag) exp_program = create_trt_exp_program(patched_module) return exp_program
    def transform( gm: torch.fx.GraphModule, + cross_compile_flag: Optional[bool] = False, ) -> torch.fx.GraphModule: """ Transforms the graphmodule by inlining Pytorch and TensorRT submodules. @@ -508,6 +513,7 @@

    Source code for torch_tensorrt.dynamo._exporter

    < Arguments: gm (torch.fx.GraphModule): Compiled Torch-TensorRT module, generated by ``torch_tensorrt.dynamo.compile`` inputs (torch.Tensor): Torch input tensors + cross_compile_flag (bool): Flag to indicated whether it is cross_compilation enabled or not Returns an inlined torch.fx.GraphModule """ @@ -516,7 +522,7 @@

    Source code for torch_tensorrt.dynamo._exporter

    < gm = copy.deepcopy(gm) # Inline TensorRT submodules - inline_trt_modules(gm) + inline_trt_modules(gm, cross_compile_flag) # Inline pytorch submodules inline_torch_modules(gm) @@ -815,7 +821,9 @@

    Source code for torch_tensorrt.dynamo._exporter

    < return trt_exp_program -def inline_trt_modules(gm: torch.fx.GraphModule) -> torch.fx.GraphModule: +def inline_trt_modules( + gm: torch.fx.GraphModule, cross_compile_flag: Optional[bool] = False +) -> torch.fx.GraphModule: """ Replace TRT submodules with trt engine nodes. """ @@ -838,25 +846,36 @@

    Source code for torch_tensorrt.dynamo._exporter

    < num_outputs = len(trt_module_node.meta["val"]) # Insert a call_function node to perform inference on TRT engine with gm.graph.inserting_before(trt_module_node): - engine_name = f"{name}_engine" - setattr(gm, engine_name, trt_module.engine) - engine_node = gm.graph.get_attr(engine_name) - - trt_node = gm.graph.call_function( - torch.ops.tensorrt.execute_engine.default, - (trt_module_node.args, engine_node), - ) + if not cross_compile_flag: + # for the normal workflow: use the execute_engine node + engine_name = f"{name}_engine" + setattr(gm, engine_name, trt_module.engine) + engine_node = gm.graph.get_attr(engine_name) + + trt_node = gm.graph.call_function( + torch.ops.tensorrt.execute_engine.default, + (trt_module_node.args, engine_node), + ) + # meta["val"] should be a lighter version of a tensor. For eg: it should be a FakeTensor (with output shape and dtype properties) + # Lighter version of a custom_obj is not defined clearly. meta["val"] does not have any type expectations but + # for custom object nodes, it should be CustomObjArgument + engine_node.meta["val"] = CustomObjArgument( + name=engine_node.name, class_fqn="" + ) + else: + # for the cross compile for windows workflow: use the no_op_placeholder node + engine_info = trt_module._pack_engine_info() + engine_bytes = engine_info[ENGINE_IDX] + engine_info[ENGINE_IDX] = base64.b64encode(engine_bytes).decode("utf-8") + # insert the no_placeholder node in the graph which should be replaced to the actual execute_engine node while load in the windows + trt_node = gm.graph.call_function( + torch.ops.tensorrt.no_op_placeholder_for_execute_engine.default, + (trt_module_node.args, *engine_info), + ) # set trt_node.meta with trt_module_node.meta assert num_outputs > 0 trt_node.meta["val"] = trt_module_node.meta["val"] - # meta["val"] should be a lighter version of a tensor. For eg: it should be a FakeTensor (with output shape and dtype properties) - # Lighter version of a custom_obj is not defined clearly. meta["val"] does not have any type expectations but - # for custom object nodes, it should be CustomObjArgument - engine_node.meta["val"] = CustomObjArgument( - name=engine_node.name, class_fqn="" - ) - if num_outputs == 1: # Insert getitem nodes as outputs (for export serialization to work) with gm.graph.inserting_after(trt_node): @@ -876,6 +895,60 @@

    Source code for torch_tensorrt.dynamo._exporter

    < gm.graph.erase_node(trt_module_node) return gm + + +def replace_execute_engine_no_op_node( + exp_program: ExportedProgram, +) -> ExportedProgram: + gm = exp_program.graph_module + no_op_placeholder_nodes = [] + for node in gm.graph.nodes: + if "no_op_placeholder_for_execute_engine" in node.name: + no_op_placeholder_nodes.append(node) + assert len(no_op_placeholder_nodes) > 0 + for no_op_placeholder_node in no_op_placeholder_nodes: + if "val" not in no_op_placeholder_node.meta: + raise ValueError(f"metadata info is missing for the node: {node.name}") + with gm.graph.inserting_before(no_op_placeholder_node): + packed_engine_info = list(no_op_placeholder_node.args[1:]) + engine_bytes = packed_engine_info[ENGINE_IDX] + engine_name = packed_engine_info[NAME_IDX] + + packed_engine_info[ENGINE_IDX] = base64.b64decode( + engine_bytes.encode("utf-8") + ) + trt_engine = torch.classes.tensorrt.Engine(tuple(packed_engine_info)) + setattr(gm, engine_name, trt_engine) + engine_node = gm.graph.get_attr(engine_name) + + trt_node = gm.graph.call_function( + torch.ops.tensorrt.execute_engine.default, + (no_op_placeholder_node.args[0], engine_node), + ) + trt_node.meta["val"] = no_op_placeholder_node.meta["val"] + engine_node.meta["val"] = CustomObjArgument( + name=engine_node.name, class_fqn="" + ) + + if len(no_op_placeholder_node.meta["val"]) == 1: + with gm.graph.inserting_after(trt_node): + getitem_output = gm.graph.call_function(operator.getitem, (trt_node, 0)) + getitem_output.meta["val"] = trt_node.meta["val"] + no_op_placeholder_node.replace_all_uses_with(getitem_output) + else: + no_op_placeholder_node.replace_all_uses_with(trt_node) + getitem_nodes = trt_node.users + for idx, getitem_node in enumerate(getitem_nodes): + getitem_node.meta["val"] = trt_node.meta["val"][idx] + + gm.graph.erase_node(no_op_placeholder_node) + + gm.delete_all_unused_submodules() + gm.graph.eliminate_dead_code() + gm.graph.lint() + gm.recompile() + + return exp_program
    diff --git a/docs/_modules/torch_tensorrt/dynamo/_refit.html b/docs/_modules/torch_tensorrt/dynamo/_refit.html index 3cbebaf5f6..ec930e6f69 100644 --- a/docs/_modules/torch_tensorrt/dynamo/_refit.html +++ b/docs/_modules/torch_tensorrt/dynamo/_refit.html @@ -9,7 +9,7 @@ - torch_tensorrt.dynamo._refit — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.dynamo._refit — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/dynamo/_settings.html b/docs/_modules/torch_tensorrt/dynamo/_settings.html index 259e97766e..41bd62ee7e 100644 --- a/docs/_modules/torch_tensorrt/dynamo/_settings.html +++ b/docs/_modules/torch_tensorrt/dynamo/_settings.html @@ -9,7 +9,7 @@ - torch_tensorrt.dynamo._settings — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.dynamo._settings — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    @@ -478,6 +478,7 @@

    Source code for torch_tensorrt.dynamo._settings

    < DLA_LOCAL_DRAM_SIZE, DLA_SRAM_SIZE, DRYRUN, + ENABLE_CROSS_COMPILE_FOR_WINDOWS, ENABLE_EXPERIMENTAL_DECOMPOSITIONS, ENABLE_WEIGHT_STREAMING, ENABLED_PRECISIONS, @@ -549,6 +550,8 @@

    Source code for torch_tensorrt.dynamo._settings

    < use_strong_typing (bool): This flag enables strong typing in TensorRT compilation which respects the precisions set in the Pytorch model. This is useful when users have mixed precision graphs. use_fp32_acc (bool): This option inserts cast to FP32 nodes around matmul layers and TensorRT ensures the accumulation of matmul happens in FP32. Use this only when FP16 precision is configured in enabled_precisions. enable_weight_streaming (bool): Enable weight streaming. + enable_cross_compile_for_windows (bool): By default this is False means TensorRT engines can only be executed on the same platform where they were built. + True will enable cross-platform compatibility which allows the engine to be built on Linux and run on Windows """ enabled_precisions: Set[dtype] = field(default_factory=lambda: ENABLED_PRECISIONS) @@ -585,7 +588,8 @@

    Source code for torch_tensorrt.dynamo._settings

    < reuse_cached_engines: bool = REUSE_CACHED_ENGINES use_explicit_typing: bool = USE_EXPLICIT_TYPING use_fp32_acc: bool = USE_FP32_ACC - enable_weight_streaming: bool = ENABLE_WEIGHT_STREAMING
    + enable_weight_streaming: bool = ENABLE_WEIGHT_STREAMING + enable_cross_compile_for_windows: bool = ENABLE_CROSS_COMPILE_FOR_WINDOWS
    _SETTINGS_TO_BE_ENGINE_INVARIANT = ( diff --git a/docs/_modules/torch_tensorrt/dynamo/_tracer.html b/docs/_modules/torch_tensorrt/dynamo/_tracer.html index ba9e41c6b8..4b2f618d2e 100644 --- a/docs/_modules/torch_tensorrt/dynamo/_tracer.html +++ b/docs/_modules/torch_tensorrt/dynamo/_tracer.html @@ -9,7 +9,7 @@ - torch_tensorrt.dynamo._tracer — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.dynamo._tracer — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/dynamo/runtime/_MutableTorchTensorRTModule.html b/docs/_modules/torch_tensorrt/dynamo/runtime/_MutableTorchTensorRTModule.html index 28c3a44bd4..6c29913343 100644 --- a/docs/_modules/torch_tensorrt/dynamo/runtime/_MutableTorchTensorRTModule.html +++ b/docs/_modules/torch_tensorrt/dynamo/runtime/_MutableTorchTensorRTModule.html @@ -9,7 +9,7 @@ - torch_tensorrt.dynamo.runtime._MutableTorchTensorRTModule — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.dynamo.runtime._MutableTorchTensorRTModule — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/dynamo/runtime/_PythonTorchTensorRTModule.html b/docs/_modules/torch_tensorrt/dynamo/runtime/_PythonTorchTensorRTModule.html index e421038739..a9f03017b1 100644 --- a/docs/_modules/torch_tensorrt/dynamo/runtime/_PythonTorchTensorRTModule.html +++ b/docs/_modules/torch_tensorrt/dynamo/runtime/_PythonTorchTensorRTModule.html @@ -9,7 +9,7 @@ - torch_tensorrt.dynamo.runtime._PythonTorchTensorRTModule — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.dynamo.runtime._PythonTorchTensorRTModule — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/dynamo/runtime/_TorchTensorRTModule.html b/docs/_modules/torch_tensorrt/dynamo/runtime/_TorchTensorRTModule.html index fe880a7d17..6519d9a754 100644 --- a/docs/_modules/torch_tensorrt/dynamo/runtime/_TorchTensorRTModule.html +++ b/docs/_modules/torch_tensorrt/dynamo/runtime/_TorchTensorRTModule.html @@ -9,7 +9,7 @@ - torch_tensorrt.dynamo.runtime._TorchTensorRTModule — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.dynamo.runtime._TorchTensorRTModule — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    @@ -597,7 +597,11 @@

    Source code for torch_tensorrt.dynamo.runtime._TorchTensorRTModule

    self.serialized_engine = serialized_engine self.engine = None - if serialized_engine and not self.settings.lazy_engine_init: + if ( + serialized_engine + and not self.settings.lazy_engine_init + and not self.settings.enable_cross_compile_for_windows + ): self.setup_engine() def _pack_engine_info(self) -> List[str | bytes]: @@ -609,16 +613,16 @@

    Source code for torch_tensorrt.dynamo.runtime._TorchTensorRTModule

    metadata = {"settings": self.settings, "weight_name_map": self.weight_name_map} target_platform = ( Platform.current_platform() + if not self.settings.enable_cross_compile_for_windows + else Platform.WIN_X86_64 ) # Change to match target for engine engine_info: List[str | bytes] = [""] * SERIALIZATION_LEN - engine_info[ABI_TARGET_IDX] = torch.ops.tensorrt.ABI_VERSION() engine_info[NAME_IDX] = ( self.name + "_engine" if self.name != "" else "tensorrt_engine" ) engine_info[DEVICE_IDX] = target_device._to_serialized_rt_device() - assert self.serialized_engine engine_info[ENGINE_IDX] = self.serialized_engine diff --git a/docs/_modules/torch_tensorrt/fx/fx2trt.html b/docs/_modules/torch_tensorrt/fx/fx2trt.html index aa21dce754..b6de4cde0b 100644 --- a/docs/_modules/torch_tensorrt/fx/fx2trt.html +++ b/docs/_modules/torch_tensorrt/fx/fx2trt.html @@ -9,7 +9,7 @@ - torch_tensorrt.fx.fx2trt — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.fx.fx2trt — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/fx/input_tensor_spec.html b/docs/_modules/torch_tensorrt/fx/input_tensor_spec.html index 720eadd8b2..30dd39a936 100644 --- a/docs/_modules/torch_tensorrt/fx/input_tensor_spec.html +++ b/docs/_modules/torch_tensorrt/fx/input_tensor_spec.html @@ -9,7 +9,7 @@ - torch_tensorrt.fx.input_tensor_spec — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.fx.input_tensor_spec — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/fx/lower.html b/docs/_modules/torch_tensorrt/fx/lower.html index 9b97a502c8..03d4d0ae65 100644 --- a/docs/_modules/torch_tensorrt/fx/lower.html +++ b/docs/_modules/torch_tensorrt/fx/lower.html @@ -9,7 +9,7 @@ - torch_tensorrt.fx.lower — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.fx.lower — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/fx/trt_module.html b/docs/_modules/torch_tensorrt/fx/trt_module.html index b0ee9a0e1f..f7a57b18b9 100644 --- a/docs/_modules/torch_tensorrt/fx/trt_module.html +++ b/docs/_modules/torch_tensorrt/fx/trt_module.html @@ -9,7 +9,7 @@ - torch_tensorrt.fx.trt_module — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.fx.trt_module — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/logging.html b/docs/_modules/torch_tensorrt/logging.html index aa798e0766..977301159b 100644 --- a/docs/_modules/torch_tensorrt/logging.html +++ b/docs/_modules/torch_tensorrt/logging.html @@ -9,7 +9,7 @@ - torch_tensorrt.logging — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.logging — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/runtime/_multi_device_safe_mode.html b/docs/_modules/torch_tensorrt/runtime/_multi_device_safe_mode.html index 33183066b9..c9ff854d62 100644 --- a/docs/_modules/torch_tensorrt/runtime/_multi_device_safe_mode.html +++ b/docs/_modules/torch_tensorrt/runtime/_multi_device_safe_mode.html @@ -9,7 +9,7 @@ - torch_tensorrt.runtime._multi_device_safe_mode — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.runtime._multi_device_safe_mode — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/ts/_compile_spec.html b/docs/_modules/torch_tensorrt/ts/_compile_spec.html index a613defb91..0ca5ae7d64 100644 --- a/docs/_modules/torch_tensorrt/ts/_compile_spec.html +++ b/docs/_modules/torch_tensorrt/ts/_compile_spec.html @@ -9,7 +9,7 @@ - torch_tensorrt.ts._compile_spec — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.ts._compile_spec — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/ts/_compiler.html b/docs/_modules/torch_tensorrt/ts/_compiler.html index 2a4988ad2e..8ec9e198d9 100644 --- a/docs/_modules/torch_tensorrt/ts/_compiler.html +++ b/docs/_modules/torch_tensorrt/ts/_compiler.html @@ -9,7 +9,7 @@ - torch_tensorrt.ts._compiler — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.ts._compiler — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_modules/torch_tensorrt/ts/ptq.html b/docs/_modules/torch_tensorrt/ts/ptq.html index df13b1409d..7f3e59fbfc 100644 --- a/docs/_modules/torch_tensorrt/ts/ptq.html +++ b/docs/_modules/torch_tensorrt/ts/ptq.html @@ -9,7 +9,7 @@ - torch_tensorrt.ts.ptq — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.ts.ptq — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/_sources/tutorials/_rendered_examples/dynamo/cross_runtime_compilation_for_windows.rst.txt b/docs/_sources/tutorials/_rendered_examples/dynamo/cross_runtime_compilation_for_windows.rst.txt new file mode 100644 index 0000000000..dfc8544c0c --- /dev/null +++ b/docs/_sources/tutorials/_rendered_examples/dynamo/cross_runtime_compilation_for_windows.rst.txt @@ -0,0 +1,142 @@ + +.. DO NOT EDIT. +.. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY. +.. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE: +.. "tutorials/_rendered_examples/dynamo/cross_runtime_compilation_for_windows.py" +.. LINE NUMBERS ARE GIVEN BELOW. + +.. only:: html + + .. note:: + :class: sphx-glr-download-link-note + + :ref:`Go to the end ` + to download the full example code + +.. rst-class:: sphx-glr-example-title + +.. _sphx_glr_tutorials__rendered_examples_dynamo_cross_runtime_compilation_for_windows.py: + + +.. _resnet_cross_runtime_compilation_for_windows_example: + +cross runtime compilation limitations: +The cross compile and saved model can only be loaded in Windows, it can no longer be loaded in Linux +The cross compile and saved model can only be loaded in the same Compute Capability as the Linux which it was cross compiled +(for example, if the model was cross compiled in Linux with GeForceRTX 4080 which has Compute Capability of 8.9, +It cannot be loaded in Windows with GeForceRTX 3080 which has Compute Capability of 8.6) + +Cross runtime compilation for windows example +====================================================== + +Compile and save the Resnet Model using Torch-TensorRT in Linux: + +python examples/dynamo/cross_runtime_compilation_for_windows.py --path trt_resnet.ep + +Load the Resnet Model saved in Windows: + +python examples/dynamo/cross_runtime_compilation_for_windows.py --path trt_resnet.ep --load True + +.. GENERATED FROM PYTHON SOURCE LINES 24-26 + +Imports and Model Definition +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +.. GENERATED FROM PYTHON SOURCE LINES 26-53 + +.. code-block:: python + + + import argparse + import platform + + import torch + import torch_tensorrt as torchtrt + import torchvision.models as models + + PARSER = argparse.ArgumentParser( + description="Cross runtime comilation for windows example: Resnet Model" + ) + PARSER.add_argument( + "--load", default=False, type=bool, required=False, help="Load the model in Windows" + ) + PARSER.add_argument( + "--path", + type=str, + required=True, + help="Path to the saved model file", + ) + + args = PARSER.parse_args() + torch.manual_seed(0) + model = models.resnet18().eval().cuda() + input = torch.rand((1, 3, 224, 224)).to("cuda") + inputs = [input] + + +.. GENERATED FROM PYTHON SOURCE LINES 54-57 + +According to the argument, it is either cross compile and save resnet model for windows in Linux +or load the saved resnet model in Windows +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +.. GENERATED FROM PYTHON SOURCE LINES 57-83 + +.. code-block:: python + + if args.load: + # load the saved model in Windows + if platform.system() != "Windows" or platform.machine() != "AMD64": + raise ValueError( + "cross runtime compiled model for windows can only be loaded in Windows system" + ) + loaded_model = torchtrt.load_cross_compiled_exported_program(args.path).module() + print(f"model has been successfully loaded from ${args.path}") + # inference + trt_output = loaded_model(input) + print(f"inference result: {trt_output}") + else: + if platform.system() != "Linux" or platform.architecture()[0] != "64bit": + raise ValueError( + "cross runtime compiled model for windows can only be compiled in Linux system" + ) + compile_spec = { + "debug": True, + "min_block_size": 1, + } + torchtrt.cross_compile_for_windows( + model, file_path=args.path, inputs=inputs, **compile_spec + ) + print( + f"model has been successfully cross compiled and saved in Linux to {args.path}" + ) + + +.. rst-class:: sphx-glr-timing + + **Total running time of the script:** ( 0 minutes 0.000 seconds) + + +.. _sphx_glr_download_tutorials__rendered_examples_dynamo_cross_runtime_compilation_for_windows.py: + +.. only:: html + + .. container:: sphx-glr-footer sphx-glr-footer-example + + + + + .. container:: sphx-glr-download sphx-glr-download-python + + :download:`Download Python source code: cross_runtime_compilation_for_windows.py ` + + .. container:: sphx-glr-download sphx-glr-download-jupyter + + :download:`Download Jupyter notebook: cross_runtime_compilation_for_windows.ipynb ` + + +.. only:: html + + .. rst-class:: sphx-glr-signature + + `Gallery generated by Sphinx-Gallery `_ diff --git a/docs/_sources/tutorials/_rendered_examples/dynamo/index.rst.txt b/docs/_sources/tutorials/_rendered_examples/dynamo/index.rst.txt index d3f55ac92c..da129a36fc 100644 --- a/docs/_sources/tutorials/_rendered_examples/dynamo/index.rst.txt +++ b/docs/_sources/tutorials/_rendered_examples/dynamo/index.rst.txt @@ -64,6 +64,23 @@ Model Zoo
    +.. raw:: html + +
    + +.. only:: html + + .. image:: /tutorials/_rendered_examples/dynamo/images/thumb/sphx_glr_cross_runtime_compilation_for_windows_thumb.png + :alt: + + :ref:`sphx_glr_tutorials__rendered_examples_dynamo_cross_runtime_compilation_for_windows.py` + +.. raw:: html + +
    cross runtime compilation limitations:
    +
    + + .. raw:: html
    @@ -295,6 +312,7 @@ Model Zoo /tutorials/_rendered_examples/dynamo/torch_compile_stable_diffusion /tutorials/_rendered_examples/dynamo/torch_export_cudagraphs + /tutorials/_rendered_examples/dynamo/cross_runtime_compilation_for_windows /tutorials/_rendered_examples/dynamo/refit_engine_example /tutorials/_rendered_examples/dynamo/torch_compile_transformers_example /tutorials/_rendered_examples/dynamo/torch_compile_advanced_usage diff --git a/docs/_sources/tutorials/_rendered_examples/index.rst.txt b/docs/_sources/tutorials/_rendered_examples/index.rst.txt index 6a994d6a40..c84129b82d 100644 --- a/docs/_sources/tutorials/_rendered_examples/index.rst.txt +++ b/docs/_sources/tutorials/_rendered_examples/index.rst.txt @@ -76,6 +76,23 @@ Model Zoo
    +.. raw:: html + +
    + +.. only:: html + + .. image:: /tutorials/_rendered_examples/dynamo/images/thumb/sphx_glr_cross_runtime_compilation_for_windows_thumb.png + :alt: + + :ref:`sphx_glr_tutorials__rendered_examples_dynamo_cross_runtime_compilation_for_windows.py` + +.. raw:: html + +
    cross runtime compilation limitations:
    +
    + + .. raw:: html
    diff --git a/docs/_static/documentation_options.js b/docs/_static/documentation_options.js index fab9dcff18..aea1c97b9a 100644 --- a/docs/_static/documentation_options.js +++ b/docs/_static/documentation_options.js @@ -1,6 +1,6 @@ var DOCUMENTATION_OPTIONS = { URL_ROOT: document.getElementById("documentation_options").getAttribute('data-url_root'), - VERSION: 'v2.6.0.dev0+e43833d', + VERSION: 'v2.6.0.dev0+bc95015', LANGUAGE: 'en', COLLAPSE_INDEX: false, BUILDER: 'html', diff --git a/docs/cli/torchtrtc.html b/docs/cli/torchtrtc.html index 823baebf5c..dee9d2ac08 100644 --- a/docs/cli/torchtrtc.html +++ b/docs/cli/torchtrtc.html @@ -10,7 +10,7 @@ - torchtrtc — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torchtrtc — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/contributors/conversion.html b/docs/contributors/conversion.html index ffedb16e73..7b2a125b58 100644 --- a/docs/contributors/conversion.html +++ b/docs/contributors/conversion.html @@ -10,7 +10,7 @@ - Conversion Phase — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Conversion Phase — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/contributors/dynamo_converters.html b/docs/contributors/dynamo_converters.html index fa5359a3da..2f8cf9cf48 100644 --- a/docs/contributors/dynamo_converters.html +++ b/docs/contributors/dynamo_converters.html @@ -10,7 +10,7 @@ - Writing Dynamo Converters — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Writing Dynamo Converters — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/contributors/lowering.html b/docs/contributors/lowering.html index ce709a213c..b25bbd1410 100644 --- a/docs/contributors/lowering.html +++ b/docs/contributors/lowering.html @@ -10,7 +10,7 @@ - Lowering Phase — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Lowering Phase — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/contributors/partitioning.html b/docs/contributors/partitioning.html index 962b4e70b3..ace43e223b 100644 --- a/docs/contributors/partitioning.html +++ b/docs/contributors/partitioning.html @@ -10,7 +10,7 @@ - Partitioning Phase — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Partitioning Phase — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/contributors/phases.html b/docs/contributors/phases.html index cb31b0631f..7fce911f26 100644 --- a/docs/contributors/phases.html +++ b/docs/contributors/phases.html @@ -10,7 +10,7 @@ - Compiler Phases — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Compiler Phases — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/contributors/runtime.html b/docs/contributors/runtime.html index 5c496df381..30a71c3a91 100644 --- a/docs/contributors/runtime.html +++ b/docs/contributors/runtime.html @@ -10,7 +10,7 @@ - Runtime Phase — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Runtime Phase — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/contributors/system_overview.html b/docs/contributors/system_overview.html index ca4b7ec1e1..9eefbca1f3 100644 --- a/docs/contributors/system_overview.html +++ b/docs/contributors/system_overview.html @@ -10,7 +10,7 @@ - System Overview — Torch-TensorRT v2.6.0.dev0+e43833d documentation + System Overview — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/contributors/ts_converters.html b/docs/contributors/ts_converters.html index d7c28adbe9..7d66f772d9 100644 --- a/docs/contributors/ts_converters.html +++ b/docs/contributors/ts_converters.html @@ -10,7 +10,7 @@ - Writing TorchScript Converters — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Writing TorchScript Converters — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/contributors/useful_links.html b/docs/contributors/useful_links.html index 0b9cd6a1dd..ffe89b0070 100644 --- a/docs/contributors/useful_links.html +++ b/docs/contributors/useful_links.html @@ -10,7 +10,7 @@ - Useful Links for Torch-TensorRT Development — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Useful Links for Torch-TensorRT Development — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/contributors/writing_dynamo_aten_lowering_passes.html b/docs/contributors/writing_dynamo_aten_lowering_passes.html index 134cec11c0..41fdacdc86 100644 --- a/docs/contributors/writing_dynamo_aten_lowering_passes.html +++ b/docs/contributors/writing_dynamo_aten_lowering_passes.html @@ -10,7 +10,7 @@ - Writing Dynamo ATen Lowering Passes — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Writing Dynamo ATen Lowering Passes — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/dynamo/dynamo_export.html b/docs/dynamo/dynamo_export.html index d6a79b8697..22bc36f2f4 100644 --- a/docs/dynamo/dynamo_export.html +++ b/docs/dynamo/dynamo_export.html @@ -10,7 +10,7 @@ - Compiling Exported Programs with Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Compiling Exported Programs with Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/dynamo/torch_compile.html b/docs/dynamo/torch_compile.html index 63703b6cf3..dc076f8a48 100644 --- a/docs/dynamo/torch_compile.html +++ b/docs/dynamo/torch_compile.html @@ -10,7 +10,7 @@ - TensorRT Backend for torch.compile — Torch-TensorRT v2.6.0.dev0+e43833d documentation + TensorRT Backend for torch.compile — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    @@ -488,7 +488,7 @@

    Key Features¶

    -class torch_tensorrt.dynamo.CompilationSettings(enabled_precisions: ~typing.Set[~torch_tensorrt._enums.dtype] = <factory>, debug: bool = False, workspace_size: int = 0, min_block_size: int = 5, torch_executed_ops: ~typing.Collection[~typing.Union[~typing.Callable[[...], ~typing.Any], str]] = <factory>, pass_through_build_failures: bool = False, max_aux_streams: ~typing.Optional[int] = None, version_compatible: bool = False, optimization_level: ~typing.Optional[int] = None, use_python_runtime: ~typing.Optional[bool] = False, truncate_double: bool = False, use_fast_partitioner: bool = True, enable_experimental_decompositions: bool = False, device: ~torch_tensorrt._Device.Device = <factory>, require_full_compilation: bool = False, disable_tf32: bool = False, assume_dynamic_shape_support: bool = False, sparse_weights: bool = False, make_refittable: bool = False, engine_capability: ~torch_tensorrt._enums.EngineCapability = <factory>, num_avg_timing_iters: int = 1, dla_sram_size: int = 1048576, dla_local_dram_size: int = 1073741824, dla_global_dram_size: int = 536870912, dryrun: ~typing.Union[bool, str] = False, hardware_compatible: bool = False, timing_cache_path: str = '/tmp/torch_tensorrt_engine_cache/timing_cache.bin', lazy_engine_init: bool = False, cache_built_engines: bool = False, reuse_cached_engines: bool = False, use_explicit_typing: bool = False, use_fp32_acc: bool = False, enable_weight_streaming: bool = False)[source]¶
    +class torch_tensorrt.dynamo.CompilationSettings(enabled_precisions: ~typing.Set[~torch_tensorrt._enums.dtype] = <factory>, debug: bool = False, workspace_size: int = 0, min_block_size: int = 5, torch_executed_ops: ~typing.Collection[~typing.Union[~typing.Callable[[...], ~typing.Any], str]] = <factory>, pass_through_build_failures: bool = False, max_aux_streams: ~typing.Optional[int] = None, version_compatible: bool = False, optimization_level: ~typing.Optional[int] = None, use_python_runtime: ~typing.Optional[bool] = False, truncate_double: bool = False, use_fast_partitioner: bool = True, enable_experimental_decompositions: bool = False, device: ~torch_tensorrt._Device.Device = <factory>, require_full_compilation: bool = False, disable_tf32: bool = False, assume_dynamic_shape_support: bool = False, sparse_weights: bool = False, make_refittable: bool = False, engine_capability: ~torch_tensorrt._enums.EngineCapability = <factory>, num_avg_timing_iters: int = 1, dla_sram_size: int = 1048576, dla_local_dram_size: int = 1073741824, dla_global_dram_size: int = 536870912, dryrun: ~typing.Union[bool, str] = False, hardware_compatible: bool = False, timing_cache_path: str = '/tmp/torch_tensorrt_engine_cache/timing_cache.bin', lazy_engine_init: bool = False, cache_built_engines: bool = False, reuse_cached_engines: bool = False, use_explicit_typing: bool = False, use_fp32_acc: bool = False, enable_weight_streaming: bool = False, enable_cross_compile_for_windows: bool = False)[source]¶

    Compilation settings for Torch-TensorRT Dynamo Paths

    Parameters
    @@ -532,6 +532,8 @@

    Customizable Settings

    use_strong_typing (bool) – This flag enables strong typing in TensorRT compilation which respects the precisions set in the Pytorch model. This is useful when users have mixed precision graphs.

  • use_fp32_acc (bool) – This option inserts cast to FP32 nodes around matmul layers and TensorRT ensures the accumulation of matmul happens in FP32. Use this only when FP16 precision is configured in enabled_precisions.

  • enable_weight_streaming (bool) – Enable weight streaming.

  • +
  • enable_cross_compile_for_windows (bool) – By default this is False means TensorRT engines can only be executed on the same platform where they were built. +True will enable cross-platform compatibility which allows the engine to be built on Linux and run on Windows

  • diff --git a/docs/fx/getting_started_with_fx_path.html b/docs/fx/getting_started_with_fx_path.html index 3c66c40926..3c7c15cace 100644 --- a/docs/fx/getting_started_with_fx_path.html +++ b/docs/fx/getting_started_with_fx_path.html @@ -10,7 +10,7 @@ - Torch-TensorRT (FX Frontend) User Guide — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Torch-TensorRT (FX Frontend) User Guide — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/genindex.html b/docs/genindex.html index aa28b75f87..8b3bf23186 100644 --- a/docs/genindex.html +++ b/docs/genindex.html @@ -9,7 +9,7 @@ - Index — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Index — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/getting_started/installation.html b/docs/getting_started/installation.html index 632895bed3..9350c63fcc 100644 --- a/docs/getting_started/installation.html +++ b/docs/getting_started/installation.html @@ -10,7 +10,7 @@ - Installation — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Installation — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/getting_started/jetpack.html b/docs/getting_started/jetpack.html index b035f01dfc..83c31f2a35 100644 --- a/docs/getting_started/jetpack.html +++ b/docs/getting_started/jetpack.html @@ -10,7 +10,7 @@ - Overview — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Overview — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/getting_started/quick_start.html b/docs/getting_started/quick_start.html index fcdb0d4660..b44b907fee 100644 --- a/docs/getting_started/quick_start.html +++ b/docs/getting_started/quick_start.html @@ -10,7 +10,7 @@ - Quick Start — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Quick Start — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/index.html b/docs/index.html index fbec4afbb8..b70f5cb06b 100644 --- a/docs/index.html +++ b/docs/index.html @@ -10,7 +10,7 @@ - Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -274,7 +274,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/indices/supported_ops.html b/docs/indices/supported_ops.html index ec5a424566..63db9cdba4 100644 --- a/docs/indices/supported_ops.html +++ b/docs/indices/supported_ops.html @@ -10,7 +10,7 @@ - Operators Supported — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Operators Supported — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -274,7 +274,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/objects.inv b/docs/objects.inv index 17390e6a73..22a52045c1 100644 Binary files a/docs/objects.inv and b/docs/objects.inv differ diff --git a/docs/py-modindex.html b/docs/py-modindex.html index 1510b93d70..336d62b086 100644 --- a/docs/py-modindex.html +++ b/docs/py-modindex.html @@ -9,7 +9,7 @@ - Python Module Index — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Python Module Index — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/py_api/dynamo.html b/docs/py_api/dynamo.html index 19d038712f..462e46af1e 100644 --- a/docs/py_api/dynamo.html +++ b/docs/py_api/dynamo.html @@ -10,7 +10,7 @@ - torch_tensorrt.dynamo — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.dynamo — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    @@ -617,13 +617,14 @@

    Functions
    -torch_tensorrt.dynamo.export(gm: GraphModule) ExportedProgram[source]¶
    +torch_tensorrt.dynamo.export(gm: GraphModule, cross_compile_flag: Optional[bool] = False) ExportedProgram[source]¶

    Export the result of TensorRT compilation into the desired output format.

    Parameters
    • gm (torch.fx.GraphModule) – Compiled Torch-TensorRT module, generated by torch_tensorrt.dynamo.compile

    • inputs (torch.Tensor) – Torch input tensors

    • +
    • cross_compile_flag (bool) – Flag to indicated whether it is cross_compilation enabled or not

    @@ -656,7 +657,7 @@

    Functions¶

    -class torch_tensorrt.dynamo.CompilationSettings(enabled_precisions: ~typing.Set[~torch_tensorrt._enums.dtype] = <factory>, debug: bool = False, workspace_size: int = 0, min_block_size: int = 5, torch_executed_ops: ~typing.Collection[~typing.Union[~typing.Callable[[...], ~typing.Any], str]] = <factory>, pass_through_build_failures: bool = False, max_aux_streams: ~typing.Optional[int] = None, version_compatible: bool = False, optimization_level: ~typing.Optional[int] = None, use_python_runtime: ~typing.Optional[bool] = False, truncate_double: bool = False, use_fast_partitioner: bool = True, enable_experimental_decompositions: bool = False, device: ~torch_tensorrt._Device.Device = <factory>, require_full_compilation: bool = False, disable_tf32: bool = False, assume_dynamic_shape_support: bool = False, sparse_weights: bool = False, make_refittable: bool = False, engine_capability: ~torch_tensorrt._enums.EngineCapability = <factory>, num_avg_timing_iters: int = 1, dla_sram_size: int = 1048576, dla_local_dram_size: int = 1073741824, dla_global_dram_size: int = 536870912, dryrun: ~typing.Union[bool, str] = False, hardware_compatible: bool = False, timing_cache_path: str = '/tmp/torch_tensorrt_engine_cache/timing_cache.bin', lazy_engine_init: bool = False, cache_built_engines: bool = False, reuse_cached_engines: bool = False, use_explicit_typing: bool = False, use_fp32_acc: bool = False, enable_weight_streaming: bool = False)[source]¶
    +class torch_tensorrt.dynamo.CompilationSettings(enabled_precisions: ~typing.Set[~torch_tensorrt._enums.dtype] = <factory>, debug: bool = False, workspace_size: int = 0, min_block_size: int = 5, torch_executed_ops: ~typing.Collection[~typing.Union[~typing.Callable[[...], ~typing.Any], str]] = <factory>, pass_through_build_failures: bool = False, max_aux_streams: ~typing.Optional[int] = None, version_compatible: bool = False, optimization_level: ~typing.Optional[int] = None, use_python_runtime: ~typing.Optional[bool] = False, truncate_double: bool = False, use_fast_partitioner: bool = True, enable_experimental_decompositions: bool = False, device: ~torch_tensorrt._Device.Device = <factory>, require_full_compilation: bool = False, disable_tf32: bool = False, assume_dynamic_shape_support: bool = False, sparse_weights: bool = False, make_refittable: bool = False, engine_capability: ~torch_tensorrt._enums.EngineCapability = <factory>, num_avg_timing_iters: int = 1, dla_sram_size: int = 1048576, dla_local_dram_size: int = 1073741824, dla_global_dram_size: int = 536870912, dryrun: ~typing.Union[bool, str] = False, hardware_compatible: bool = False, timing_cache_path: str = '/tmp/torch_tensorrt_engine_cache/timing_cache.bin', lazy_engine_init: bool = False, cache_built_engines: bool = False, reuse_cached_engines: bool = False, use_explicit_typing: bool = False, use_fp32_acc: bool = False, enable_weight_streaming: bool = False, enable_cross_compile_for_windows: bool = False)[source]¶

    Compilation settings for Torch-TensorRT Dynamo Paths

    Parameters
    @@ -700,6 +701,8 @@

    Classes - torch_tensorrt.fx — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.fx — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/py_api/logging.html b/docs/py_api/logging.html index a40cca11ef..668ac2ff8c 100644 --- a/docs/py_api/logging.html +++ b/docs/py_api/logging.html @@ -10,7 +10,7 @@ - torch_tensorrt.logging — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.logging — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/py_api/ptq.html b/docs/py_api/ptq.html index 4493c11807..573eb71f00 100644 --- a/docs/py_api/ptq.html +++ b/docs/py_api/ptq.html @@ -10,7 +10,7 @@ - torch_tensorrt.ts.ptq — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.ts.ptq — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/py_api/runtime.html b/docs/py_api/runtime.html index 03a6ac5b2d..ad57bfcc37 100644 --- a/docs/py_api/runtime.html +++ b/docs/py_api/runtime.html @@ -10,7 +10,7 @@ - torch_tensorrt.runtime — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.runtime — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    @@ -577,13 +577,13 @@

    Classes
    -class torch_tensorrt.runtime.PythonTorchTensorRTModule(serialized_engine: ~typing.Optional[bytes] = None, input_binding_names: ~typing.Optional[~typing.List[str]] = None, output_binding_names: ~typing.Optional[~typing.List[str]] = None, *, name: str = '', settings: ~torch_tensorrt.dynamo._settings.CompilationSettings = CompilationSettings(enabled_precisions={<dtype.f32: 7>}, debug=False, workspace_size=0, min_block_size=5, torch_executed_ops=set(), pass_through_build_failures=False, max_aux_streams=None, version_compatible=False, optimization_level=None, use_python_runtime=False, truncate_double=False, use_fast_partitioner=True, enable_experimental_decompositions=False, device=Device(type=DeviceType.GPU, gpu_id=0), require_full_compilation=False, disable_tf32=False, assume_dynamic_shape_support=False, sparse_weights=False, make_refittable=False, engine_capability=<EngineCapability.STANDARD: 1>, num_avg_timing_iters=1, dla_sram_size=1048576, dla_local_dram_size=1073741824, dla_global_dram_size=536870912, dryrun=False, hardware_compatible=False, timing_cache_path='/tmp/torch_tensorrt_engine_cache/timing_cache.bin', lazy_engine_init=False, cache_built_engines=False, reuse_cached_engines=False, use_explicit_typing=False, use_fp32_acc=False, enable_weight_streaming=False), weight_name_map: ~typing.Any = None)[source]¶
    +class torch_tensorrt.runtime.PythonTorchTensorRTModule(serialized_engine: ~typing.Optional[bytes] = None, input_binding_names: ~typing.Optional[~typing.List[str]] = None, output_binding_names: ~typing.Optional[~typing.List[str]] = None, *, name: str = '', settings: ~torch_tensorrt.dynamo._settings.CompilationSettings = CompilationSettings(enabled_precisions={<dtype.f32: 7>}, debug=False, workspace_size=0, min_block_size=5, torch_executed_ops=set(), pass_through_build_failures=False, max_aux_streams=None, version_compatible=False, optimization_level=None, use_python_runtime=False, truncate_double=False, use_fast_partitioner=True, enable_experimental_decompositions=False, device=Device(type=DeviceType.GPU, gpu_id=0), require_full_compilation=False, disable_tf32=False, assume_dynamic_shape_support=False, sparse_weights=False, make_refittable=False, engine_capability=<EngineCapability.STANDARD: 1>, num_avg_timing_iters=1, dla_sram_size=1048576, dla_local_dram_size=1073741824, dla_global_dram_size=536870912, dryrun=False, hardware_compatible=False, timing_cache_path='/tmp/torch_tensorrt_engine_cache/timing_cache.bin', lazy_engine_init=False, cache_built_engines=False, reuse_cached_engines=False, use_explicit_typing=False, use_fp32_acc=False, enable_weight_streaming=False, enable_cross_compile_for_windows=False), weight_name_map: ~typing.Any = None)[source]¶

    PythonTorchTensorRTModule is a PyTorch module which encompasses an arbitrary TensorRT Engine.

    This module is backed by the Torch-TensorRT runtime and is only compatible with FX / Dynamo / Python deployments. This module cannot be serialized to torchscript via torch.jit.trace for C++ deployment.

    -__init__(serialized_engine: ~typing.Optional[bytes] = None, input_binding_names: ~typing.Optional[~typing.List[str]] = None, output_binding_names: ~typing.Optional[~typing.List[str]] = None, *, name: str = '', settings: ~torch_tensorrt.dynamo._settings.CompilationSettings = CompilationSettings(enabled_precisions={<dtype.f32: 7>}, debug=False, workspace_size=0, min_block_size=5, torch_executed_ops=set(), pass_through_build_failures=False, max_aux_streams=None, version_compatible=False, optimization_level=None, use_python_runtime=False, truncate_double=False, use_fast_partitioner=True, enable_experimental_decompositions=False, device=Device(type=DeviceType.GPU, gpu_id=0), require_full_compilation=False, disable_tf32=False, assume_dynamic_shape_support=False, sparse_weights=False, make_refittable=False, engine_capability=<EngineCapability.STANDARD: 1>, num_avg_timing_iters=1, dla_sram_size=1048576, dla_local_dram_size=1073741824, dla_global_dram_size=536870912, dryrun=False, hardware_compatible=False, timing_cache_path='/tmp/torch_tensorrt_engine_cache/timing_cache.bin', lazy_engine_init=False, cache_built_engines=False, reuse_cached_engines=False, use_explicit_typing=False, use_fp32_acc=False, enable_weight_streaming=False), weight_name_map: ~typing.Any = None)[source]¶
    +__init__(serialized_engine: ~typing.Optional[bytes] = None, input_binding_names: ~typing.Optional[~typing.List[str]] = None, output_binding_names: ~typing.Optional[~typing.List[str]] = None, *, name: str = '', settings: ~torch_tensorrt.dynamo._settings.CompilationSettings = CompilationSettings(enabled_precisions={<dtype.f32: 7>}, debug=False, workspace_size=0, min_block_size=5, torch_executed_ops=set(), pass_through_build_failures=False, max_aux_streams=None, version_compatible=False, optimization_level=None, use_python_runtime=False, truncate_double=False, use_fast_partitioner=True, enable_experimental_decompositions=False, device=Device(type=DeviceType.GPU, gpu_id=0), require_full_compilation=False, disable_tf32=False, assume_dynamic_shape_support=False, sparse_weights=False, make_refittable=False, engine_capability=<EngineCapability.STANDARD: 1>, num_avg_timing_iters=1, dla_sram_size=1048576, dla_local_dram_size=1073741824, dla_global_dram_size=536870912, dryrun=False, hardware_compatible=False, timing_cache_path='/tmp/torch_tensorrt_engine_cache/timing_cache.bin', lazy_engine_init=False, cache_built_engines=False, reuse_cached_engines=False, use_explicit_typing=False, use_fp32_acc=False, enable_weight_streaming=False, enable_cross_compile_for_windows=False), weight_name_map: ~typing.Any = None)[source]¶

    Takes a name, target device, serialized TensorRT engine, and binding names / order and constructs a PyTorch torch.nn.Module around it. Uses TensorRT Python APIs to run the engine

    diff --git a/docs/py_api/torch_tensorrt.html b/docs/py_api/torch_tensorrt.html index 3e426876f4..338d46c2a4 100644 --- a/docs/py_api/torch_tensorrt.html +++ b/docs/py_api/torch_tensorrt.html @@ -10,7 +10,7 @@ - torch_tensorrt — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/py_api/ts.html b/docs/py_api/ts.html index 79ddefc111..1c1a98b163 100644 --- a/docs/py_api/ts.html +++ b/docs/py_api/ts.html @@ -10,7 +10,7 @@ - torch_tensorrt.ts — Torch-TensorRT v2.6.0.dev0+e43833d documentation + torch_tensorrt.ts — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    @@ -691,7 +691,7 @@

    Functions
    -torch_tensorrt.ts.TensorRTCompileSpec(inputs: Optional[List[torch.Tensor | Input]] = None, input_signature: Optional[Any] = None, device: torch.device | Device = Device(type=DeviceType.GPU, gpu_id=0), disable_tf32: bool = False, sparse_weights: bool = False, enabled_precisions: Optional[Set[Union[dtype, dtype]]] = None, refit: bool = False, debug: bool = False, capability: EngineCapability = EngineCapability.STANDARD, num_avg_timing_iters: int = 1, workspace_size: int = 0, dla_sram_size: int = 1048576, dla_local_dram_size: int = 1073741824, dla_global_dram_size: int = 536870912, truncate_long_and_double: bool = False, calibrator: object = None, allow_shape_tensors: bool = False) <torch.ScriptClass object at 0x7efd148567b0>[source]¶
    +torch_tensorrt.ts.TensorRTCompileSpec(inputs: Optional[List[torch.Tensor | Input]] = None, input_signature: Optional[Any] = None, device: torch.device | Device = Device(type=DeviceType.GPU, gpu_id=0), disable_tf32: bool = False, sparse_weights: bool = False, enabled_precisions: Optional[Set[Union[dtype, dtype]]] = None, refit: bool = False, debug: bool = False, capability: EngineCapability = EngineCapability.STANDARD, num_avg_timing_iters: int = 1, workspace_size: int = 0, dla_sram_size: int = 1048576, dla_local_dram_size: int = 1073741824, dla_global_dram_size: int = 536870912, truncate_long_and_double: bool = False, calibrator: object = None, allow_shape_tensors: bool = False) <torch.ScriptClass object at 0x7fdbd3710cf0>[source]¶

    Utility to create a formatted spec dictionary for using the PyTorch TensorRT backend

    Keyword Arguments
    diff --git a/docs/search.html b/docs/search.html index b64d9bc7d9..62bc88d19f 100644 --- a/docs/search.html +++ b/docs/search.html @@ -9,7 +9,7 @@ - Search — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Search — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -272,7 +272,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/searchindex.js b/docs/searchindex.js index fb41a175b4..afbb8d4006 100644 --- a/docs/searchindex.js +++ b/docs/searchindex.js @@ -1 +1 @@ -Search.setIndex({"docnames": ["_cpp_api/classtorch__tensorrt_1_1DataType", "_cpp_api/classtorch__tensorrt_1_1Device_1_1DeviceType", "_cpp_api/classtorch__tensorrt_1_1TensorFormat", "_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8CacheCalibrator", "_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8Calibrator", "_cpp_api/define_macros_8h_1a18d295a837ac71add5578860b55e5502", "_cpp_api/define_macros_8h_1a282fd3c0b1c3a215148ae372070e1268", "_cpp_api/define_macros_8h_1a31398a6d4d27e28817afb0f0139e909e", "_cpp_api/define_macros_8h_1a35703561b26b1a9d2738ad7d58b27827", "_cpp_api/define_macros_8h_1abd1465eb38256d3f22cc1426b23d516b", "_cpp_api/define_macros_8h_1abe87b341f562fd1cf40b7672e4d759da", "_cpp_api/define_macros_8h_1ad19939408f7be171a74a89928b36eb59", "_cpp_api/define_macros_8h_1adad592a7b1b7eed529cdf6acd584c883", "_cpp_api/dir_cpp", "_cpp_api/dir_cpp_include", "_cpp_api/dir_cpp_include_torch_tensorrt", "_cpp_api/enum_namespacetorch__tensorrt_1_1logging_1a130f65408ad8cbaee060f05e8db69558", "_cpp_api/enum_namespacetorch__tensorrt_1a3fbe5d72e4fc624dbd038853079620eb", "_cpp_api/file_cpp_include_torch_tensorrt_logging.h", "_cpp_api/file_cpp_include_torch_tensorrt_macros.h", "_cpp_api/file_cpp_include_torch_tensorrt_ptq.h", "_cpp_api/file_cpp_include_torch_tensorrt_torch_tensorrt.h", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0593f776f469c20469e2f729fc7861a3", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0c012cb374addd90eb1f42eaec570650", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a56e110feaaba2c3fd44bd201fd21a76a", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a7cb50492421ea9de4e3db895819df6f2", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ac46ac0901cb97e3ae6e93b45f24e90b8", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ad2efd47b6c3689e58ccc595680579ae5", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1af8f3443813315af7901903d25dd495cc", "_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a226e3c83379d1012cde8578c1c86b16c", "_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a6186e305f47c1d94b6130ef6c7f7e178", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a5b405fd3bf3c8fc2e2a54cbbab979797", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a6e19490a08fb1553c9dd347a5ae79db9", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a81f9783517335dda877d8cfcf38987c9", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1ae8d56472106eeef37fbe51ff7f40c9b2", "_cpp_api/function_namespacetorch__tensorrt_1ac4ab8313ae72c2c899ea31548b528528", "_cpp_api/function_namespacetorch__tensorrt_1ad1acd06eaeaffbbcf6e7ebf426891384", "_cpp_api/function_namespacetorch__tensorrt_1ad6a4ee8ca6c8f6e5519eb1128ec7f4a1", "_cpp_api/namespace_torch_tensorrt", "_cpp_api/namespace_torch_tensorrt__logging", "_cpp_api/namespace_torch_tensorrt__ptq", "_cpp_api/namespace_torch_tensorrt__torchscript", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_logging.h", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_macros.h", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_ptq.h", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_torch_tensorrt.h", "_cpp_api/structtorch__tensorrt_1_1Device", "_cpp_api/structtorch__tensorrt_1_1GraphInputs", "_cpp_api/structtorch__tensorrt_1_1Input", "_cpp_api/structtorch__tensorrt_1_1torchscript_1_1CompileSpec", "_cpp_api/torch_tensort_cpp", "_cpp_api/unabridged_orphan", "cli/torchtrtc", "contributors/conversion", "contributors/dynamo_converters", "contributors/lowering", "contributors/partitioning", "contributors/phases", "contributors/runtime", "contributors/system_overview", "contributors/ts_converters", "contributors/useful_links", "contributors/writing_dynamo_aten_lowering_passes", "dynamo/dynamo_export", "dynamo/torch_compile", "fx/getting_started_with_fx_path", "getting_started/installation", "getting_started/jetpack", "getting_started/quick_start", "index", "indices/supported_ops", "py_api/dynamo", "py_api/fx", "py_api/logging", "py_api/ptq", "py_api/runtime", "py_api/torch_tensorrt", "py_api/ts", "sg_execution_times", "src/pytorch-sphinx-theme/docs/changelog", "src/pytorch-sphinx-theme/docs/configuring", "src/pytorch-sphinx-theme/docs/demo/api", "src/pytorch-sphinx-theme/docs/demo/demo", "src/pytorch-sphinx-theme/docs/demo/lists_tables", "src/pytorch-sphinx-theme/docs/demo/long", "src/pytorch-sphinx-theme/docs/demo/structure", "src/pytorch-sphinx-theme/docs/index", "src/pytorch-sphinx-theme/docs/installing", "ts/creating_torchscript_module_in_python", "ts/getting_started_with_cpp_api", "ts/getting_started_with_python_api", "ts/ptq", "ts/torchscript_frontend_from_pytorch", "tutorials/_rendered_examples/dynamo/converter_overloading", "tutorials/_rendered_examples/dynamo/custom_kernel_plugins", "tutorials/_rendered_examples/dynamo/engine_caching_bert_example", "tutorials/_rendered_examples/dynamo/engine_caching_example", "tutorials/_rendered_examples/dynamo/index", "tutorials/_rendered_examples/dynamo/mutable_torchtrt_module_example", "tutorials/_rendered_examples/dynamo/refit_engine_example", "tutorials/_rendered_examples/dynamo/torch_compile_advanced_usage", "tutorials/_rendered_examples/dynamo/torch_compile_resnet_example", "tutorials/_rendered_examples/dynamo/torch_compile_stable_diffusion", "tutorials/_rendered_examples/dynamo/torch_compile_transformers_example", "tutorials/_rendered_examples/dynamo/torch_export_cudagraphs", "tutorials/_rendered_examples/dynamo/torch_export_gpt2", "tutorials/_rendered_examples/dynamo/torch_export_llama2", "tutorials/_rendered_examples/dynamo/vgg16_ptq", "tutorials/_rendered_examples/dynamo/weight_streaming_example", "tutorials/_rendered_examples/index", "tutorials/notebooks", "tutorials/serving_torch_tensorrt_with_triton", "user_guide/dynamic_shapes", "user_guide/mixed_precision", "user_guide/runtime", "user_guide/saving_models", "user_guide/torch_tensorrt_explained", "user_guide/using_dla"], "filenames": ["_cpp_api/classtorch__tensorrt_1_1DataType.rst", "_cpp_api/classtorch__tensorrt_1_1Device_1_1DeviceType.rst", "_cpp_api/classtorch__tensorrt_1_1TensorFormat.rst", "_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8CacheCalibrator.rst", "_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8Calibrator.rst", "_cpp_api/define_macros_8h_1a18d295a837ac71add5578860b55e5502.rst", "_cpp_api/define_macros_8h_1a282fd3c0b1c3a215148ae372070e1268.rst", "_cpp_api/define_macros_8h_1a31398a6d4d27e28817afb0f0139e909e.rst", "_cpp_api/define_macros_8h_1a35703561b26b1a9d2738ad7d58b27827.rst", "_cpp_api/define_macros_8h_1abd1465eb38256d3f22cc1426b23d516b.rst", "_cpp_api/define_macros_8h_1abe87b341f562fd1cf40b7672e4d759da.rst", "_cpp_api/define_macros_8h_1ad19939408f7be171a74a89928b36eb59.rst", "_cpp_api/define_macros_8h_1adad592a7b1b7eed529cdf6acd584c883.rst", "_cpp_api/dir_cpp.rst", "_cpp_api/dir_cpp_include.rst", "_cpp_api/dir_cpp_include_torch_tensorrt.rst", "_cpp_api/enum_namespacetorch__tensorrt_1_1logging_1a130f65408ad8cbaee060f05e8db69558.rst", "_cpp_api/enum_namespacetorch__tensorrt_1a3fbe5d72e4fc624dbd038853079620eb.rst", "_cpp_api/file_cpp_include_torch_tensorrt_logging.h.rst", "_cpp_api/file_cpp_include_torch_tensorrt_macros.h.rst", "_cpp_api/file_cpp_include_torch_tensorrt_ptq.h.rst", "_cpp_api/file_cpp_include_torch_tensorrt_torch_tensorrt.h.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0593f776f469c20469e2f729fc7861a3.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0c012cb374addd90eb1f42eaec570650.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a56e110feaaba2c3fd44bd201fd21a76a.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a7cb50492421ea9de4e3db895819df6f2.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ac46ac0901cb97e3ae6e93b45f24e90b8.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ad2efd47b6c3689e58ccc595680579ae5.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1af8f3443813315af7901903d25dd495cc.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a226e3c83379d1012cde8578c1c86b16c.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a6186e305f47c1d94b6130ef6c7f7e178.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a5b405fd3bf3c8fc2e2a54cbbab979797.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a6e19490a08fb1553c9dd347a5ae79db9.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a81f9783517335dda877d8cfcf38987c9.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1ae8d56472106eeef37fbe51ff7f40c9b2.rst", "_cpp_api/function_namespacetorch__tensorrt_1ac4ab8313ae72c2c899ea31548b528528.rst", "_cpp_api/function_namespacetorch__tensorrt_1ad1acd06eaeaffbbcf6e7ebf426891384.rst", "_cpp_api/function_namespacetorch__tensorrt_1ad6a4ee8ca6c8f6e5519eb1128ec7f4a1.rst", "_cpp_api/namespace_torch_tensorrt.rst", "_cpp_api/namespace_torch_tensorrt__logging.rst", "_cpp_api/namespace_torch_tensorrt__ptq.rst", "_cpp_api/namespace_torch_tensorrt__torchscript.rst", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_logging.h.rst", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_macros.h.rst", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_ptq.h.rst", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_torch_tensorrt.h.rst", "_cpp_api/structtorch__tensorrt_1_1Device.rst", "_cpp_api/structtorch__tensorrt_1_1GraphInputs.rst", "_cpp_api/structtorch__tensorrt_1_1Input.rst", "_cpp_api/structtorch__tensorrt_1_1torchscript_1_1CompileSpec.rst", "_cpp_api/torch_tensort_cpp.rst", "_cpp_api/unabridged_orphan.rst", "cli/torchtrtc.rst", "contributors/conversion.rst", "contributors/dynamo_converters.rst", "contributors/lowering.rst", "contributors/partitioning.rst", "contributors/phases.rst", "contributors/runtime.rst", "contributors/system_overview.rst", "contributors/ts_converters.rst", "contributors/useful_links.rst", "contributors/writing_dynamo_aten_lowering_passes.rst", "dynamo/dynamo_export.rst", "dynamo/torch_compile.rst", "fx/getting_started_with_fx_path.rst", "getting_started/installation.rst", "getting_started/jetpack.rst", "getting_started/quick_start.rst", "index.rst", "indices/supported_ops.rst", "py_api/dynamo.rst", "py_api/fx.rst", "py_api/logging.rst", "py_api/ptq.rst", "py_api/runtime.rst", "py_api/torch_tensorrt.rst", "py_api/ts.rst", "sg_execution_times.rst", "src/pytorch-sphinx-theme/docs/changelog.rst", "src/pytorch-sphinx-theme/docs/configuring.rst", "src/pytorch-sphinx-theme/docs/demo/api.rst", "src/pytorch-sphinx-theme/docs/demo/demo.rst", "src/pytorch-sphinx-theme/docs/demo/lists_tables.rst", "src/pytorch-sphinx-theme/docs/demo/long.rst", "src/pytorch-sphinx-theme/docs/demo/structure.rst", "src/pytorch-sphinx-theme/docs/index.rst", "src/pytorch-sphinx-theme/docs/installing.rst", "ts/creating_torchscript_module_in_python.rst", "ts/getting_started_with_cpp_api.rst", "ts/getting_started_with_python_api.rst", "ts/ptq.rst", "ts/torchscript_frontend_from_pytorch.rst", "tutorials/_rendered_examples/dynamo/converter_overloading.rst", "tutorials/_rendered_examples/dynamo/custom_kernel_plugins.rst", "tutorials/_rendered_examples/dynamo/engine_caching_bert_example.rst", "tutorials/_rendered_examples/dynamo/engine_caching_example.rst", "tutorials/_rendered_examples/dynamo/index.rst", "tutorials/_rendered_examples/dynamo/mutable_torchtrt_module_example.rst", "tutorials/_rendered_examples/dynamo/refit_engine_example.rst", "tutorials/_rendered_examples/dynamo/torch_compile_advanced_usage.rst", "tutorials/_rendered_examples/dynamo/torch_compile_resnet_example.rst", "tutorials/_rendered_examples/dynamo/torch_compile_stable_diffusion.rst", "tutorials/_rendered_examples/dynamo/torch_compile_transformers_example.rst", "tutorials/_rendered_examples/dynamo/torch_export_cudagraphs.rst", "tutorials/_rendered_examples/dynamo/torch_export_gpt2.rst", "tutorials/_rendered_examples/dynamo/torch_export_llama2.rst", "tutorials/_rendered_examples/dynamo/vgg16_ptq.rst", "tutorials/_rendered_examples/dynamo/weight_streaming_example.rst", "tutorials/_rendered_examples/index.rst", "tutorials/notebooks.rst", "tutorials/serving_torch_tensorrt_with_triton.rst", "user_guide/dynamic_shapes.rst", "user_guide/mixed_precision.rst", "user_guide/runtime.rst", "user_guide/saving_models.rst", "user_guide/torch_tensorrt_explained.rst", "user_guide/using_dla.rst"], "titles": ["Class DataType", "Class Device::DeviceType", "Class TensorFormat", "Template Class Int8CacheCalibrator", "Template Class Int8Calibrator", "Define STR", "Define TORCH_TENSORRT_PATCH_VERSION", "Define TORCH_TENSORRT_MAJOR_VERSION", "Define TORCH_TENSORRT_MINOR_VERSION", "Define TORCHTRT_API", "Define XSTR", "Define TORCHTRT_HIDDEN", "Define TORCH_TENSORRT_VERSION", "Directory cpp", "Directory include", "Directory torch_tensorrt", "Enum Level", "Enum EngineCapability", "File logging.h", "File macros.h", "File ptq.h", "File torch_tensorrt.h", "Function torch_tensorrt::logging::get_logging_prefix", "Function torch_tensorrt::logging::get_reportable_log_level", "Function torch_tensorrt::logging::get_is_colored_output_on", "Function torch_tensorrt::logging::set_reportable_log_level", "Function torch_tensorrt::logging::log", "Function torch_tensorrt::logging::set_is_colored_output_on", "Function torch_tensorrt::logging::set_logging_prefix", "Template Function torch_tensorrt::ptq::make_int8_cache_calibrator", "Template Function torch_tensorrt::ptq::make_int8_calibrator", "Function torch_tensorrt::torchscript::check_method_operator_support", "Function torch_tensorrt::torchscript::compile", "Function torch_tensorrt::torchscript::embed_engine_in_new_module", "Function torch_tensorrt::torchscript::convert_method_to_trt_engine", "Function torch_tensorrt::get_build_info", "Function torch_tensorrt::set_device", "Function torch_tensorrt::dump_build_info", "Namespace torch_tensorrt", "Namespace torch_tensorrt::logging", "Namespace torch_tensorrt::ptq", "Namespace torch_tensorrt::torchscript", "Program Listing for File logging.h", "Program Listing for File macros.h", "Program Listing for File ptq.h", "Program Listing for File torch_tensorrt.h", "Struct Device", "Struct GraphInputs", "Struct Input", "Struct CompileSpec", "Torch-TensorRT C++ API", "Full API", "torchtrtc", "Conversion Phase", "Writing Dynamo Converters", "Lowering Phase", "Partitioning Phase", "Compiler Phases", "Runtime Phase", "System Overview", "Writing TorchScript Converters", "Useful Links for Torch-TensorRT Development", "Writing Dynamo ATen Lowering Passes", "Compiling Exported Programs with Torch-TensorRT", "TensorRT Backend for torch.compile", "Torch-TensorRT (FX Frontend) User Guide", "Installation", "Overview", "Quick Start", "Torch-TensorRT", "Operators Supported", "torch_tensorrt.dynamo", "torch_tensorrt.fx", "torch_tensorrt.logging", "torch_tensorrt.ts.ptq", "torch_tensorrt.runtime", "torch_tensorrt", "torch_tensorrt.ts", "Computation times", "Changelog", "Configuration", "5. :mod:`test_py_module`", "3. Paragraph Level Markup", "4. Lists & Tables", "1. Long Sticky Nav", "1. Structural Elements", "<no title>", "Installation", "Creating a TorchScript Module", "Using Torch-TensorRT in C++", "Using Torch-TensorRT in Python", "Post Training Quantization (PTQ)", "Using Torch-TensorRT TorchScript Frontend Directly From PyTorch", "Overloading Torch-TensorRT Converters with Custom Converters", "Using Custom Kernels within TensorRT Engines with Torch-TensorRT", "Engine Caching (BERT)", "Engine Caching", "Dependencies", "Mutable Torch TensorRT Module", "Refitting Torch-TensorRT Programs with New Weights", "Torch Compile Advanced Usage", "Compiling ResNet with dynamic shapes using the torch.compile backend", "Compiling Stable Diffusion model using the torch.compile backend", "Compiling BERT using the torch.compile backend", "Torch Export with Cudagraphs", "Compiling GPT2 using the dynamo backend", "Compiling Llama2 using the dynamo backend", "Deploy Quantized Models using Torch-TensorRT", "Weight Streaming", "Torch-TensorRT Tutorials", "Legacy notebooks", "Serving a Torch-TensorRT model with Triton", "Dynamic shapes with Torch-TensorRT", "Compile Mixed Precision models with Torch-TensorRT", "Deploying Torch-TensorRT Programs", "Saving models compiled with Torch-TensorRT", "Torch-TensorRT Explained", "DLA"], "terms": {"defin": [0, 1, 2, 3, 4, 33, 43, 46, 47, 48, 49, 51, 52, 54, 65, 68, 75, 76, 80, 88, 89, 90, 91, 93, 94, 96, 100, 103, 104, 105, 106, 110], "file": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 46, 47, 48, 49, 52, 54, 56, 58, 59, 64, 65, 66, 67, 68, 71, 72, 74, 76, 77, 78, 80, 81, 83, 87, 89, 91, 111, 112, 115], "torch_tensorrt": [0, 1, 2, 14, 16, 17, 42, 43, 44, 46, 47, 48, 49, 50, 51, 52, 54, 56, 62, 63, 64, 65, 68, 69, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 102, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 117], "h": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 15, 46, 47, 48, 49, 50, 51, 52, 55, 68, 76, 89, 91], "support": [0, 1, 2, 27, 31, 46, 48, 49, 52, 54, 56, 61, 63, 65, 67, 68, 69, 72, 75, 76, 77, 80, 81, 88, 89, 90, 93, 94, 99, 101, 103, 105, 106, 107, 108, 111, 113, 116, 117], "data": [0, 2, 3, 4, 29, 30, 44, 46, 48, 49, 52, 53, 56, 57, 59, 60, 64, 65, 70, 71, 72, 74, 76, 77, 82, 86, 90, 91, 94, 96, 107, 108, 110], "type": [0, 1, 2, 30, 49, 50, 52, 53, 56, 58, 60, 62, 63, 64, 65, 71, 72, 74, 75, 76, 77, 82, 89, 90, 91, 93, 94, 96, 107, 108, 110, 113, 115], "can": [0, 1, 4, 29, 30, 34, 46, 47, 48, 49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 62, 63, 64, 65, 66, 67, 68, 71, 74, 75, 76, 77, 80, 82, 88, 89, 90, 91, 92, 93, 94, 96, 98, 99, 100, 103, 104, 107, 108, 110, 111, 112, 113, 114, 115, 116], "us": [0, 1, 2, 3, 4, 29, 30, 32, 34, 36, 43, 44, 45, 46, 48, 49, 52, 53, 54, 56, 58, 59, 60, 62, 63, 64, 65, 66, 67, 68, 69, 71, 72, 74, 75, 76, 77, 78, 80, 81, 82, 83, 88, 91, 96, 97, 98, 99, 108, 109, 111, 113, 114, 115, 116, 117], "tensorrt": [0, 1, 3, 4, 29, 30, 31, 32, 33, 34, 37, 44, 45, 46, 48, 49, 52, 53, 54, 55, 56, 57, 59, 60, 62, 67, 68, 71, 72, 74, 75, 76, 77, 88, 91, 96, 97, 100, 101, 102, 103, 104, 108], "engin": [0, 1, 17, 32, 33, 34, 45, 46, 48, 49, 52, 53, 56, 57, 59, 62, 63, 64, 69, 71, 72, 75, 76, 77, 80, 89, 90, 91, 92, 93, 97, 99, 101, 103, 108, 109, 112, 114, 116, 117], "thi": [0, 1, 2, 29, 30, 42, 43, 44, 45, 46, 47, 48, 49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 62, 63, 64, 65, 66, 67, 68, 71, 72, 75, 76, 77, 80, 81, 82, 84, 85, 88, 89, 91, 92, 93, 94, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 110, 111, 112, 113, 114, 115, 116], "compat": [0, 1, 46, 55, 58, 64, 65, 71, 75, 76, 77, 116], "c10": [0, 1, 45, 46, 48, 49, 89, 91], "check": [0, 1, 31, 46, 52, 55, 60, 65, 67, 71, 75, 77, 89, 94, 98, 99, 111, 114], "trt": [0, 1, 3, 4, 46, 48, 53, 55, 58, 60, 62, 64, 65, 67, 68, 70, 71, 75, 76, 89, 93, 94, 103, 105, 106, 108, 112, 114, 115], "so": [0, 44, 52, 53, 54, 55, 58, 59, 60, 62, 64, 65, 66, 67, 72, 75, 76, 81, 82, 83, 89, 91, 93, 94, 96, 100, 101, 103, 105, 106, 112], "should": [0, 3, 4, 29, 45, 49, 52, 53, 54, 55, 56, 57, 59, 60, 63, 64, 65, 67, 71, 75, 76, 77, 80, 82, 85, 91, 93, 94, 95, 96, 99, 104, 111], "reason": [0, 65, 88, 93, 94, 96, 116], "you": [0, 1, 2, 29, 30, 46, 48, 49, 52, 53, 54, 55, 56, 58, 59, 60, 63, 65, 66, 67, 68, 71, 75, 76, 77, 80, 82, 83, 84, 88, 89, 90, 91, 92, 93, 94, 96, 97, 98, 99, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116], "need": [0, 1, 2, 25, 29, 43, 46, 53, 54, 55, 60, 65, 66, 67, 71, 72, 75, 76, 82, 89, 90, 91, 93, 94, 95, 96, 98, 99, 110, 111, 112, 114], "explictli": 0, "public": [0, 1, 2, 3, 4, 44, 45, 46, 47, 48, 49, 83, 91], "enum": [0, 1, 2, 42, 45, 46, 71, 77, 91, 93], "valu": [0, 1, 2, 16, 17, 45, 46, 48, 53, 56, 58, 60, 63, 70, 71, 74, 76, 80, 89, 98, 100, 101, 103, 108, 110], "underli": [0, 1, 2, 46, 60], "In": [0, 1, 2, 46, 53, 54, 56, 57, 58, 59, 60, 64, 65, 66, 75, 76, 82, 83, 85, 90, 91, 93, 94, 98, 110, 111, 112, 113, 114, 115], "case": [0, 1, 2, 46, 49, 53, 54, 56, 58, 60, 62, 64, 65, 66, 67, 75, 76, 91, 93, 94, 98, 99, 112, 113, 114], "itself": [0, 1, 2, 46, 52, 55, 92, 93, 111], "interfac": [0, 1, 2, 46, 58, 59, 60, 64, 69, 91], "vs": [0, 1, 2, 46, 55, 66, 71, 76, 77, 92], "normal": [0, 1, 2, 46, 65, 82, 88, 89, 91, 93, 98, 99, 104, 107, 111, 117], "instatin": [0, 1, 2, 46], "ex": [0, 1, 2, 33, 46, 67, 77, 83, 85], "kfloat": [0, 45, 49], "enumer": [0, 1, 2, 16, 17, 46], "klong": [0, 45], "int64": [0, 76, 77, 108], "kdoubl": [0, 45], "fp64": [0, 76], "fp32": [0, 48, 49, 52, 64, 65, 71, 76, 77, 91, 105, 106, 110, 111, 113], "khalf": [0, 45, 89], "fp16": [0, 48, 49, 52, 64, 65, 71, 72, 76, 89, 90, 98, 102, 105, 106, 108, 113, 117], "kchar": [0, 45], "int8": [0, 44, 48, 49, 52, 64, 71, 76, 77, 91, 107, 117], "kint": [0, 45], "int": [0, 3, 4, 36, 44, 45, 49, 52, 54, 56, 63, 64, 70, 71, 72, 76, 77, 80, 89, 94, 107, 108], "kbool": [0, 45], "bool": [0, 1, 2, 3, 4, 24, 27, 30, 31, 42, 44, 45, 46, 49, 55, 60, 64, 70, 71, 72, 74, 75, 76, 77, 80, 89, 91, 94], "kunknown": [0, 2, 45], "sentinel": [0, 2, 76], "function": [0, 1, 2, 3, 4, 46, 48, 49, 54, 55, 56, 58, 60, 62, 64, 65, 66, 88, 89, 91, 92, 93, 94, 99, 100, 103, 104, 105, 106, 110, 111, 112, 114, 116, 117], "default": [0, 1, 2, 3, 4, 16, 29, 30, 33, 43, 45, 46, 48, 49, 52, 54, 56, 62, 64, 65, 66, 71, 72, 75, 76, 77, 80, 81, 82, 89, 90, 91, 92, 93, 94, 96, 107, 112, 114, 115, 116], "construct": [0, 1, 2, 3, 4, 46, 48, 49, 53, 54, 55, 57, 59, 60, 65, 74, 75, 76, 82, 83, 89, 91, 93, 94, 96, 112], "new": [0, 1, 2, 3, 4, 32, 33, 46, 48, 49, 56, 58, 59, 60, 62, 64, 65, 68, 69, 71, 77, 82, 89, 96, 97, 98, 101, 103, 104, 109, 111, 114], "object": [0, 1, 2, 3, 4, 46, 48, 49, 52, 58, 60, 62, 63, 64, 71, 75, 76, 77, 91, 92, 93, 112, 115], "inlin": [0, 1, 2, 3, 4, 29, 30, 44, 46, 48, 55, 83, 86, 89], "constexpr": [0, 1, 2, 45, 46, 94], "t": [0, 1, 2, 45, 46, 55, 60, 65, 66, 70, 76, 80, 82, 83, 88, 89, 91, 93, 94, 107, 111, 112], "constructor": [0, 2, 46, 48, 49, 58, 88], "from": [0, 1, 2, 3, 4, 29, 30, 44, 46, 48, 49, 52, 53, 55, 56, 57, 58, 59, 60, 63, 64, 65, 67, 69, 71, 72, 75, 76, 77, 78, 80, 81, 82, 83, 88, 89, 91, 93, 94, 95, 96, 98, 99, 102, 103, 105, 106, 107, 108, 110, 111, 114, 115, 116], "torchtrt_api": [0, 2, 19, 22, 23, 24, 25, 26, 27, 28, 31, 32, 33, 34, 35, 36, 37, 42, 43, 44, 45, 48, 49, 50], "scalartyp": [0, 45, 70], "torch": [0, 1, 2, 4, 20, 21, 29, 30, 31, 32, 33, 34, 37, 44, 45, 46, 47, 48, 49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 62, 67, 71, 72, 74, 75, 76, 77, 78, 88, 91, 95, 96, 97, 108, 117], "paramet": [0, 1, 2, 3, 4, 25, 26, 27, 29, 30, 31, 32, 33, 34, 36, 46, 48, 49, 53, 54, 55, 60, 64, 65, 71, 72, 74, 75, 76, 77, 86, 88, 89, 99, 105, 106], "oper": [0, 1, 2, 3, 4, 31, 44, 45, 46, 49, 52, 53, 55, 56, 57, 58, 59, 60, 62, 63, 65, 69, 71, 76, 77, 90, 91, 93, 99, 101, 103, 116, 117], "const": [0, 1, 2, 3, 4, 29, 30, 31, 32, 33, 34, 36, 44, 45, 46, 55, 60, 70, 89, 91], "get": [0, 1, 2, 3, 4, 23, 35, 44, 46, 55, 56, 60, 62, 63, 65, 67, 75, 76, 89, 91, 93, 96, 105, 106, 108, 110, 111], "return": [0, 1, 2, 3, 4, 23, 24, 29, 30, 31, 32, 33, 34, 35, 42, 43, 44, 45, 46, 54, 55, 56, 57, 58, 59, 60, 62, 64, 65, 71, 72, 75, 76, 77, 88, 89, 90, 91, 93, 94, 96, 99, 100, 107, 108, 111, 112, 113], "explicit": [0, 1, 2, 3, 4, 45, 46, 55, 65, 72, 75, 82, 91, 116], "delet": [0, 1, 2, 45, 46, 55], "other": [0, 1, 2, 45, 46, 52, 53, 55, 58, 62, 64, 65, 66, 70, 71, 75, 76, 81, 82, 89, 90, 93, 114], "comparis": [0, 2], "true": [0, 1, 2, 4, 46, 49, 55, 56, 60, 62, 64, 65, 70, 71, 72, 75, 76, 77, 80, 83, 89, 91, 92, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 111, 113, 114, 117], "fals": [0, 1, 2, 3, 4, 44, 45, 46, 49, 54, 62, 64, 65, 70, 71, 72, 75, 76, 77, 80, 81, 82, 83, 89, 91, 92, 93, 94, 95, 96, 98, 99, 100, 102, 103, 104, 105, 106, 107, 108, 114], "struct": [1, 21, 38, 41, 45, 54, 91], "onli": [1, 3, 4, 16, 29, 44, 46, 48, 52, 54, 55, 56, 59, 60, 64, 65, 67, 68, 71, 72, 75, 76, 82, 91, 93, 94, 98, 99, 106, 108, 113, 114, 117], "applic": [1, 29, 46, 52, 55, 59, 64, 71, 75, 76, 89, 90, 92, 114, 117], "kcuda": [1, 46, 56, 89], "which": [1, 2, 29, 32, 34, 46, 49, 53, 54, 55, 56, 57, 58, 59, 60, 62, 63, 64, 65, 66, 71, 72, 74, 75, 76, 77, 80, 82, 83, 88, 89, 90, 91, 92, 93, 94, 96, 100, 101, 104, 105, 106, 110, 111, 112, 113, 114, 115, 116], "map": [1, 46, 53, 54, 55, 57, 59, 60, 65, 76, 89, 91, 92, 96, 100, 110, 111], "kgpu": [1, 45, 46], "To": [1, 46, 52, 54, 56, 64, 66, 71, 80, 88, 89, 90, 92, 94, 99, 105, 106, 111], "datatyp": [1, 21, 38, 45, 46, 48, 49, 50, 71, 76, 77, 90, 94, 111, 113], "target": [1, 33, 45, 46, 48, 49, 52, 54, 56, 58, 59, 64, 65, 66, 69, 71, 75, 76, 77, 90, 91, 92, 93, 94, 99, 116, 117], "gpu": [1, 32, 34, 36, 45, 46, 52, 64, 65, 71, 75, 76, 77, 89, 91, 92, 94, 105, 106, 108, 111, 114, 116, 117], "run": [1, 34, 46, 49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 64, 65, 66, 67, 68, 71, 72, 75, 76, 77, 82, 88, 89, 90, 91, 92, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 110, 111, 112, 113, 114, 115, 116, 117], "kdla": [1, 45, 46, 117], "dla": [1, 45, 46, 49, 52, 64, 69, 71, 76, 77], "intern": [1, 16, 46, 60, 63, 73, 75, 82, 89], "note": [1, 46, 48, 54, 60, 62, 65, 66, 67, 75, 76, 80, 82, 89, 94, 99, 112, 117], "The": [1, 46, 48, 49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 62, 63, 64, 65, 66, 71, 75, 76, 77, 80, 83, 88, 90, 91, 92, 93, 94, 95, 96, 98, 99, 101, 104, 105, 108, 110, 111, 112, 115, 116], "valid": [1, 46, 56, 60, 62, 71, 75, 76, 93], "kcpu": [1, 46], "comparison": [1, 46], "an": [2, 3, 4, 48, 49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 62, 64, 65, 66, 68, 71, 72, 74, 75, 76, 77, 80, 82, 83, 88, 89, 90, 91, 93, 94, 96, 99, 100, 104, 105, 106, 108, 110, 111, 112, 114, 115, 116], "memeori": 2, "layout": [2, 48, 70, 71, 76, 77], "store": [2, 4, 49, 52, 53, 58, 60, 64, 65, 71, 75, 76, 77, 88, 89, 94, 96, 99], "tensor": [2, 33, 44, 45, 48, 49, 52, 53, 54, 55, 56, 58, 60, 62, 63, 64, 65, 70, 71, 72, 75, 76, 77, 88, 89, 90, 91, 93, 94, 100, 108, 110], "kcontigu": [2, 45, 48], "contigu": [2, 48, 49, 52, 71, 76, 77], "nchw": [2, 71, 76, 77], "linear": [2, 56, 70, 76, 88, 94, 107, 113], "kchannelslast": [2, 45], "channel": [2, 76, 81], "last": [2, 55, 65, 76, 107], "nhwc": [2, 52], "memoryformat": [2, 45], "ptq": [3, 4, 15, 18, 19, 38, 50, 51, 52, 69, 71, 76, 77], "privat": [3, 4, 44, 45, 91], "algorithm": [3, 4, 29, 30, 44, 65, 74, 91, 106], "typenam": [3, 4, 29, 30, 44], "gener": [3, 4, 29, 52, 55, 58, 59, 60, 62, 64, 65, 66, 71, 72, 80, 82, 83, 86, 88, 89, 91, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 107, 108, 109, 114], "int8calibr": [3, 20, 30, 40, 44, 50], "implement": [3, 4, 55, 56, 58, 63, 65, 75, 81, 89, 91, 94, 96, 114], "specifi": [3, 4, 33, 52, 54, 60, 64, 65, 66, 71, 76, 77, 80, 82, 90, 92, 108, 111, 112, 113, 115, 116], "calibr": [3, 4, 29, 30, 44, 49, 52, 71, 74, 76, 77, 89, 91], "read": [3, 4, 29, 30, 44, 80, 82, 91], "nvinfer1": [3, 4, 29, 30, 44, 45, 49, 60, 91], "iint8calibr": [3, 4, 29, 30, 44, 45, 49, 71, 76, 77, 91], "iint8entropycalibrator2": [3, 4, 29, 30, 44, 91], "std": [3, 4, 22, 26, 28, 29, 30, 31, 33, 34, 35, 42, 44, 45, 47, 48, 49, 56, 89, 91, 111, 117], "string": [3, 4, 18, 20, 21, 22, 26, 28, 29, 30, 31, 33, 34, 35, 42, 44, 45, 49, 54, 56, 58, 60, 64, 71, 76, 80, 89, 91], "cache_file_path": [3, 4, 29, 30, 44], "8": [3, 52, 55, 63, 64, 66, 75, 76, 82, 83, 86, 89, 94, 101, 104, 111, 112], "cach": [3, 4, 29, 30, 44, 52, 64, 65, 69, 71, 72, 74, 76, 89, 91, 97, 109, 114], "getbatchs": [3, 4, 44], "noexceptoverrid": [3, 4], "batch": [3, 4, 44, 64, 65, 72, 75, 91, 96, 101, 103, 107, 108, 111, 112, 117], "size": [3, 4, 44, 48, 49, 52, 55, 56, 64, 65, 70, 71, 72, 76, 77, 80, 89, 91, 94, 96, 101, 103, 107, 110, 112], "next": [3, 4, 53, 54, 58, 63, 72, 76, 80, 82, 83, 91, 93, 100, 104, 107, 111], "alwai": [3, 4, 27, 52, 76, 82, 99, 108], "1": [3, 4, 33, 44, 45, 48, 49, 52, 54, 55, 56, 58, 60, 62, 63, 64, 65, 66, 70, 71, 72, 74, 75, 76, 77, 79, 80, 82, 83, 86, 88, 89, 90, 91, 92, 93, 94, 95, 96, 98, 99, 101, 103, 104, 105, 106, 107, 108, 110, 112, 113, 115, 117], "due": [3, 4, 66, 81, 82, 107], "issu": [3, 4, 64, 71, 76, 89, 100, 103], "getbatch": [3, 4, 44], "void": [3, 4, 25, 26, 27, 28, 36, 37, 42, 44, 45], "bind": [3, 4, 33, 44, 75, 77, 82], "char": [3, 4, 44, 52, 89], "name": [3, 4, 31, 33, 34, 44, 54, 56, 58, 60, 65, 66, 67, 72, 74, 75, 76, 77, 82, 83, 88, 89, 92, 93, 94, 99, 104, 107, 111, 113], "nbbind": [3, 4, 44], "Not": 3, "arrai": [3, 4, 33, 53, 54, 76, 77, 93, 94, 108], "pointer": [3, 4, 91], "fed": [3, 4, 48], "buffer": [3, 4, 65, 94], "each": [3, 4, 49, 53, 55, 56, 58, 60, 64, 65, 66, 71, 72, 75, 80, 82, 89, 93, 99, 106, 114], "input": [3, 4, 21, 29, 33, 38, 44, 45, 47, 49, 50, 52, 53, 54, 55, 56, 58, 60, 62, 63, 64, 65, 68, 70, 71, 72, 73, 75, 76, 77, 83, 88, 89, 90, 91, 92, 93, 94, 95, 96, 98, 99, 100, 104, 105, 106, 107, 108, 110, 111, 112, 113, 114, 115, 116, 117], "number": [3, 4, 49, 52, 54, 55, 56, 60, 63, 64, 65, 71, 72, 76, 77, 80, 89, 90, 94, 99, 101, 103, 108, 110, 116], "readcalibrationcach": [3, 4, 44], "size_t": [3, 4, 44, 91], "length": [3, 4, 44, 65, 70, 83, 108], "how": [3, 4, 66, 67, 82, 84, 86, 88, 92, 93, 94, 96, 98, 100, 107, 108, 110, 111, 112, 114], "enabl": [3, 4, 24, 49, 52, 54, 56, 57, 59, 64, 65, 66, 71, 72, 74, 75, 76, 77, 80, 96, 98, 99, 101, 103, 104, 105, 106, 108, 113, 114], "use_cach": [3, 4, 30, 44, 74, 91, 105, 106, 108], "set": [3, 4, 16, 21, 25, 27, 29, 32, 34, 36, 45, 46, 48, 49, 52, 53, 54, 55, 56, 57, 58, 59, 65, 66, 71, 72, 75, 76, 77, 80, 84, 87, 88, 89, 90, 91, 93, 94, 99, 105, 107, 108, 110, 112, 113, 114, 116, 117], "writecalibrationcach": [3, 4, 44], "write": [3, 4, 29, 30, 44, 65, 69, 82, 89, 91, 111], "provid": [3, 4, 49, 52, 54, 56, 58, 60, 62, 64, 65, 66, 68, 71, 72, 75, 76, 77, 82, 89, 90, 91, 92, 93, 96, 97, 99, 100, 104, 108, 109, 111, 112, 114, 115, 116], "cast": [3, 4, 55, 64, 71, 105, 106, 113], "convienc": [3, 4, 49], "convert": [3, 4, 31, 32, 34, 52, 55, 56, 57, 59, 63, 64, 69, 71, 76, 77, 90, 92, 94, 97, 101, 103, 108, 109, 110, 114], "easili": [3, 4, 98], "assign": [3, 4, 81], "ptq_calibr": [3, 4, 45, 49, 91], "field": [3, 4, 63, 72, 76, 91], "compilespec": [3, 4, 21, 32, 34, 41, 45, 50, 56, 77, 89, 91, 117], "dataloaderuniqueptr": [4, 44], "libtorch": [4, 37, 60, 66, 68, 89, 91, 116], "dataload": [4, 29, 30, 44, 49, 74, 91, 107], "unique_ptr": [4, 30], "unqiue_ptr": 4, "A": [4, 29, 30, 32, 33, 47, 48, 54, 55, 56, 60, 65, 66, 71, 72, 76, 77, 83, 91, 102, 111], "uniqu": [4, 90], "what": [4, 54, 55, 65, 68, 76, 82, 88, 89, 90, 105, 106, 116], "make_data_load": [4, 91], "factori": [4, 29, 30, 64, 71, 91], "path": [4, 13, 14, 15, 29, 30, 52, 64, 65, 66, 67, 71, 74, 76, 88, 89, 91, 96, 104, 107, 111, 116], "find": [4, 65, 66, 67, 89, 94, 108], "whether": [4, 52, 54, 64, 65, 71, 72, 76, 81, 91, 101, 103, 114], "exist": [4, 31, 32, 34, 54, 63, 64, 65, 67, 71, 74, 76, 77, 91, 96, 110], "There": [4, 53, 54, 59, 60, 62, 63, 65, 66, 83, 88, 91, 99, 110, 111, 112, 114], "consum": [4, 53, 88], "macro": [5, 6, 7, 8, 9, 10, 11, 12, 15, 18, 20, 21, 42, 44, 45, 50, 51], "x": [5, 10, 33, 43, 55, 56, 66, 67, 68, 75, 77, 83, 88, 89, 93, 94, 96, 100, 104, 107, 108, 112, 113, 115], "includ": [13, 15, 16, 35, 37, 42, 43, 44, 45, 51, 52, 54, 56, 57, 58, 59, 62, 64, 65, 66, 67, 68, 71, 72, 75, 76, 80, 82, 88, 89, 91, 94, 114], "parent": [14, 15, 18, 19, 20, 21], "cpp": [14, 15, 42, 43, 44, 45, 51, 55, 59, 66, 89, 91], "log": [15, 16, 19, 20, 38, 44, 50, 51, 55, 60, 64, 65, 69, 70, 71, 72, 76, 93, 101, 103, 113], "emum": [16, 17], "messag": [16, 25, 26, 52, 73], "sever": [16, 26, 73], "kinternal_error": [16, 42], "print": [16, 31, 44, 62, 64, 67, 71, 77, 82, 89, 92, 93, 94, 95, 96, 98, 99, 101, 103, 105, 106, 107, 108, 111], "error": [16, 49, 52, 53, 55, 59, 64, 65, 71, 73, 76, 77, 82, 89, 112], "kerror": [16, 42], "all": [16, 42, 43, 44, 45, 49, 52, 54, 55, 56, 58, 62, 64, 65, 66, 67, 71, 73, 75, 76, 78, 82, 83, 88, 89, 90, 91, 93, 94, 105, 106, 109, 110, 111, 113, 114, 116], "kwarn": [16, 42], "warn": [16, 44, 52, 60, 73, 75], "kinfo": [16, 42, 44], "info": [16, 32, 34, 45, 52, 60, 73, 75, 76, 113], "kdebug": [16, 42, 44], "debug": [16, 27, 45, 49, 52, 60, 62, 64, 71, 73, 75, 76, 77, 92, 94, 95, 96, 98, 99, 100, 101, 103, 107, 113], "kgraph": [16, 42, 55], "everyth": [16, 64, 71, 76], "intermedi": [16, 49, 52, 54, 64, 71, 73, 76, 77, 88, 113, 116], "graph": [16, 31, 32, 34, 45, 49, 52, 53, 54, 56, 57, 59, 60, 62, 63, 64, 65, 71, 72, 73, 76, 77, 88, 89, 93, 94, 96, 98, 99, 101, 103, 104, 110, 112, 114], "lower": [16, 54, 63, 65, 69, 71, 72, 73, 76, 83, 94, 96, 101, 103, 108, 110, 116], "phase": [16, 60, 63, 89, 93, 99, 112, 116], "select": [17, 29, 30, 34, 49, 52, 58, 64, 65, 66, 70, 71, 76, 77, 81, 84, 90, 91, 94, 116], "capabl": [17, 45, 49, 52, 58, 71, 76, 77, 92, 93], "kstandard": [17, 45, 49], "ksafeti": [17, 45], "kdla_standalon": [17, 45], "directori": [18, 19, 20, 21, 42, 43, 44, 45, 50, 66, 67, 71, 91, 96], "program": [18, 19, 20, 21, 29, 51, 52, 57, 58, 59, 69, 71, 88, 96, 97, 105, 106, 109, 112], "list": [18, 19, 20, 21, 31, 49, 51, 53, 56, 58, 60, 62, 63, 65, 68, 70, 71, 72, 75, 76, 77, 86, 89, 90, 93, 94, 111], "torchscript": [19, 21, 38, 43, 45, 49, 50, 52, 56, 57, 58, 59, 63, 68, 71, 72, 74, 75, 76, 77, 90, 110, 112, 117], "str": [19, 43, 44, 50, 54, 64, 65, 70, 71, 74, 75, 76, 77, 93, 94, 96, 107], "torch_tensorrt_major_vers": [19, 43, 50], "torch_tensorrt_minor_vers": [19, 43, 50], "torch_tensorrt_patch_vers": [19, 43, 50], "torch_tensorrt_vers": [19, 43, 50], "torchtrt_hidden": [19, 43, 50], "xstr": [19, 43, 50], "nvinfer": [20, 44], "fstream": [20, 44], "iostream": [20, 21, 44, 45, 89], "iter": [20, 44, 49, 52, 53, 64, 71, 74, 76, 77, 95, 96, 107, 108], "memori": [20, 21, 44, 45, 55, 60, 71, 76, 77, 89, 90, 94, 96, 105, 106, 108], "sstream": [20, 44], "vector": [20, 21, 33, 44, 45, 47, 48, 49, 56, 58, 76, 89, 91, 117], "templat": [20, 40, 44, 45, 50, 80, 89], "int8cachecalibr": [20, 29, 40, 44, 50], "cuda_runtim": [21, 45], "custom_class": [21, 45], "devic": [21, 33, 36, 38, 45, 49, 50, 52, 58, 64, 70, 71, 72, 74, 75, 76, 77, 90, 91, 92, 94, 98, 102, 105, 106, 108, 110, 117], "graphinput": [21, 38, 45, 49, 50], "devicetyp": [21, 38, 45, 46, 50, 75, 76, 77, 91, 92, 94, 117], "tensorformat": [21, 38, 45, 48, 50, 76, 94], "level": [23, 25, 26, 39, 42, 44, 50, 54, 55, 56, 59, 64, 65, 71, 76, 77, 86, 88, 93, 94, 111, 116], "current": [23, 54, 56, 58, 60, 62, 63, 64, 65, 66, 67, 71, 72, 75, 76, 77, 80, 93, 94, 98, 105, 106, 107, 108, 114], "report": [23, 44, 75], "Is": [24, 76], "color": [24, 27, 82], "output": [24, 27, 33, 49, 52, 53, 54, 55, 56, 58, 60, 62, 63, 64, 65, 66, 71, 73, 75, 76, 77, 80, 82, 83, 89, 93, 94, 96, 98, 99, 102, 108, 110, 111, 112, 113, 115], "lvl": [25, 26, 42], "inform": [25, 33, 35, 37, 48, 52, 53, 56, 58, 62, 64, 65, 66, 71, 72, 73, 76, 82, 88, 89, 91, 92, 94, 96, 108, 112], "ad": [25, 52, 53, 54, 56, 62, 65, 66, 94, 98], "abov": [25, 54, 56, 62, 65, 66, 73, 81, 82, 89, 94, 101, 103, 113, 115], "msg": [26, 42], "add": [26, 53, 54, 55, 56, 60, 63, 66, 70, 80, 82, 87, 89, 90, 93, 94], "global": [26, 52, 64, 71, 76, 89], "colored_output_on": [27, 42], "prefix": [27, 28, 42, 82], "help": [27, 52, 53, 60, 64, 65, 89, 96, 107, 108, 110, 114], "when": [27, 44, 45, 46, 52, 53, 55, 56, 57, 58, 59, 60, 64, 65, 66, 71, 75, 76, 77, 80, 82, 84, 88, 89, 91, 93, 94, 96, 98, 99, 108, 110, 112, 114], "termin": [27, 52, 89], "If": [27, 33, 53, 54, 55, 56, 62, 63, 64, 65, 66, 68, 71, 72, 76, 80, 82, 89, 90, 91, 93, 94, 96, 99, 100, 104, 108, 111, 112, 113, 114, 116, 117], "build": [29, 30, 35, 49, 52, 53, 57, 59, 60, 63, 64, 65, 71, 75, 76, 81, 86, 89, 91, 93, 94, 101, 103, 108, 112], "post": [29, 30, 49, 52, 63, 69, 89, 96], "train": [29, 30, 49, 52, 69, 70, 89, 90, 96, 108], "quantiz": [29, 30, 52, 64, 69, 74, 76, 89, 97, 109], "creat": [29, 30, 33, 52, 53, 54, 56, 58, 60, 65, 69, 76, 77, 82, 89, 93, 94, 99, 108, 111], "previous": [29, 33, 89, 96, 99], "therefor": [29, 58, 65, 66, 75, 82, 89, 110, 114], "have": [29, 33, 44, 52, 53, 54, 55, 56, 60, 62, 63, 64, 65, 66, 67, 71, 72, 74, 75, 76, 77, 82, 88, 89, 90, 91, 94, 97, 101, 103, 107, 109, 110, 111, 112], "requir": [29, 49, 52, 53, 54, 55, 63, 64, 65, 66, 67, 71, 76, 77, 80, 89, 91, 93, 94, 97, 107, 108, 109, 111, 112, 114], "dataset": [29, 74, 91, 110], "save": [29, 44, 52, 58, 64, 65, 68, 69, 71, 75, 76, 77, 89, 90, 95, 96, 99, 102, 108, 110, 111, 114, 116], "later": [29, 71, 89, 99, 115, 116], "differ": [29, 55, 56, 59, 64, 65, 66, 76, 80, 88, 93, 94, 96, 98, 105, 108, 110, 114, 116], "scratch": [29, 96, 99], "depend": [29, 35, 53, 59, 64, 65, 67, 68, 71, 89, 90, 108, 111, 114], "howev": [29, 66, 80, 81, 89, 93, 94, 96, 111, 112, 116], "network": [29, 30, 54, 60, 65, 76, 89, 91, 93, 94, 108, 110, 111, 117], "also": [29, 53, 54, 60, 62, 64, 66, 68, 80, 82, 83, 89, 90, 91, 96, 104, 107, 110], "recalibr": 29, "its": [29, 53, 56, 58, 60, 66, 75, 76, 82, 94, 107, 111, 114, 116], "structur": [29, 46, 49, 56, 59, 60, 64, 71, 76, 80, 82, 86, 88, 94, 111], "chang": [29, 55, 56, 59, 62, 64, 65, 75, 76, 77, 80, 91, 93, 96, 98, 99, 111, 114, 116], "respons": [29, 54, 58, 82, 114], "ensur": [29, 54, 55, 56, 62, 64, 66, 67, 71, 75, 105, 106], "By": [29, 30, 51, 56, 66, 80, 88, 96, 112], "entropi": [29, 30, 91], "v2": [29, 30, 82], "perform": [29, 30, 54, 62, 63, 71, 75, 76, 91, 94, 104, 108, 110, 111, 113, 114, 115, 116], "recommend": [29, 30, 65, 66, 76, 82, 89, 94, 111, 112], "feed": [29, 30, 89], "forward": [29, 30, 32, 33, 56, 58, 60, 64, 68, 71, 75, 76, 77, 88, 89, 90, 91, 92, 93, 94, 100, 107, 112, 113], "overrid": [29, 30, 44, 54, 65, 91], "minmax": [29, 30, 91], "recomend": [29, 30], "nlp": [29, 30, 91], "task": [29, 30, 65, 91, 110], "call": [29, 30, 32, 49, 54, 55, 58, 60, 65, 71, 72, 75, 76, 77, 82, 88, 89, 92, 93, 94, 96, 98, 100, 103, 110, 112, 114, 116], "make_int8_calibr": [29, 40, 44, 50, 91], "class": [29, 30, 44, 45, 46, 51, 58, 60, 64, 65, 73, 77, 82, 83, 88, 89, 90, 91, 93, 94, 96, 100, 107, 110, 112, 113], "e": [29, 30, 52, 55, 60, 65, 66, 67, 68, 72, 76, 88, 89, 91, 94, 96, 99], "g": [29, 30, 52, 55, 65, 66, 67, 72, 76, 82, 91, 94, 99], "iint8minmaxcalibr": [29, 30, 91], "calibration_cache_fil": [29, 30, 91], "move": [30, 44, 55, 58, 77, 89, 91, 93, 105, 106], "calibration_dataload": [30, 91], "contain": [30, 31, 52, 53, 54, 55, 56, 60, 65, 66, 72, 75, 76, 82, 83, 88, 89, 91, 94, 96, 111, 114], "jit": [31, 32, 33, 34, 45, 47, 49, 52, 53, 55, 56, 57, 58, 59, 60, 61, 64, 68, 69, 71, 75, 76, 77, 88, 89, 90, 92, 94, 99, 111, 115, 116], "modul": [31, 32, 33, 34, 45, 49, 52, 56, 57, 58, 59, 60, 64, 65, 66, 67, 68, 69, 71, 72, 74, 75, 76, 77, 81, 82, 83, 90, 91, 92, 93, 94, 97, 99, 100, 107, 109, 110, 112, 113, 115, 117], "method_nam": [31, 34, 45, 52, 76, 77, 89], "see": [31, 55, 56, 58, 62, 64, 65, 66, 76, 77, 82, 88, 89, 90, 93, 94, 96, 99, 100], "fulli": [31, 52, 55, 64, 71, 75, 76, 77, 89, 91, 94, 117], "compil": [31, 34, 41, 45, 49, 50, 52, 54, 55, 56, 58, 60, 62, 65, 71, 72, 73, 75, 76, 77, 78, 80, 88, 90, 91, 92, 93, 94, 95, 97, 98, 107, 109, 111, 114, 117], "take": [31, 32, 33, 34, 53, 54, 57, 58, 59, 60, 62, 65, 71, 72, 75, 76, 77, 80, 82, 89, 91, 92, 93, 94, 100, 110, 112], "method": [31, 32, 33, 34, 48, 52, 55, 60, 66, 71, 76, 77, 82, 88, 89, 92, 96, 110], "pure": [31, 71, 76], "Will": 31, "out": [31, 44, 53, 55, 56, 57, 59, 60, 64, 66, 71, 76, 77, 82, 89, 94, 98, 107, 108, 111, 112], "unsupport": [31, 49, 54, 64, 76, 94, 116], "script": [31, 55, 56, 68, 76, 77, 88, 89, 90, 92, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 114, 116], "nvidia": [32, 34, 42, 43, 44, 45, 52, 61, 64, 65, 66, 67, 71, 76, 77, 89, 100, 103, 111, 116, 117], "configur": [32, 34, 48, 62, 64, 66, 71, 75, 76, 77, 86, 89, 91, 94, 108, 111, 112], "equival": [32, 57, 59, 60, 71, 76, 77, 88, 89, 91, 93, 94, 101, 103], "specif": [32, 49, 54, 55, 57, 59, 62, 64, 71, 76, 77, 82, 93, 108, 116], "traget": 32, "input_binding_nam": [33, 45, 75, 77], "output_binding_nam": [33, 45, 75, 77], "emb": [33, 52, 63, 77, 83], "pre": [33, 55, 74, 77, 91, 96, 108, 114], "built": [33, 52, 58, 59, 64, 66, 71, 75, 76, 77, 96, 99], "serial": [33, 34, 52, 57, 59, 66, 71, 75, 76, 77, 89, 94, 96, 116], "regist": [33, 54, 58, 60, 65, 75, 77, 93, 94], "execut": [33, 49, 52, 55, 57, 58, 59, 63, 64, 65, 66, 69, 71, 72, 75, 76, 77, 78, 88, 89, 91, 93, 94, 111], "must": [33, 48, 49, 52, 54, 55, 56, 60, 62, 65, 66, 71, 72, 76, 77, 82, 83, 89, 96, 112, 114, 116], "follow": [33, 52, 54, 56, 58, 62, 63, 64, 65, 66, 77, 80, 82, 83, 87, 88, 89, 91, 93, 94, 96, 97, 101, 105, 106, 109, 110, 111, 112, 113, 114], "format": [33, 45, 48, 49, 52, 70, 71, 76, 77, 82, 83, 90, 94, 96, 107, 110, 111, 113, 115], "symbol": [33, 65, 66, 77, 82, 114], "index": [33, 61, 62, 66, 67, 69, 70, 77, 80, 86, 91, 94], "0": [33, 43, 44, 45, 49, 52, 54, 56, 59, 60, 62, 64, 65, 66, 67, 69, 70, 71, 72, 74, 75, 76, 77, 78, 79, 81, 82, 89, 91, 92, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 111, 112, 113, 117], "2": [33, 43, 54, 56, 60, 63, 64, 65, 66, 67, 69, 70, 71, 74, 75, 76, 77, 80, 82, 83, 86, 88, 89, 91, 93, 94, 95, 96, 98, 99, 100, 101, 103, 105, 106, 107, 108, 112, 115], "y": [33, 56, 77, 83, 93, 94, 100], "compilesepc": 33, "order": [33, 49, 54, 56, 60, 62, 65, 66, 71, 72, 75, 76, 77, 89, 90, 93, 96, 113], "pass": [33, 53, 54, 56, 57, 58, 59, 60, 63, 64, 65, 66, 69, 73, 74, 75, 76, 77, 88, 89, 91, 93, 94, 96, 99], "origin": [33, 65, 72, 76, 94, 96, 98, 116], "pytorch": [33, 48, 49, 52, 54, 55, 56, 57, 58, 59, 60, 63, 64, 66, 67, 68, 71, 74, 75, 76, 77, 88, 89, 90, 91, 93, 96, 98, 99, 107, 111, 112, 113, 114, 115, 116], "assum": [33, 75, 92, 94, 97, 109], "convent": 33, "below": [33, 56, 60, 62, 63, 64, 65, 66, 67, 82, 89, 90, 96, 102, 111], "equivil": 34, "librari": [35, 42, 43, 44, 45, 52, 54, 57, 58, 59, 60, 76, 89, 94, 97, 109], "version": [35, 37, 59, 62, 64, 65, 67, 71, 75, 76, 80, 83, 94, 110, 111, 115], "gpu_id": [36, 45, 46, 52, 75, 76, 77, 91, 92, 94, 117], "id": [36, 45, 52, 76, 80, 81, 85, 117], "cudasetdevic": 36, "dump": [37, 52, 94], "base": [37, 50, 58, 63, 64, 66, 71, 72, 76, 82, 88, 90, 91, 95, 99, 103, 110, 116], "stdout": [37, 75], "enginecap": [38, 45, 49, 50, 64, 71, 75, 76, 77, 92, 94], "dump_build_info": [38, 45, 50], "get_build_info": [38, 45, 50], "set_devic": [38, 45, 50, 114], "get_is_colored_output_on": [39, 42, 50], "get_logging_prefix": [39, 42, 50], "get_reportable_log_level": [39, 42, 50], "set_is_colored_output_on": [39, 42, 50], "set_logging_prefix": [39, 42, 50], "set_reportable_log_level": [39, 42, 50], "make_int8_cache_calibr": [40, 44, 50, 91], "check_method_operator_support": [41, 45, 50], "convert_method_to_trt_engin": [41, 45, 50, 76, 77, 89, 92], "embed_engine_in_new_modul": [41, 45, 50, 77], "document": [42, 43, 44, 45, 50, 59, 80, 82, 83, 87, 88, 89, 91, 92, 111, 112, 114], "copyright": [42, 43, 44, 45, 83, 89], "c": [42, 43, 44, 45, 52, 59, 64, 67, 70, 71, 72, 75, 76, 83, 90, 94, 98, 111, 114, 117], "corpor": [42, 43, 44, 45], "right": [42, 43, 44, 45, 55, 59, 60, 82], "reserv": [42, 43, 44, 45, 105, 106], "licens": [42, 43, 44, 45, 89], "under": [42, 43, 44, 45, 59, 65, 82, 93, 101, 116], "bsd": [42, 43, 44, 45], "style": [42, 43, 44, 45, 64, 68, 80, 82, 83], "found": [42, 43, 44, 45, 63, 66, 75, 82, 89, 91, 93, 94, 96, 114], "root": [42, 43, 44, 45, 66, 80, 91, 107], "sourc": [42, 43, 44, 45, 54, 59, 64, 65, 67, 71, 72, 73, 74, 75, 76, 77, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "tree": [42, 43, 44, 45, 80, 91, 107, 114], "pragma": [42, 43, 44, 45, 91], "onc": [42, 43, 44, 45, 53, 55, 56, 58, 64, 65, 66, 67, 76, 91, 94, 106, 108, 111, 114], "namespac": [42, 43, 44, 45, 51, 55, 69, 76, 91, 94], "ar": [42, 46, 49, 52, 53, 54, 55, 56, 58, 59, 60, 62, 63, 64, 65, 66, 71, 74, 75, 76, 77, 80, 82, 83, 84, 88, 89, 91, 92, 93, 94, 95, 96, 98, 99, 101, 105, 106, 108, 110, 111, 112, 113, 114, 115, 116], "ones": [42, 56, 57, 59, 66, 82, 89, 93, 94, 116], "necessari": [42, 62, 64, 66, 75, 93, 99, 114], "user": [42, 48, 54, 56, 57, 58, 59, 62, 63, 64, 66, 67, 71, 82, 83, 89, 90, 91, 93, 96, 99, 108, 111, 112, 113, 114, 116], "dont": 42, "know": [42, 60, 80, 82, 93, 94], "we": [42, 44, 53, 54, 55, 56, 57, 58, 59, 60, 62, 63, 64, 65, 72, 75, 80, 82, 88, 89, 91, 93, 94, 96, 97, 98, 99, 100, 101, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 115, 116], "want": [42, 56, 65, 66, 67, 68, 72, 88, 89, 91, 92, 93, 94, 99, 100, 111], "use_cmake_generated_export_head": 43, "torch_tensorrt_export": 43, "els": [43, 44, 48, 77, 82, 83, 95, 96, 107], "__gnuc__": 43, "__attribute__": 43, "__visibility__": 43, "hidden": [43, 80], "endif": [43, 44, 45], "doe": [43, 44, 55, 56, 60, 62, 65, 66, 76, 82, 91, 94, 101, 103], "gaurd": 43, "someth": [43, 55, 82, 111], "6": [43, 55, 56, 58, 66, 70, 82, 86, 88, 89, 94], "setup": [43, 67, 91, 111], "alias": 43, "eas": 43, "ts": [43, 52, 56, 68, 69, 76, 88, 89, 90, 92, 112, 115], "torchtrt": [43, 56, 94, 107], "ifndef": [44, 45], "doxygen_should_skip_thi": [44, 45], "get_batch_impl": 44, "element_typ": 44, "super": [44, 88, 93, 94, 100, 107, 112, 113], "batchtyp": 44, "dataloader_": 44, "cache_file_path_": 44, "use_cache_": 44, "auto": [44, 56, 60, 64, 68, 71, 82, 83, 89, 91, 105, 106, 108, 117], "batched_data_": 44, "push_back": [44, 56], "it_": 44, "begin": [44, 65, 66, 82, 100, 104], "noexcept": [44, 91], "hack": 44, "explict": 44, "work": [44, 55, 59, 60, 64, 65, 68, 71, 74, 75, 76, 82, 83, 91, 94, 99, 100, 104, 108, 112], "here": [44, 53, 54, 56, 58, 63, 64, 65, 66, 68, 80, 82, 83, 88, 89, 91, 93, 94, 97, 104, 105, 106, 107, 109, 111, 112, 114, 115], "explic": 44, "just": [44, 45, 55, 56, 64, 65, 69, 73, 75, 82, 84, 88, 89, 90, 92, 94, 96, 98, 110, 114], "still": [44, 56, 65, 66, 91, 93, 100, 116], "static_cast": 44, "option": [44, 48, 52, 56, 57, 59, 62, 63, 64, 65, 71, 75, 76, 77, 82, 86, 91, 93, 94, 95, 96, 100, 102, 113, 114, 115, 117], "batch_siz": [44, 91, 107], "end": [44, 52, 60, 62, 70, 71, 76, 77, 82, 89, 91, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108], "statu": [44, 83], "reset": [44, 95, 96, 100, 103, 114], "incas": 44, "go": [44, 55, 56, 65, 68, 88, 89, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 110, 111, 116], "again": [44, 58, 60, 82, 94, 98], "stringstream": 44, "ss": 44, "cache_": 44, "clear": 44, "ifstream": 44, "io": [44, 67, 111], "binari": [44, 91], "noskipw": 44, "good": [44, 60, 65, 82, 96], "copi": [44, 60, 65, 67, 70, 74, 83, 108, 111], "istream_iter": 44, "back_insert": 44, "nullptr": [44, 45, 49], "ofstream": [44, 89], "cache_fil": [44, 74, 91], "reinterpret_cast": 44, "cache_size_": 44, "int8_t": 45, "arrayref": [45, 48, 49], "friend": 45, "ostream": 45, "os": [45, 67, 96], "dtype": [45, 48, 49, 52, 63, 64, 65, 70, 71, 72, 75, 76, 77, 90, 94, 95, 101, 103, 104, 108, 110, 112, 113], "device_typ": [45, 46, 76, 91, 92, 117], "int64_t": [45, 46, 48, 49, 91, 117], "core": [45, 52, 55, 56, 59, 64, 71, 76, 89, 93, 116, 117], "agx": 45, "platform": [45, 52, 59, 66, 67, 111, 117], "xavier": [45, 117], "dla_cor": [45, 46, 52, 76, 91, 92, 117], "allow_gpu_fallback": [45, 46, 71, 76, 77, 91, 92, 117], "customclasshold": [45, 48], "min_shap": [45, 48, 63, 65, 71, 76, 77, 90, 101, 104, 110, 112], "opt_shap": [45, 48, 63, 71, 76, 77, 90, 101, 104, 110, 112], "max_shap": [45, 48, 63, 65, 71, 76, 77, 90, 101, 104, 110, 112], "shape": [45, 47, 48, 49, 52, 56, 60, 63, 65, 69, 70, 71, 72, 75, 76, 77, 78, 90, 93, 94, 97, 104, 107, 108, 109, 111, 114, 117], "doubl": [45, 48, 49, 52, 63, 71, 76, 77, 82, 114], "tensor_domain": [45, 48, 76], "input_is_dynam": 45, "ivalu": [45, 47, 49, 53, 58, 60, 89], "input_signatur": [45, 47, 49, 77, 90], "nest": [45, 49, 50, 82, 83], "full": [45, 49, 52, 60, 64, 71, 73, 76, 89, 91, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 111, 114, 117], "spec": [45, 48, 49, 52, 73, 76, 77, 92, 96], "flatten": [45, 47, 70, 88, 89, 107], "fixed_s": [45, 49], "reflect": [45, 76], "builderconfig": 45, "graph_input": [45, 49], "enabled_precis": [45, 49, 63, 64, 71, 75, 76, 77, 89, 90, 91, 92, 94, 95, 96, 98, 99, 100, 101, 102, 103, 105, 106, 107, 108, 111, 113, 117], "disable_tf32": [45, 49, 64, 71, 75, 76, 77, 91, 94, 105, 106], "sparse_weight": [45, 49, 64, 65, 71, 75, 76, 77, 94], "refit": [45, 49, 64, 69, 71, 76, 77, 92, 94, 96, 97, 98, 109], "truncate_long_and_doubl": [45, 49, 63, 64, 77, 102], "allow_shape_tensor": [45, 49, 77], "uint64_t": [45, 49], "num_avg_timing_it": [45, 49, 64, 71, 75, 76, 77, 92, 94], "workspace_s": [45, 49, 52, 64, 71, 75, 76, 77, 94, 99, 101, 103], "dla_sram_s": [45, 49, 52, 64, 71, 75, 76, 77, 94], "1048576": [45, 49, 64, 71, 75, 76, 77, 94], "dla_local_dram_s": [45, 49, 52, 64, 71, 75, 76, 77, 94], "1073741824": [45, 49, 64, 71, 75, 76, 77, 94], "dla_global_dram_s": [45, 49, 52, 64, 71, 75, 76, 77, 94], "536870912": [45, 49, 64, 71, 75, 76, 77, 94], "require_full_compil": [45, 49, 64, 71, 75, 76, 77, 94], "min_block_s": [45, 49, 56, 63, 64, 71, 75, 76, 77, 93, 94, 95, 96, 99, 100, 101, 103, 107], "3": [45, 49, 52, 55, 56, 58, 63, 64, 65, 67, 68, 70, 71, 74, 76, 77, 82, 83, 86, 88, 89, 91, 92, 94, 95, 96, 98, 99, 101, 104, 105, 106, 107, 108, 110, 112, 115, 117], "torch_executed_op": [45, 49, 56, 63, 64, 71, 75, 76, 77, 94, 99, 100, 101, 103], "torch_executed_modul": [45, 49, 56, 71, 76, 77], "member": [46, 47, 48, 49], "hold": [46, 47, 48, 53, 60, 76, 91], "relat": [46, 82, 100, 103], "let": [46, 52, 55, 60, 65, 71, 76, 77, 80, 82, 110, 111, 116], "layer": [46, 49, 52, 53, 55, 60, 62, 64, 65, 71, 75, 76, 77, 89, 91, 93, 94, 105, 106, 107, 110, 111, 112, 113, 116, 117], "thei": [46, 52, 53, 54, 55, 58, 60, 65, 74, 75, 76, 80, 82, 90, 93, 96], "complex": [47, 49, 64, 66, 88, 90, 98, 106], "either": [47, 48, 52, 60, 62, 71, 76, 77, 80, 82, 88, 89, 90, 93, 94, 96, 115], "one": [47, 54, 55, 60, 64, 65, 67, 71, 75, 76, 82, 88, 89, 90, 93, 94, 100, 103, 105, 106, 111], "rang": [48, 49, 52, 65, 76, 94, 95, 96, 101, 108, 110, 112], "optim": [48, 52, 63, 64, 65, 69, 71, 72, 74, 76, 88, 89, 90, 99, 101, 102, 103, 108, 110, 112, 116], "profil": [48, 72, 75, 113], "singl": [48, 52, 55, 56, 65, 76, 82, 88, 89, 91, 108, 114], "repres": [48, 49, 54, 60, 65, 68, 82], "signifi": [48, 55], "static": [48, 49, 53, 60, 63, 64, 71, 76, 77, 80, 89, 107, 112], "three": [48, 57, 59, 65, 72, 76, 82, 83, 110, 111], "min": [48, 52, 60, 70, 76, 96, 101, 112], "optimin": 48, "max": [48, 52, 60, 70, 76, 80, 96, 101, 107, 112], "allow": [48, 49, 52, 53, 54, 55, 56, 62, 64, 65, 66, 71, 76, 77, 80, 93, 94, 96, 99, 101, 103, 108, 114], "argument": [48, 52, 54, 55, 58, 60, 62, 64, 65, 71, 75, 76, 77, 82, 83, 89, 90, 93, 94, 112], "expect": [48, 54, 55, 60, 76, 89, 90, 110], "tradit": [48, 71, 76, 77, 91], "convect": 48, "produc": [48, 53, 54, 58, 60, 63, 76, 82, 89, 110], "low": [48, 65, 93, 98], "high": [48, 55, 56, 80, 93, 94, 116], "weight": [48, 49, 52, 53, 64, 65, 69, 70, 71, 76, 77, 82, 89, 96, 97, 98, 102, 109, 110], "first": [48, 53, 54, 55, 65, 68, 82, 83, 89, 90, 91, 93, 94, 96, 98, 100, 111, 112, 115, 116], "calcul": [48, 53, 56, 89, 94, 108], "detect": [48, 58, 76], "float32": [48, 49, 52, 63, 64, 65, 71, 76, 77, 94, 98, 102, 105, 106, 108, 112, 113], "dynam": [48, 49, 63, 65, 69, 71, 72, 76, 77, 78, 93, 96, 97, 100, 102, 103, 106, 108, 109, 114], "opt": [48, 66, 75, 76, 104], "minimum": [48, 49, 52, 56, 63, 64, 71, 76, 77, 94, 108], "maximum": [48, 49, 52, 64, 65, 71, 72, 76, 77, 101, 103, 108, 111], "accept": [48, 52, 54, 58, 60, 66, 76, 89, 90, 100, 115], "exampl": [48, 56, 58, 59, 60, 65, 66, 71, 73, 75, 76, 77, 78, 80, 81, 83, 86, 88, 89, 90, 91, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 114, 115], "s": [48, 49, 53, 56, 58, 60, 63, 65, 66, 67, 69, 71, 72, 75, 76, 80, 82, 83, 88, 89, 91, 93, 94, 96, 108, 110, 111, 112, 114, 115], "cannot": [48, 55, 56, 65, 66, 71, 75, 76, 77, 81, 88, 94], "through": [48, 53, 54, 55, 56, 58, 64, 65, 71, 73, 74, 82, 89, 90, 94, 98, 99, 110, 116], "altern": [48, 56, 62, 63, 76, 90, 93, 104, 110, 115], "refer": [48, 54, 57, 59, 65, 81, 86, 89, 91, 94, 107, 111, 112, 115], "given": [48, 49, 52, 54, 55, 65, 71, 72, 74, 76, 77, 88, 89, 90, 92, 93, 112], "kernel": [48, 49, 52, 60, 64, 65, 69, 71, 76, 77, 93, 97, 109, 113, 114], "ani": [48, 52, 53, 54, 60, 62, 64, 65, 70, 71, 74, 75, 76, 77, 80, 82, 89, 90, 91, 93, 94, 101, 112], "event": [48, 64, 95, 96], "place": [48, 55, 62, 65, 82, 83, 84, 91, 94, 107], "variabl": [48, 65, 75, 76], "dimens": [48, 55, 65, 72, 76, 101, 110, 112, 113], "domain": [48, 76, 83, 91], "convien": 49, "fix": [49, 65, 82, 94, 114, 117], "describ": [49, 56, 60, 76, 88, 92, 111], "entri": [49, 60, 96], "okai": 49, "ha": [49, 53, 54, 55, 56, 57, 59, 60, 62, 64, 65, 66, 67, 71, 72, 76, 82, 83, 88, 89, 91, 93, 96, 99, 107, 110, 112, 116], "flaten": 49, "precis": [49, 52, 63, 64, 65, 69, 71, 76, 89, 90, 91, 101, 103, 105, 106, 108, 117], "dure": [49, 52, 54, 56, 60, 63, 64, 71, 74, 76, 91, 93, 105, 106, 108, 110, 112, 114], "prevent": [49, 52, 54, 56], "tf32": [49, 52, 64, 71], "comput": [49, 64, 65, 66, 67, 71, 75, 82, 91, 97, 109, 110], "inner": [49, 83, 110], "product": [49, 67, 76], "round": [49, 71, 76, 77, 94], "10": [49, 66, 67, 71, 72, 76, 77, 86, 88, 89, 91, 107, 108, 110, 111, 112, 113], "bit": [49, 60, 65, 66, 71, 76, 77, 89], "mantissa": [49, 71, 76, 77], "befor": [49, 54, 55, 56, 59, 60, 65, 71, 76, 77, 89, 111, 112], "multipli": [49, 71, 76, 77], "accumul": [49, 64, 71, 76, 77, 105, 106], "sum": [49, 65, 70, 71, 76, 77, 94, 107], "23": [49, 55, 71, 76, 77, 83], "behavior": [49, 56, 65, 71, 76, 77, 93, 105, 106, 112, 114, 115], "sparsiti": [49, 52, 65, 71, 76, 77], "conv": [49, 52, 89, 94], "fc": [49, 52, 55], "truncat": [49, 52, 63, 64, 71, 76, 77], "long": [49, 52, 53, 63, 76, 82, 83], "float": [49, 52, 63, 64, 70, 76, 88, 89, 90, 91, 92, 94, 95, 96, 99, 100, 103, 104, 113], "ishap": 49, "restrict": [49, 64, 71, 76, 77, 112], "cuda": [49, 58, 63, 65, 67, 68, 71, 72, 75, 76, 89, 90, 91, 92, 93, 94, 95, 96, 98, 99, 101, 102, 104, 105, 106, 107, 108, 111, 112, 113, 114, 115], "safeti": [49, 52, 76], "averag": [49, 52, 64, 71, 76, 77, 94], "time": [49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 64, 65, 66, 68, 69, 71, 72, 75, 76, 77, 80, 82, 89, 91, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108], "workspac": [49, 52, 64, 65, 66, 71, 72, 76, 77, 94, 100, 101, 103], "fast": [49, 52, 64, 68, 71, 76, 77], "softwar": [49, 52, 64, 71, 76, 77, 82], "manag": [49, 52, 53, 55, 57, 59, 60, 64, 66, 67, 71, 73, 75, 76, 77, 89, 104, 114], "ram": [49, 52, 64, 71, 76, 77], "commun": [49, 52, 64, 71, 76, 77, 89], "within": [49, 52, 57, 59, 64, 69, 71, 75, 76, 77, 80, 82, 97, 105, 106, 109, 110], "host": [49, 52, 64, 66, 71, 76, 77, 94, 108, 111], "share": [49, 52, 64, 66, 71, 75, 76, 77, 96], "across": [49, 52, 55, 56, 64, 71, 76, 77, 80], "metadata": [49, 52, 54, 58, 60, 64, 71, 76, 77, 80, 99, 112, 113], "quantizatiom": 49, "instead": [49, 52, 53, 54, 55, 66, 71, 75, 76, 89, 93, 99, 107, 114], "potenti": [49, 71, 76, 85], "subgraph": [49, 52, 53, 54, 55, 60, 62, 89, 94, 96, 116], "aten": [49, 54, 55, 56, 60, 61, 64, 69, 70, 71, 76, 77, 89, 93, 100, 116], "thrown": [49, 71, 76, 77], "empti": [49, 71, 72, 76, 77, 83, 88, 94], "torch_tensorrtnamespac": 50, "loggingenum": 50, "levelnamespac": 50, "ptqtemplat": 50, "int8cachecalibratortempl": 50, "int8calibratornamespac": 50, "torchscriptstruct": 50, "compilespecenum": 50, "enginecapabilitystruct": 50, "deviceclass": 50, "devicetypestruct": 50, "graphinputsstruct": 50, "inputclass": 50, "datatypeclass": 50, "cppdirectori": 50, "includedirectori": 50, "torch_tensorrtfil": 50, "hfile": 50, "relationship": 50, "inherit": [50, 65, 71, 91], "subdirectori": 51, "definit": [51, 54, 60, 82], "cli": [52, 90], "It": [52, 54, 55, 56, 57, 59, 60, 65, 66, 69, 76, 80, 82, 94, 108, 110, 114, 116], "serv": [52, 58, 65, 69, 71, 76], "easi": [52, 53, 55, 89, 91], "wai": [52, 64, 65, 66, 88, 89, 91, 93, 94, 96, 99, 110, 114, 115], "command": [52, 64, 66, 82, 83, 88, 89, 111], "line": [52, 66, 83, 89, 98], "quickli": [52, 89, 91], "part": [52, 56, 59, 65, 75, 80, 81, 82, 94, 96], "deploy": [52, 75, 89, 90, 91, 110, 111, 114, 117], "pipelin": [52, 89, 98, 102, 117], "basic": [52, 56, 65, 83, 111], "featur": [52, 56, 65, 66, 89, 91, 92, 102, 107, 108, 110, 116], "though": [52, 59, 60, 88, 89, 116], "alreadi": [52, 53, 54, 55, 89, 91, 93, 94, 97, 109, 112], "two": [52, 55, 60, 62, 64, 65, 66, 76, 82, 83, 87, 88, 90, 91, 93, 96, 111, 112], "embed": [52, 54, 58, 70, 77, 82, 117], "plan": [52, 59, 63, 64, 71], "after": [52, 53, 55, 56, 62, 65, 71, 75, 76, 88, 89, 90, 100, 103, 111, 114], "link": [52, 53, 62, 69, 80, 81, 86, 89, 94, 114], "against": [52, 89, 93], "libtorchtrt": [52, 66, 89], "python": [52, 56, 59, 62, 64, 65, 67, 71, 72, 75, 76, 77, 82, 83, 89, 92, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 114, 117], "import": [52, 55, 56, 63, 64, 65, 66, 67, 68, 75, 80, 82, 88, 89, 90, 92, 93, 94, 95, 96, 98, 111, 112, 114, 115], "packag": [52, 55, 64, 67, 89], "aspect": 52, "ident": [52, 62, 71, 76, 99], "standard": [52, 58, 66, 69, 71, 75, 76, 77, 82, 92, 93, 94, 98, 110, 114], "load": [52, 56, 58, 64, 65, 68, 71, 74, 75, 76, 77, 89, 90, 91, 92, 94, 95, 96, 98, 99, 108, 110, 111, 114, 116], "like": [52, 53, 55, 58, 60, 65, 66, 68, 76, 81, 82, 88, 89, 90, 91, 93, 94, 96, 98, 99, 108, 111, 114], "would": [52, 54, 60, 64, 65, 66, 67, 75, 89, 90, 92, 93, 94, 111, 114], "input_file_path": [52, 117], "output_file_path": [52, 117], "input_spec": [52, 65, 72], "displai": [52, 62, 64, 73, 80, 114], "menu": [52, 80, 82], "verbios": 52, "v": [52, 67, 83, 107, 111], "verbos": [52, 64, 65, 71, 72, 83, 101, 103], "about": [52, 53, 58, 60, 66, 75, 80, 89, 111, 112], "process": [52, 56, 64, 76, 81, 82, 88, 91, 92, 99, 100, 104, 110, 111, 114], "onto": [52, 58], "consol": 52, "w": [52, 66, 76], "disabl": [52, 64, 66, 71, 75, 80, 81, 93, 96, 108, 114], "i": [52, 55, 60, 66, 68, 70, 82, 83, 88, 89, 91, 94, 95, 96, 105, 107], "debugg": [52, 71, 76, 77], "fallback": [52, 57, 59, 60, 99, 117], "model": [52, 56, 58, 63, 68, 71, 72, 73, 74, 76, 78, 88, 89, 90, 91, 92, 95, 96, 98, 112, 114, 116], "throw": [52, 55, 76, 89], "spars": [52, 54, 64, 70, 71], "p": [52, 70, 89, 111, 117], "repeat": [52, 70], "f32": [52, 71, 75, 76, 94], "half": [52, 64, 76, 82, 89, 90, 91, 92, 94, 100, 101, 105, 106, 108, 111, 113, 117], "float16": [52, 76, 94, 98, 102, 113], "f16": [52, 76, 89, 117], "i8": [52, 76], "d": [52, 67, 76, 82, 83, 89, 117], "multi": [52, 75], "dlacor": 52, "avail": [52, 54, 60, 62, 64, 65, 66, 67, 71, 75, 76, 80, 94, 108, 110, 116, 117], "dla_standalon": [52, 76], "file_path": [52, 76, 115], "teo": 52, "op_nam": 52, "op": [52, 53, 54, 55, 56, 57, 59, 60, 62, 63, 64, 75, 76, 89, 93, 100, 114, 116], "partial": [52, 82], "tem": 52, "module_nam": 52, "mod": [52, 56, 65, 71, 86, 89, 91, 113], "mb": [52, 78], "num_op": 52, "block": [52, 53, 55, 56, 64, 71, 86, 116], "treat": 52, "num": 52, "avg": 52, "num_it": 52, "sram": 52, "local": [52, 55, 66, 67, 80, 89], "dram": 52, "atol": 52, "absolut": [52, 66], "toler": 52, "threshold": 52, "numer": [52, 65, 83], "deviat": 52, "1e": [52, 98, 99], "rtol": 52, "rel": [52, 56], "5": [52, 56, 58, 59, 64, 65, 66, 67, 71, 75, 76, 82, 83, 86, 88, 89, 93, 94, 98, 100, 108, 111], "skip": 52, "complianc": 52, "64bit": 52, "32bit": 52, "custom": [52, 62, 63, 65, 66, 69, 97, 105, 106, 109], "dll": 52, "n": [52, 60, 62, 76, 89, 91, 93, 94, 95], "min_n": 52, "min_c": 52, "min_h": 52, "min_w": 52, "opt_n": 52, "opt_c": 52, "opt_h": 52, "opt_w": 52, "max_n": 52, "max_c": 52, "max_h": 52, "max_w": 52, "32": [52, 76, 88, 89, 90, 91, 105, 106, 107, 117], "flag": [52, 56, 57, 59, 64, 66, 71, 74, 76, 90, 104, 105, 106, 114, 115], "forc": [52, 63, 65, 71, 76, 77, 80], "posit": [52, 54, 65, 76, 80], "test": [52, 56, 59, 65, 66, 67, 71, 76, 82, 83, 91, 107, 110, 111], "ssd_trace": 52, "pt": [52, 65, 89, 105, 106, 111], "ssd_trt": 52, "300": [52, 92], "512": [52, 71, 76, 77, 107, 110], "1024": [52, 71, 76, 77, 105, 110], "simplifi": [53, 94], "form": [53, 75, 76, 82, 90, 111], "up": [53, 55, 56, 57, 58, 59, 62, 65, 66, 71, 76, 82, 88, 93, 94, 96, 99, 100, 103, 108, 110], "context": [53, 57, 58, 59, 64, 73, 75, 93, 104, 114], "inetworkdefinit": [53, 54], "record": [53, 88, 95, 96, 104, 114], "togeth": [53, 60, 89], "start": [53, 56, 65, 70, 74, 76, 83, 89, 92, 94, 95, 96, 110], "look": [53, 54, 55, 68, 71, 76, 88, 91, 92, 93, 96, 111, 112], "assembl": [53, 62, 89], "resourc": [53, 91, 94], "coupl": [53, 59, 65, 114], "state": [53, 54, 60, 62, 75, 89, 93, 98], "been": [53, 60, 64, 66, 67, 74, 83, 89, 96, 99, 116], "evaluated_value_map": [53, 60], "stage": [53, 65], "arg": [53, 54, 62, 65, 71, 74, 75, 76, 86, 89, 93, 94, 96, 107, 110], "itensor": [53, 54, 60, 65, 89, 93, 94], "value_tensor_map": [53, 60], "typic": [53, 60, 76, 111], "abl": [53, 55, 60, 62, 65, 91, 92, 94, 99], "system": [53, 60, 62, 64, 69, 71, 75, 76, 77, 93, 94, 96, 99, 116], "registri": [53, 54, 89, 94], "enter": [53, 76], "recurs": 53, "resolv": [53, 55, 57, 59, 100, 103], "until": [53, 56, 59, 60, 66, 71, 76, 116], "final": [53, 56, 57, 59, 66, 93, 94, 100, 103, 110], "some": [53, 54, 55, 56, 57, 58, 59, 60, 62, 63, 65, 66, 76, 81, 82, 89, 91, 93, 94, 96, 112, 116], "These": [53, 54, 56, 58, 62, 64, 66, 71, 74, 75, 76, 80, 82, 91, 93, 111, 116], "those": [53, 54, 62, 64, 82], "do": [53, 54, 55, 56, 60, 63, 65, 81, 83, 88, 89, 90, 91, 93, 94, 105, 106, 117], "theori": [53, 82], "kind": [53, 65], "common": [53, 55, 65, 72, 82, 93, 96], "prim": [53, 55, 56, 58, 70, 88, 89], "constant": [53, 54, 55, 56, 89, 94], "emit": 53, "listconstruct": [53, 56, 58, 89], "make": [53, 54, 65, 66, 67, 71, 76, 82, 84, 89, 90, 91, 94, 96, 110, 111, 117], "associ": [53, 60, 89, 96, 114], "where": [53, 54, 55, 60, 62, 64, 65, 71, 75, 76, 77, 83, 89, 91, 93, 99], "result": [53, 55, 56, 66, 68, 71, 73, 75, 76, 77, 80, 88, 90, 94, 98, 99, 108, 111, 113, 116], "done": [53, 56, 59, 94, 99, 111, 115], "mai": [53, 54, 56, 58, 59, 65, 66, 71, 75, 76, 77, 82, 83, 88, 89, 90, 91, 93, 94, 99, 100, 103, 108, 111, 114], "For": [53, 56, 62, 63, 64, 65, 66, 68, 72, 76, 80, 82, 83, 88, 89, 91, 92, 93, 94, 98, 100, 107, 110, 111, 114, 115], "more": [53, 64, 65, 66, 67, 69, 71, 76, 80, 83, 88, 89, 90, 91, 92, 94, 96, 98, 101, 103, 111, 114], "writing_convert": [53, 89], "locat": [54, 62, 66, 91, 93, 94], "py": [54, 55, 59, 62, 65, 66, 67, 78, 80, 82, 87, 88, 89, 91, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 112], "convers": [54, 55, 56, 58, 63, 64, 65, 71, 76, 77, 89, 93, 94, 110, 112], "decror": 54, "dynamo_tensorrt_convert": [54, 93, 94], "signatur": [54, 77], "leaky_relu": [54, 70], "def": [54, 62, 65, 82, 88, 90, 93, 94, 95, 96, 100, 107, 108, 111, 112, 113], "leaky_relu_convert": 54, "ctx": [54, 60, 89, 93, 94, 108], "conversionctx": [54, 60, 89, 93], "tupl": [54, 58, 63, 65, 71, 72, 75, 76, 77, 90, 93, 94, 96, 99, 112, 113], "kwarg": [54, 65, 71, 74, 75, 76, 93, 94, 110], "dict": [54, 71, 75, 76, 77, 93, 94, 96], "union": [54, 60, 64, 71, 75, 76, 77, 89, 93], "sequenc": [54, 62, 65, 71, 72, 75, 76, 77, 82, 93, 94, 108, 110], "decor": [54, 62, 65, 93], "kei": [54, 82, 88, 96, 111, 112], "node": [54, 55, 56, 57, 59, 60, 62, 64, 65, 71, 72, 89, 93, 94, 107, 110, 112], "capability_valid": [54, 93], "lambda": [54, 60, 82, 89, 93, 111], "fx": [54, 62, 63, 71, 75, 76, 89, 90, 93, 94, 99, 115], "determin": [54, 55, 64, 65, 76, 93, 108, 112, 114], "properli": [54, 66], "handl": [54, 55, 56, 58, 64, 65, 75, 76, 94], "partition": [54, 71, 76, 94], "sure": [54, 66, 67, 89, 90, 111, 117], "prioriti": [54, 93], "develop": [54, 65, 66, 67, 69, 82, 83, 89, 93, 94], "bodi": [54, 82, 83], "nativ": [54, 59, 61, 89, 93, 94, 99], "numpi": [54, 76, 94, 95, 96, 98, 99, 108, 111], "frozen": 54, "attribut": [54, 55, 56, 58, 65, 76, 82, 89], "previou": [54, 80, 100], "correspond": [54, 60, 65, 66, 75, 76, 93, 96, 98, 107, 114], "edg": [54, 82], "well": [54, 63, 66, 69, 73, 75, 82, 89, 91, 93, 96, 104, 115], "being": [54, 65, 66, 71, 89, 93, 94, 99], "truth": 54, "http": [54, 61, 64, 66, 67, 80, 82, 88, 89, 91, 93, 94, 98, 100, 103, 107, 110, 111, 112, 114], "github": [54, 61, 64, 66, 67, 80, 89, 91, 100, 103, 107, 111, 114], "com": [54, 61, 64, 66, 67, 89, 91, 98, 100, 103, 107, 111, 114], "blob": [54, 61, 66, 80, 91, 96], "main": [54, 55, 56, 57, 58, 59, 60, 63, 65, 66, 80, 82, 84, 89, 93, 94, 105, 107], "src": [54, 58, 61, 70], "native_funct": [54, 61], "yaml": [54, 61], "sinc": [54, 55, 64, 65, 67, 75, 82, 88, 89, 91, 93, 95, 96, 99], "mani": [54, 56, 64, 65, 80, 82, 83, 93, 96, 99, 116], "composit": [54, 89], "raw": [54, 80, 93], "impl": [54, 93], "subpackag": 54, "chain": [54, 60], "primarili": [54, 59, 66, 89, 93], "manipul": [54, 62, 76], "net": [54, 60, 82, 83, 89, 94], "addit": [54, 55, 64, 65, 75, 76, 89, 93, 94, 96, 99, 110, 112], "call_modul": 54, "call_funct": [54, 62, 65], "eg": [54, 111, 113], "aten_": 54, "_leaky_relu": 54, "opoverloadpacket": 54, "while": [54, 56, 66, 75, 91, 93, 98, 108, 110, 111, 114, 116], "opoverload": 54, "particular": [54, 64, 96], "collect": [54, 56, 64, 71, 76, 77, 89, 90, 107], "trtinterpret": [54, 65, 72], "along": [54, 76], "match": [54, 55, 93, 99], "special": [54, 56], "account": [54, 111], "illustr": [54, 65, 101, 105, 106, 110], "scale_grad_by_freq": [54, 70], "embedding_param_valid": 54, "establish": 54, "subset": [54, 64, 71, 76, 91, 110], "converter_util": [54, 94], "enforce_tensor_typ": 54, "dictionari": [54, 76, 77, 92, 100], "between": [54, 55, 56, 60, 66, 76, 82, 83, 91, 96, 98, 105, 108], "possibl": [54, 66, 82, 93, 94, 96, 110, 111], "prefer": [54, 64, 66, 89], "keyword": [54, 62, 71, 75, 76, 77, 93, 100, 103], "both": [54, 56, 64, 66, 69, 71, 72, 75, 76, 80, 82, 88, 91, 93, 94, 96], "enforc": [54, 89], "situat": 54, "partit": [54, 55, 63, 64, 71, 76, 93, 116], "greater": [54, 71, 73, 76], "than": [54, 55, 64, 66, 71, 76, 81, 82, 93, 95, 96, 98, 108, 110, 114], "3d": [54, 65], "autocast": 54, "therebi": [54, 58, 94, 110], "limit": [54, 55, 73, 81, 91, 96, 108, 116], "author": [54, 83], "conv_nod": 54, "7": [54, 56, 58, 59, 75, 76, 86, 89, 94, 100, 101, 103, 107, 112], "ignor": [54, 71, 75, 76, 94], "misc": [54, 94], "trttensor": 54, "np": [54, 93, 94, 95, 96, 98, 99, 108, 111], "ndarrai": [54, 94], "aten_ops_convolut": 54, "conversioncontext": [54, 93, 94], "side": [54, 55, 80, 89, 93], "effect": [54, 55, 64, 65, 71, 80, 89, 91, 93, 94, 110], "term": [54, 76, 82, 83, 91, 93, 94, 110], "getitem": 54, "categor": 54, "modif": [54, 62, 76], "op_evalu": 54, "capbility_valid": 54, "opcod": 54, "decompos": 54, "suboper": 54, "separ": [54, 56, 57, 59, 66], "Such": 54, "via": [54, 64, 65, 67, 69, 71, 75, 76, 77, 80, 86, 90, 91, 100, 101, 103, 105, 106, 110, 112, 114, 115, 116], "register_torch_trt_decomposit": 54, "addmm_replac": 54, "replac": [54, 56, 62, 66, 67, 74, 94, 107, 116], "input_": 54, "mat1": 54, "mat2": [54, 70], "beta": [54, 65, 70, 77], "alpha": [54, 65, 70, 83], "mul": [54, 56, 70, 93], "matmul": [54, 55, 64, 70, 71, 89, 105, 106, 112], "modifi": [54, 56, 62, 65, 83, 98, 112], "edit": [54, 66, 80], "torch_enabled_decomposit": 54, "torch_disabled_decomposit": 54, "disjoint": 54, "preced": [54, 82], "over": [54, 57, 59, 65, 82, 107, 108, 111, 116], "much": [54, 60, 80, 82, 91], "significantli": [54, 55, 80, 96], "easier": [54, 57, 59, 60, 65, 71, 75, 76, 89, 91, 94, 98], "tri": 54, "made": [55, 57, 59, 76, 82], "represent": [55, 60, 65, 88, 110, 116], "instanc": [55, 62, 64, 66, 71, 74, 75, 88, 89, 93, 110, 114], "idea": [55, 82, 93], "reduc": [55, 56, 57, 59, 65, 71, 76, 91, 94, 96, 110, 114], "actual": [55, 58, 60, 65, 88, 89, 94], "aim": [55, 116], "closer": 55, "scope": [55, 94, 100, 103], "csrc": [55, 61], "common_subexpression_elimin": 55, "subexpress": 55, "dead_code_elimin": 55, "exception_elimin": 55, "wa": [55, 58, 62, 64, 65, 71, 75, 76, 82, 89, 93, 116], "1013": 55, "ne": [55, 70], "1012": 55, "24": [55, 67], "lib": [55, 66, 67, 89], "python3": [55, 66, 89], "site": [55, 66, 82, 89], "nn": [55, 61, 65, 71, 72, 75, 76, 77, 88, 89, 90, 93, 94, 100, 107, 112, 113, 116], "batchnorm": 55, "248": 55, "11": [55, 66, 82, 86, 89, 111], "block0": 55, "raiseexcept": 55, "249": 55, "12": [55, 56, 67, 82, 86, 88, 89, 101, 111, 112], "block1": 55, "guard_elimin": 55, "whose": [55, 65, 101], "freeze_modul": 55, "propag": 55, "fuse_addmm_branch": 55, "variant": [55, 114], "caught": 55, "ret": 55, "622": 55, "self": [55, 58, 60, 70, 75, 76, 88, 89, 90, 93, 94, 96, 100, 107, 110, 112, 113, 117], "bia": [55, 70, 89, 107], "x9": 55, "3677": 55, "output0": [55, 113], "add_": [55, 70, 89, 93], "fuse_linear": 55, "back": [55, 56, 58, 59, 75, 76, 82, 88, 89, 94, 116], "fuse_flatten_linear": 55, "implicitli": [55, 76], "connect": [55, 71, 76, 77, 82, 98, 111, 117], "higher": [55, 64, 71, 76, 80, 82, 88, 108], "1d": 55, "lower_graph": 55, "access": [55, 60, 65, 80, 89, 92, 116], "rather": 55, "getattr": [55, 58, 88, 89], "trainabl": 55, "remain": [55, 76, 91, 116], "lower_tupl": 55, "lowersimpletupl": 55, "tupleconstruct": [55, 58], "tupleunpack": 55, "leav": [55, 62, 64, 71], "statement": [55, 82, 93], "loweralltupl": 55, "_all_": 55, "rais": [55, 65, 76], "onnx": 55, "module_fallback": 55, "consist": [55, 65, 82, 94, 114, 116], "pair": [55, 60, 66, 82, 91, 110], "delimit": 55, "around": [55, 58, 60, 64, 66, 71, 75, 82, 85, 88, 94], "second": [55, 65, 82, 90, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108], "mark": [55, 56, 80, 96], "notatemoduleforfallback": 55, "marknodesforfallback": 55, "tell": [55, 56, 57, 58, 59, 60, 82, 116], "them": [55, 56, 58, 63, 64, 65, 66, 71, 75, 80, 89, 94, 96, 106, 110, 112, 116], "peephole_optimz": 55, "intent": [55, 82], "catch": [55, 76, 89], "small": [55, 94, 95, 111], "might": [55, 66, 80, 99, 112], "interest": [55, 82], "now": [55, 56, 59, 60, 65, 66, 76, 82, 89, 92, 93, 94, 96, 99, 108, 113, 114], "expand": [55, 70], "simpli": [55, 100, 110], "remove_contigu": 55, "remove_dropout": 55, "infer": [55, 64, 65, 71, 76, 77, 89, 91, 99, 100, 108, 110, 112, 114, 115, 116], "remove_to": 55, "unpack_addmm": 55, "reus": [55, 65, 91, 96], "dedic": [55, 83], "unpack_log_softmax": 55, "softmax": [55, 65, 70, 107], "loop_unrol": 55, "suffici": [55, 66, 76], "short": [55, 64, 71, 82, 83, 99], "tile_to_repeat": 55, "instruct": [56, 57, 59, 65, 66, 89, 111], "criteria": [56, 57, 59, 64], "lack": [56, 57, 59, 65, 94, 108], "explicitli": [56, 57, 59, 66, 77, 90, 91, 92, 105, 106, 113], "On": 56, "segment": [56, 63, 94, 101, 103, 110], "verifi": [56, 71, 93, 94, 99], "Then": [56, 91, 92, 99], "roughli": 56, "analysi": 56, "everi": [56, 72, 75, 76, 89, 114], "complet": [56, 63, 71, 76, 88, 89], "mean": [56, 60, 65, 70, 72, 100, 108, 111, 116], "trace": [56, 65, 71, 75, 77, 88, 89, 112, 115, 116], "tensorlist": [56, 60], "figur": [56, 83, 85], "our": [56, 59, 63, 88, 89, 111], "stitch": [56, 89], "altogeth": [56, 80], "brief": 56, "descript": [56, 83, 107], "partitioninfo": 56, "api": [56, 59, 60, 62, 63, 64, 65, 75, 76, 77, 81, 89, 90, 91, 92, 94, 100, 101, 104, 108, 110, 111, 112, 114, 115], "maintain": [56, 58, 60, 76, 98, 116], "code": [56, 59, 62, 64, 65, 66, 81, 83, 88, 89, 91, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 112], "mymodel": [56, 63, 68, 90, 94, 112, 115], "ts_model": [56, 89], "trt_model": [56, 92, 94, 101, 105, 106, 107, 108, 111, 115], "off": [56, 58, 104], "consecut": [56, 63], "satisfi": [56, 62, 65], "forced_fallback_op": 56, "randn": [56, 63, 68, 71, 76, 77, 89, 92, 93, 96, 101, 104, 112, 113, 115], "224": [56, 63, 68, 71, 72, 76, 77, 89, 96, 98, 99, 101, 104, 110, 111, 112, 115], "trt_ts_modul": [56, 90], "input_s": 56, "inputrang": 56, "cfg": [56, 89], "relu": [56, 70, 88, 89, 100, 107], "trt_mod": [56, 68, 89, 91, 117], "consid": [56, 77, 89, 94, 113], "segmentmodelwithdependencyawar": 56, "test_segment": 56, "20": [56, 67, 86, 99, 101, 103], "x_lgamma": 56, "lgamma": 56, "y_lgamma": 56, "div": [56, 70], "div_lgamma": 56, "27": [56, 89], "cat": [56, 66, 67, 70, 107, 108], "greedi": [56, 105, 106, 108], "strategi": [56, 76], "travers": [56, 57, 59, 64], "gather": 56, "same": [56, 58, 62, 64, 65, 66, 71, 76, 80, 82, 88, 89, 92, 94, 96, 99, 101, 103, 111, 112, 114, 115], "encount": [56, 64, 66, 93, 100, 103], "4": [56, 58, 63, 64, 65, 66, 70, 76, 78, 80, 82, 83, 86, 89, 94, 100, 102, 103, 104, 107, 112], "suboptim": 56, "arithmet": 56, "split": [56, 65, 70], "own": [56, 60, 64, 66, 71, 82, 89, 96, 107, 111], "could": [56, 64, 65, 94, 101, 103, 114], "rewrit": [56, 62], "portion": [56, 82, 94, 102], "without": [56, 60, 68, 71, 80, 82, 89, 91, 94, 95, 96, 99, 114], "reorder": 56, "seri": 56, "cleanli": 56, "approach": [56, 96], "achiev": [56, 110], "hit": 56, "larger": [56, 71, 76, 80, 108, 110], "boundari": [56, 74, 76], "guarante": [56, 75], "trigger": [56, 64, 65, 76, 89, 96, 98, 99, 116], "appear": [56, 82], "adjac": [56, 71, 76, 82], "As": [56, 65, 66, 76, 89, 93, 94, 96, 99, 116], "clean": [56, 62, 82, 100, 103], "step": [56, 65, 67, 70, 76, 91, 94, 99, 110], "consolid": [56, 88], "further": [56, 64, 65, 114, 116], "merg": 56, "identifi": 56, "do_not_merg": 56, "combin": [56, 64, 65], "condit": [56, 82, 116], "loop": [56, 64, 65, 105, 106], "ir": [57, 59, 60, 63, 64, 68, 71, 76, 88, 89, 90, 97, 100, 101, 103, 104, 109, 112], "larg": [57, 59, 80, 82, 89, 91, 99, 108, 110], "opset": [57, 59, 93], "compon": [57, 59, 66, 67, 74, 88, 114, 116], "evalu": [57, 58, 59, 107], "deploi": [57, 59, 69, 89, 91, 97, 109, 111], "instanti": [57, 58, 59, 60, 89, 102], "wrap": [57, 58, 59, 65, 82, 85, 89, 92, 100, 103], "extend": [57, 59, 60, 70, 89, 96, 110], "providi": [57, 59], "stand": [58, 82], "interpret": [58, 65, 82], "execute_engin": [58, 75, 89], "stack": [58, 70, 91, 107, 116], "machin": [58, 66, 91, 111], "pop": 58, "push": 58, "element": [58, 65, 82, 83, 86], "realiz": 58, "abstract": [58, 60, 83, 93], "__torch__": [58, 88, 89], "portabl": [58, 66, 77], "serializ": [58, 64, 88, 116], "instnanti": 58, "whatev": [58, 65, 94], "self_1": [58, 89], "torchvis": [58, 91, 92, 96, 98, 99, 101, 104, 107, 111], "resnet": [58, 69, 78, 97, 98, 109, 110, 111], "___torch_mangle_4847": 58, "resnet_trt": 58, "input_0": [58, 89], "__torch___torchvision_models_resnet____torch_mangle_4847_resnet_trt_engin": 58, "listunpack": [58, 89], "multipl": [58, 66, 71, 75, 76, 82, 83, 91, 108, 111, 114], "repack": 58, "ssd": 58, "ssd300_trt": 58, "__torch___pytorch_detection_ssd_src_model_ssd300_trt_engin": 58, "holder": [58, 84], "torchbind": 58, "pickler": 58, "seril": 58, "zip": [58, 66, 98, 99, 109], "depickl": 58, "encod": [58, 110], "sm": 58, "correct": [58, 66, 80, 97, 98, 99, 107, 109], "bazel": [59, 66, 67], "linux": [59, 67, 89], "x86_64": [59, 66], "aarch64": 59, "gcc": [59, 89], "untest": 59, "try": [59, 76, 82, 83, 89, 92, 94, 96, 116], "older": 59, "repositori": [59, 66, 80, 87, 111], "notebook": [59, 69, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "doc": [59, 61, 66, 67, 80, 81, 82, 87, 93, 94, 112], "docsrc": 59, "third_parti": [59, 66], "toolchain": [59, 66, 67], "unstabl": 59, "subject": [59, 62, 116], "matur": 59, "most": [59, 65, 66, 72, 94, 99, 111, 114, 116], "hood": [59, 101, 116], "major": [59, 65, 76], "top": [59, 80, 84], "coordin": [59, 76], "ingest": 59, "flow": [60, 65, 82, 88, 110], "ilay": 60, "analogu": 60, "goal": [60, 64, 96], "registernodeconversionpattern": [60, 89], "helper": [60, 93], "pattern": [60, 76, 89, 108], "schema": [60, 89, 93, 94], "caus": [60, 64, 80, 100, 101, 103, 108, 114], "acthardtanh": 60, "torchtrt_unus": 60, "hardtanh": [60, 70], "scalar": [60, 70], "min_val": [60, 70], "max_val": [60, 70], "unwraptodoubl": 60, "new_lay": 60, "addactiv": 60, "activationtyp": [60, 65], "kclip": 60, "torchtrt_check": 60, "unabl": [60, 89, 94], "setalpha": 60, "setbeta": 60, "setnam": [60, 89], "util": [60, 62, 74, 77, 89, 91, 100, 103, 105, 106, 107, 108, 110, 111, 116], "node_info": [60, 89], "c_str": [60, 89], "out_tensor": [60, 89], "associatevalueandtensor": [60, 89], "getoutput": [60, 89], "log_debug": 60, "getdimens": [60, 89], "accord": [60, 64, 77], "unwrap": 60, "tool": [60, 64, 65, 66, 89, 93, 96, 110], "don": [60, 65, 80, 82, 83, 91, 93, 107, 111, 112], "annot": [60, 89], "your": [60, 63, 64, 66, 67, 68, 75, 80, 82, 83, 87, 88, 89, 90, 92, 96, 112, 114], "Its": [60, 82], "track": [60, 91], "sort": [60, 70, 92], "live": [60, 82], "directli": [60, 62, 63, 66, 69, 74, 76, 91, 93, 94, 100, 115], "associatevalueandivalu": 60, "inspect": [60, 88, 89], "dataflow": [60, 89], "mechan": [60, 64, 65, 94, 99, 110], "safe": [60, 64, 71, 75, 76, 77], "unsur": 60, "deep": [60, 64, 69, 80, 91, 94, 117], "straight": 60, "chanc": 60, "none": [60, 64, 65, 70, 71, 72, 74, 75, 76, 77, 80, 82, 93, 94, 96, 100, 107, 108], "wrapper": [60, 65, 115], "similar": [60, 63, 64, 65, 66, 89, 92, 94, 105, 106], "tocustomclass": 60, "tensorcontain": 60, "istensor": 60, "iscustomclass": 60, "lot": [60, 63], "singular": 60, "becaus": [60, 65, 66, 72, 88, 89, 93, 94, 95, 96, 108, 113], "alloc": 60, "freed": 60, "destructor": 60, "destroi": [60, 83], "realli": 60, "think": [60, 82], "becom": [60, 66, 98], "benefit": [60, 89, 96, 108], "deal": [60, 96], "quit": [60, 66, 89, 110], "effici": 60, "batch_norm": [60, 70], "fusion": [60, 62, 65], "deeplearn": [61, 65, 67], "sdk": [61, 67, 116], "matrix": 61, "html": [61, 66, 67, 82, 88, 91, 93, 94, 112], "c_api": 61, "python_api": 61, "org": [61, 66, 80, 82, 88, 89, 91, 93, 94, 112, 114], "stabl": [61, 67, 69, 77, 78, 80, 97, 109, 112], "master": [61, 66, 91, 114], "overview": [61, 69, 100, 104], "md": 61, "appli": [62, 63, 91, 99], "desir": [62, 71, 83, 91, 96], "coalesc": 62, "insert": [62, 64, 71, 89, 91, 93, 96, 99], "graphmodul": [62, 63, 71, 72, 76, 89, 90, 94, 99, 115, 116], "caller": 62, "invok": [62, 64, 65, 88, 89, 114], "lint": 62, "recompil": [62, 71, 76, 93, 96, 99, 103, 112, 116], "repair": 62, "disallow": 62, "repair_input_as_output": 62, "gm": [62, 71], "sample_input": [62, 65, 100], "scenario": [62, 64, 98, 108], "clone": [62, 66, 70, 94], "modified_graph": 62, "extract": [62, 89, 110], "placehold": [62, 93], "isinst": [62, 65, 94, 107], "issubclass": 62, "direct": [62, 86, 99, 114], "len": [62, 70, 94], "direct_output": 62, "inserting_aft": 62, "cloned_placehold": 62, "replace_input_with": 62, "date": [62, 83, 116], "eliminate_dead_cod": 62, "logger": [62, 73], "f": [62, 64, 65, 67, 76, 82, 88, 93, 94, 107, 108], "__init__": [62, 75, 76, 82, 88, 93, 94, 96, 100, 107, 112, 113], "pass_manag": 62, "passmanag": 62, "backend": [62, 68, 69, 77, 78, 81, 92, 95, 96, 97, 100, 107, 109, 112], "offer": [62, 64], "registr": [62, 65], "conveni": [62, 91, 103, 110, 114, 116], "control": [62, 65, 88, 99, 108, 114], "_aten_lowering_pass": 62, "my_custom_pass": 62, "front": [62, 71], "passlist": 62, "arbitrari": [62, 75], "remov": [62, 63, 71, 80, 95, 96, 107], "dump_lowering_pass": 62, "apply_lowering_pass": 62, "graph_modul": [62, 71], "_remove_lowering_pass": 62, "evolv": 62, "introduc": [63, 65, 110], "exportedprogram": [63, 68, 71, 76, 99, 105, 106, 108, 112, 116], "dynamo": [63, 64, 66, 68, 74, 75, 76, 78, 89, 93, 94, 95, 96, 97, 99, 100, 101, 103, 104, 107, 108, 109, 112, 113], "frontend": [63, 71, 74, 90, 94, 97, 101, 103, 107, 109, 112], "simpl": [63, 64, 65, 82, 83, 88, 110, 111, 112], "usag": [63, 65, 69, 74, 78, 82, 89, 97, 108, 109, 112, 115], "eval": [63, 68, 89, 90, 93, 95, 96, 98, 99, 100, 101, 103, 104, 105, 106, 107, 108, 111, 112, 113, 115], "exp_program": [63, 96, 99, 107, 112], "trt_gm": [63, 68, 96, 99, 112, 113, 115], "interact": [63, 82, 98, 100, 101, 102, 103, 104], "ideal": 63, "discuss": [63, 64, 111], "section": [63, 65, 80, 82, 83, 84, 86, 89, 91, 111, 115], "frequent": 63, "builder": [63, 64, 65, 71], "respect": [63, 64, 66, 71, 76, 105, 106, 113], "releas": [63, 64, 67, 82], "insid": [63, 82, 94, 111], "decomposit": [63, 64, 71, 76, 94], "downstream": [63, 110], "constraint": [63, 108], "guid": [64, 81], "present": [64, 99], "learn": [64, 66, 69, 89, 91, 94, 111, 117], "acceler": [64, 72, 76, 114, 116, 117], "workflow": [64, 65, 68, 69, 71, 72, 76, 89, 92, 96, 97, 98, 101, 102, 103, 105, 106, 109, 110], "wide": [64, 76, 86], "varieti": [64, 111], "primari": [64, 93, 96, 115], "simplic": 64, "optimized_model": [64, 68, 95, 100, 101, 103], "depth": [64, 80, 110], "challeng": [64, 98, 111], "addition": [64, 94], "fit": [64, 82], "compilationset": [64, 71, 75, 93, 94, 100], "_enum": [64, 71], "callabl": [64, 71, 76], "pass_through_build_failur": [64, 71, 75, 76, 94], "max_aux_stream": [64, 71, 75, 76, 94], "version_compat": [64, 71, 75, 76, 94], "optimization_level": [64, 71, 75, 76, 94, 100], "use_python_runtim": [64, 71, 75, 76, 94, 95, 96, 98, 99, 100], "truncate_doubl": [64, 71, 75, 76, 94, 95, 105, 106, 108], "use_fast_partition": [64, 71, 75, 76, 94], "enable_experimental_decomposit": [64, 71, 75, 76, 94], "_devic": [64, 71], "assume_dynamic_shape_support": [64, 71, 75, 76], "make_refitt": [64, 71, 75, 76, 95, 96, 98, 99], "engine_cap": [64, 71, 75, 76, 94], "dryrun": [64, 71, 75, 76, 94], "hardware_compat": [64, 71, 75, 76, 94], "timing_cache_path": [64, 71, 75, 76, 96], "tmp": [64, 71, 75, 76, 89, 95], "torch_tensorrt_engine_cach": [64, 71, 75, 76], "timing_cach": [64, 65, 71, 75, 76], "bin": [64, 66, 67, 71, 75, 76], "lazy_engine_init": [64, 71, 75, 76], "cache_built_engin": [64, 71, 75, 95, 96], "reuse_cached_engin": [64, 71, 75, 95, 96, 99], "use_explicit_typ": [64, 71, 75, 105, 106, 108, 113], "use_fp32_acc": [64, 71, 75, 105, 106], "enable_weight_stream": [64, 71, 75, 108], "dpython": [64, 71, 76, 77], "per": [64, 71, 94, 114], "regardless": [64, 71, 83, 101, 103], "fail": [64, 71, 76, 89, 98, 99, 107, 117], "auxiliari": [64, 71], "stream": [64, 69, 71, 76, 94, 97, 109], "impli": [64, 71], "longer": [64, 66, 71, 76, 80, 114], "search": [64, 69, 71, 76, 80], "strictli": [64, 71], "runtim": [64, 66, 68, 69, 71, 76, 89, 93, 98, 100, 103, 104, 108, 116], "presenc": [64, 71], "preferenti": [64, 71], "choos": [64, 65, 71, 88], "float64": [64, 71, 76, 77], "refitt": [64, 71, 96], "toggl": [64, 71, 76], "mode": [64, 65, 71, 75, 76, 90, 91, 93, 104, 107], "detail": [64, 65, 67, 71, 88, 89, 94, 96, 111, 114], "natur": [64, 71, 82], "architectur": [64, 66, 69, 71, 76, 96, 110], "amper": [64, 71, 76], "newer": [64, 66, 71, 76], "storag": [64, 71, 91], "use_strong_typ": [64, 71], "strong": [64, 71, 82], "mix": [64, 69, 71], "happen": [64, 65, 71, 88, 98, 101, 112], "sub": [64, 70, 82, 88, 100], "slate": 64, "futur": [64, 65, 71, 76, 77, 114], "occur": [64, 108], "first_output": 64, "subsequ": [64, 96], "second_output": 64, "session": [64, 68, 82, 96, 104], "point": [64, 66, 76, 80, 81, 82, 89, 107, 111], "cover": [64, 93], "benchmark": [64, 70], "automat": [64, 67, 76, 82, 89, 99, 112, 116], "vari": [64, 72, 108, 112], "distribut": [64, 67, 89, 91, 108, 114], "inf": 64, "dynamo_convers": 64, "contribut": 64, "demonstr": [64, 82, 83, 84, 91, 93, 94, 96, 98, 107, 110, 111], "break": [64, 65, 71, 75, 76, 82, 94, 106], "successfulli": [64, 98, 99], "_dynamo": [64, 95, 96, 100, 101, 103, 112], "explain": [64, 65, 69], "veri": [64, 65, 83, 84, 91, 92, 105, 106, 111], "explan": [64, 65], "graph_break_count": 64, "furthermor": 64, "durat": [64, 82], "latter": [64, 75], "logic": [64, 65, 93], "guard": 64, "compos": [65, 88, 91, 93, 107, 111], "variou": [65, 117], "etc": [65, 80, 82, 94, 117], "environ": [65, 68, 111], "research": 65, "few": [65, 66, 76, 93], "nightli": 65, "lower_exampl": 65, "welcom": [65, 89], "finish": 65, "converison": 65, "pleas": [65, 67, 76, 82, 89, 97, 107, 109, 111, 112], "max_batch_s": [65, 72, 111], "2048": [65, 72], "max_workspace_s": [65, 72], "33554432": [65, 72], "explicit_batch_dimens": [65, 72], "lower_precis": [65, 72], "lowerprecis": [65, 72], "verbose_log": [65, 72], "timing_cache_prefix": [65, 72], "save_timing_cach": [65, 72], "cuda_graph_batch_s": [65, 72], "dynamic_batch": [65, 72], "turn": [65, 72, 104], "trtmodul": [65, 72], "otherwis": [65, 66, 72, 96, 114], "implicit": [65, 70, 72, 82], "config": [65, 66, 72, 111], "updat": [65, 66, 67, 71, 72, 76, 94, 99], "dim": [65, 70, 72, 94, 96, 107, 108, 111, 112], "fx2trt_exampl": 65, "acc_trac": 65, "come": [65, 66, 81, 94, 98, 111], "my_pytorch_model": 65, "build_model": 65, "prepar": [65, 111], "acc_mod": 65, "earli": [65, 99], "deprec": [65, 70], "continu": [65, 82, 114], "backward": [65, 75, 94, 116], "vision": [65, 97, 109, 111], "activ": [65, 75, 77, 82, 89, 91, 93, 110, 114, 117], "except": [65, 71, 76], "permut": [65, 70], "transpos": [65, 70, 112], "ll": [65, 96], "inputtensorspec": [65, 72, 76], "experiment": [65, 76, 77], "dataclass": [65, 100], "re": [65, 76, 82, 96, 98, 104, 114], "manual": [65, 76, 81, 82, 99, 108], "sampl": [65, 71, 82, 90, 91, 98, 99, 100, 101, 102, 103, 104, 105, 106, 111], "rand": [65, 89, 96, 98, 99, 100], "from_tensor": [65, 76], "slightli": [65, 66, 94], "promis": 65, "optimize_target_shap": 65, "input_tensor_spec": 65, "shape_rang": [65, 72], "100": [65, 72, 94, 96, 107, 108], "accordingli": [65, 80, 112, 114], "trtinterpreterresult": [65, 72], "namedtupl": 65, "input_nam": [65, 72], "output_nam": [65, 72], "serialized_cach": [65, 72], "bytearrai": [65, 75, 77], "afford": 65, "temporari": [65, 96], "best": [65, 71, 76, 82, 98, 108, 113], "perforamnc": 65, "examin": 65, "suitabl": [65, 93], "force_fp32_output": 65, "strict_type_constraint": 65, "usual": [65, 66, 80], "unless": 65, "certain": [65, 66, 100, 105, 106, 108, 114], "algorithm_selector": 65, "profiling_verbos": 65, "trt_interpreter_result": 65, "64": [65, 76, 90, 106, 107, 112], "25": [65, 72, 89], "runtimeerror": [65, 107], "xxx": 65, "One": [65, 82, 83, 89, 110, 114], "reload_trt_mod": 65, "reload_model_output": 65, "far": [65, 82], "give": [65, 80, 82], "convtert": 65, "scheme": [65, 71, 76], "action": [65, 82], "tensort": [65, 116], "thing": [65, 66, 82], "compar": [65, 71, 76, 90, 99], "vanilla": 65, "mainli": 65, "builtin": 65, "purpos": [65, 110, 111], "acc_op": 65, "leverag": [65, 91], "power": [65, 82, 89, 108, 110], "goe": [65, 82], "whole": 65, "sigmoid": [65, 70], "tensorrt_convert": 65, "acc_ops_sigmoid": 65, "rest": [65, 82, 83], "input_v": [65, 93], "receiv": 65, "region": 65, "add_activ": 65, "get_output": [65, 94], "wherev": 65, "rememb": [65, 66], "mapper": 65, "todo": [65, 67, 80], "logist": 65, "down": [65, 66, 80, 106], "acc_norm": 65, "foo": [65, 82, 83], "register_acc_op": 65, "register_acc_op_map": 65, "this_arg_is_opt": 65, "op_and_target": 65, "arg_replacement_tupl": 65, "rule": [65, 66, 77], "third": [65, 83], "boolean": [65, 76, 93], "matter": [65, 94], "register_custom_acc_mapper_fn": 65, "design": [65, 74, 93, 98, 105, 108, 110, 117], "redund": 65, "throught": 65, "custom_mapp": 65, "_": [65, 82, 94, 107, 108, 113], "foo_kwarg": 65, "inserting_befor": 65, "foo_nod": 65, "meta": [65, 67, 86, 106, 108], "children": 65, "unit": [65, 76], "test_acc_trac": 65, "acc_op_convert": 65, "essenti": 65, "plugin": [65, 94], "yet": [65, 110], "folder": 65, "center": 66, "pypi": 66, "m": [66, 67, 83, 107], "pip": [66, 67, 97, 109, 111], "upload": [66, 111], "x86": [66, 114], "extra": [66, 75, 89, 94, 98], "url": [66, 80, 111], "download": [66, 67, 86, 91, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 111], "whl": [66, 67], "cu118": 66, "cu124": 66, "tarbal": [66, 89, 91], "easiest": [66, 94], "bazelisk": [66, 67], "bazelbuild": [66, 67], "export": [66, 67, 69, 71, 76, 96, 97, 99, 101, 105, 106, 107, 108, 109, 113, 115, 116], "bazel_vers": 66, "path_to_torchtrt_root": 66, "bazelvers": 66, "mkdir": 66, "cd": [66, 111], "curl": [66, 82], "fssl": 66, "o": [66, 82, 111], "dist": 66, "unzip": 66, "bash": 66, "sh": 66, "cp": [66, 67, 94], "usr": [66, 67], "driver": 66, "branch": [66, 67], "4e5b0f6e860910eb510fa70a76ee3eb9825e7a4d": 66, "l46": 66, "pull": [66, 96, 111], "latest": [66, 67, 80], "l53c1": 66, "fact": 66, "reproduc": 66, "l71": 66, "http_archiv": 66, "build_fil": 66, "archiv": [66, 67], "sha256": 66, "strip_prefix": 66, "OR": 66, "TO": [66, 89], "gnu": 66, "tar": [66, 67, 82, 91], "gz": [66, 82, 83, 91], "ld_library_path": 66, "comment": [66, 82], "uncom": 66, "l114c1": 66, "l124c3": 66, "uv": 66, "astral": 66, "project": [66, 81, 86], "simpler": [66, 91], "wheel": [66, 67], "dep": 66, "lighter": 66, "executor": 66, "avoid": [66, 93, 94, 99, 112], "implic": 66, "python_onli": 66, "legaci": [66, 74], "mainten": 66, "torchdynamo": [66, 112, 116], "technolog": [66, 116], "exclud": [66, 94], "speed": [66, 96, 99], "no_torchscript": 66, "dbg": 66, "pre_cxx11_abi": 66, "complic": 66, "incompat": 66, "popular": [66, 81, 97, 105, 106, 109, 110], "ngc": [66, 67, 111], "tabl": [66, 86], "bdist_wheel": 66, "preinstal": 66, "forum": 66, "correctli": [66, 94], "declar": 66, "intend": [66, 100, 101, 102, 103, 104], "microsoft": 66, "2022": [66, 69], "open": [66, 110, 111], "app": 66, "x64": 66, "prompt": [66, 98, 102, 105, 106], "admin": 66, "privileg": 66, "launcher": 66, "chocolatei": 66, "navig": [66, 80], "ninja": 66, "setuptool": 66, "r": [66, 67, 82, 97, 109], "txt": [66, 67, 97, 109], "distutils_use_sdk": 66, "cuda_win": 66, "libtorch_win": 66, "tensorrt_win": 66, "non": [66, 76, 83, 85, 114], "similarli": [66, 96, 104, 114], "ci_workspac": 66, "win": 66, "tmpl": [66, 67], "torchtrtc": [66, 69, 117], "websit": 66, "finder": 66, "dcmake_module_path": 66, "doesn": [66, 82, 88, 89], "dtorch_dir": 66, "dtensorrt_root": 66, "choic": [66, 74], "b": [66, 70, 76, 83, 108, 111], "dcmake_build_typ": 66, "72048": 66, "jp_workspac": [66, 67], "new_local_repositori": 66, "sudo": [66, 67], "home": 66, "unlik": [66, 92], "libtorch_pre_cxx11_abi": 66, "shift": [66, 70, 82], "jetpack": 66, "jetpack_x": 66, "jetpack_5": 66, "drop": [66, 80, 107], "nvida": 67, "ofjetpack": 67, "With": [67, 80, 82, 89, 91, 96, 111], "incorpor": [67, 83], "cudnn": 67, "9": [67, 86, 89, 94, 111], "dlfw": 67, "09": 67, "jetson": [67, 110], "framework": 67, "instal": [67, 69, 86, 89, 97, 109, 111, 114], "kit": 67, "flash": 67, "board": 67, "apt": 67, "show": [67, 80, 82, 96, 102, 108, 110], "dev": 67, "everth": 67, "nvcc": 67, "cmd": 67, "toolkit": [67, 74], "libcusparselt": 67, "lib64": 67, "wget": [67, 111], "cusparselt": 67, "redist": 67, "libcusparse_lt": 67, "sbsa": 67, "xz": 67, "xf": 67, "v1": [67, 98, 102], "arm64": 67, "mv": 67, "chmod": 67, "pypa": 67, "en": [67, 80], "bootstrap": 67, "jp": 67, "v61": 67, "0a0": 67, "872d972e41": 67, "nv24": 67, "08": 67, "17622132": 67, "cp310": 67, "linux_aarch64": 67, "test_requir": 67, "jetpack6": 67, "lanl": 67, "cuda_vers": 67, "grep": 67, "cut": [67, 82, 99], "sed": [67, 83, 85], "torch_install_path": 67, "dirnam": 67, "__file__": 67, "site_package_path": 67, "cuda_hom": 67, "envsubst": 67, "cxx11": [67, 114], "abi": [67, 114], "anywher": 68, "ahead": [68, 69, 89, 98], "ep": [68, 70, 99, 113, 115], "output_format": [68, 76, 115], "input_tensor": [68, 94, 107, 108], "fill": 68, "aot": [69, 89, 97, 98, 99, 109, 116], "integr": [69, 98, 100], "seamlessli": [69, 76], "ecosystem": [69, 116], "hybrid": [69, 71, 76, 77, 116], "advanc": [69, 78, 83, 91, 97, 109], "bert": [69, 78, 97, 109], "triton": [69, 94], "cudagraph": [69, 97, 109], "overload": [69, 97, 109], "mutabl": [69, 97, 109], "diffus": [69, 78, 97, 109], "gpt2": [69, 97, 109], "llama2": [69, 97, 109], "page": [69, 84, 86, 111], "introductori": 69, "blog": [69, 114], "gtc": 69, "2020": [69, 89], "talk": 69, "fall": [69, 76, 94], "2021": 69, "dai": 69, "confer": 69, "_convolut": [70, 89], "stride": [70, 76, 94, 107], "pad": [70, 76, 94, 107], "dilat": 70, "output_pad": 70, "group": [70, 82, 83], "determinist": 70, "cudnn_en": 70, "allow_tf32": 70, "ab": 70, "aco": 70, "acosh": 70, "adaptive_avg_pool1d": 70, "output_s": 70, "adaptive_avg_pool2d": 70, "adaptive_avg_pool3d": 70, "adaptive_max_pool1d": 70, "adaptive_max_pool2d": 70, "adaptive_max_pool3d": 70, "argmax": [70, 108], "keepdim": 70, "argmin": 70, "asin": 70, "asinh": 70, "atan": 70, "atanh": 70, "avg_pool1d": 70, "kernel_s": [70, 94, 107], "ceil_mod": 70, "count_include_pad": 70, "avg_pool2d": 70, "divisor_overrid": 70, "avg_pool3d": 70, "gamma": 70, "var": 70, "momentum": 70, "bitwise_not": 70, "bmm": 70, "ceil": 70, "clamp": 70, "clamp_max": 70, "clamp_min": 70, "constant_pad_nd": 70, "co": [70, 83, 110], "cosh": 70, "cumsum": 70, "tensor_mod": 70, "rounding_mod": 70, "div_": 70, "elu": 70, "scale": [70, 91, 110], "input_scal": 70, "indic": [70, 80, 82, 93, 99, 101, 112, 113], "padding_idx": 70, "eq": [70, 82], "erf": [70, 93], "exp": 70, "expand_a": 70, "fake_quantize_per_channel_affin": 70, "zero_point": 70, "axi": [70, 76], "quant_min": 70, "quant_max": 70, "fake_quantize_per_tensor_affin": 70, "using_int": [70, 89], "start_dim": [70, 89], "end_dim": [70, 89], "floor": 70, "floor_divid": 70, "ge": 70, "gru_cel": 70, "hx": 70, "w_ih": 70, "w_hh": 70, "b_ih": 70, "b_hh": 70, "gt": 70, "hardtanh_": 70, "instance_norm": 70, "running_mean": 70, "running_var": 70, "use_input_stat": 70, "layer_norm": 70, "normalized_shap": 70, "le": 70, "negative_slop": 70, "01": [70, 83, 89, 107], "leaky_relu_": 70, "lstm_cell": 70, "lt": 70, "masked_fil": 70, "mask": [70, 94], "max_pool1d": 70, "max_pool2d": [70, 88, 89], "max_pool3d": 70, "mul_": [70, 93], "narrow": 70, "neg": [70, 98], "norm": 70, "scalaropt_dim": 70, "pixel_shuffl": 70, "upscale_factor": 70, "pow": 70, "tensor_scalar": 70, "expon": 70, "tensor_tensor": 70, "prelu": 70, "prod": [70, 94], "dim_int": 70, "reciproc": 70, "reflection_pad1d": 70, "reflection_pad2d": 70, "relu_": 70, "repeat_interleav": 70, "self_int": 70, "replication_pad1d": 70, "replication_pad2d": 70, "replication_pad3d": 70, "reshap": [70, 94, 111], "roll": 70, "rsub": 70, "scatter": 70, "sigmoid_": 70, "sin": [70, 82], "sinh": 70, "slice": 70, "split_siz": 70, "split_with_s": 70, "sqrt": 70, "squar": 70, "squeez": [70, 110], "sub_": 70, "dim_intlist": 70, "tan": 70, "tanh": [70, 93], "tanh_": [70, 93], "non_block": [70, 107], "memory_format": [70, 76], "prim_devic": 70, "topk": 70, "k": [70, 91, 107], "largest": 70, "dim0": [70, 96], "dim1": 70, "unbind": 70, "unsqueez": 70, "upsample_bilinear2d": 70, "align_corn": 70, "scales_h": 70, "scales_w": 70, "vec": 70, "scale_factor": 70, "upsample_linear1d": 70, "upsample_nearest1d": 70, "upsample_nearest2d": 70, "upsample_nearest3d": 70, "scales_d": 70, "upsample_trilinear3d": 70, "view": [70, 80], "__and__": 70, "__derive_index": 70, "idx": 70, "__getitem__": 70, "__is__": 70, "t1": 70, "t2": 70, "obj": 70, "__isnot__": 70, "__not__": 70, "__or__": 70, "__range_length": 70, "lo": 70, "hi": [70, 82, 83], "__round_to_zero_floordiv": 70, "__xor__": 70, "append": [70, 93, 95, 96, 107, 108], "el": 70, "arang": [70, 94], "pin_memori": 70, "start_step": 70, "copy_": 70, "float_int": 70, "int_float": 70, "floordiv": 70, "is_floating_point": 70, "numel": 70, "l": [70, 107], "9223372036854775807": 70, "requires_grad": 70, "tupleindex": 70, "tup": 70, "exported_program": [71, 76, 115], "arg_input": [71, 76, 93, 99], "kwarg_input": [71, 76, 99], "engine_cache_dir": [71, 95, 96], "engine_cache_s": [71, 95, 96], "custom_engine_cach": [71, 96], "baseenginecach": [71, 96], "int32": [71, 76, 77, 94, 95, 103, 110], "channel_last": [71, 76, 77, 110], "244": [71, 76, 77], "alia": [71, 76], "better": [71, 76, 88, 110, 116], "understand": [71, 76, 112], "convolut": [71, 76, 77, 91, 94, 117], "_c": [71, 76, 77, 92], "oppos": [71, 76, 77], "lean": [71, 76], "spend": [71, 76], "integ": [71, 76, 85], "faster": [71, 76, 95, 96, 110], "parition": [71, 76], "increas": [71, 76, 96, 108], "amount": [71, 76, 108], "defer": [71, 76, 116], "lead": [71, 76, 82, 108, 114], "oversubscript": [71, 76], "hard": [71, 99], "disk": [71, 76, 96], "space": [71, 82, 83, 91], "byte": [71, 75, 76, 77, 94, 96, 108, 110], "1gb": [71, 95, 96], "exce": 71, "oldest": 71, "gear": [71, 91], "toward": [71, 91], "refit_module_weight": [71, 99], "compiled_modul": [71, 99], "new_weight_modul": [71, 99], "verify_output": [71, 99], "use_weight_map_cach": [71, 99], "in_plac": [71, 99], "compmil": 71, "coverag": [71, 94], "min_acc_module_s": 72, "is_aten": 72, "use_experimental_fx_rt": 72, "correctness_atol": 72, "correctness_rtol": 72, "minim": [72, 91, 94], "submodul": [72, 88, 94], "fx2trt": 72, "cpu": [72, 105, 106, 108], "has_batch_dim": 72, "dtyep": 72, "prop": 72, "min_input_shap": 72, "optimized_input_shap": 72, "max_input_shap": 72, "popul": 72, "225": [72, 111], "explicit_precis": 72, "logger_level": 72, "model_trt": 73, "model_torchtrt": 73, "internal_error": 73, "dataloadercalibr": [74, 91], "preprocess": [74, 91, 111], "algo_typ": [74, 91], "calibrationalgo": [74, 91], "cachecalibr": [74, 91], "qualnam": [74, 76], "entropy_calibr": 74, "entropy_calibration_2": [74, 91], "legacy_calibr": 74, "minmax_calibr": 74, "set_multi_device_safe_mod": [75, 114], "_multidevicesafemodecontextmanag": 75, "impact": 75, "suppress": 75, "unsaf": 75, "trt_compiled_modul": 75, "torchtensorrtmodul": [75, 94], "encompass": [75, 77], "simpili": 75, "de": 75, "initi": [75, 76, 82, 99, 100, 101, 103, 104, 105, 106], "scriptmodul": [75, 76, 77, 89, 90, 115, 116], "overridden": [75, 76], "subclass": 75, "although": [75, 82], "recip": [75, 91], "afterward": 75, "former": 75, "care": 75, "hook": 75, "silent": 75, "get_extra_st": 75, "state_dict": [75, 76, 98], "set_extra_st": 75, "picklabl": 75, "pickl": [75, 94, 96], "load_state_dict": [75, 98, 107], "pythontorchtensorrtmodul": 75, "serialized_engin": [75, 77], "_set": [75, 100], "weight_name_map": 75, "trt_modul": 75, "engine_str": 75, "my_modul": 75, "current_devic": 75, "cudagraphs_validate_shap": 75, "versu": 75, "disable_profil": 75, "enable_profil": 75, "iprofil": 75, "spent": 75, "get_layer_info": 75, "request": [76, 89, 111], "decid": 76, "deseri": [76, 77, 89, 94], "retrac": 76, "strict": [76, 114], "valueerror": 76, "mutabletorchtensorrtmodul": [76, 98], "pytorch_model": 76, "regular": 76, "whenev": 76, "refit_gm": 76, "shape_mod": 76, "_shapemod": 76, "interv": 76, "notat": 76, "bound": 76, "torch_tensor": 76, "tracer": 76, "example_tensor": 76, "optimization_profile_field": 76, "classmethod": 76, "disable_memory_format_check": 76, "core_id": 76, "schedul": [76, 111], "use_default": 76, "try_to": 76, "anoth": [76, 82, 83, 88, 90, 99], "typeerror": 76, "unknown": 76, "succe": 76, "float_dtyp": 76, "failur": 76, "bf16": 76, "try_from": [76, 94], "complex128": 76, "16": [76, 86, 88, 89, 90, 101, 104], "brain": 76, "bfloat16": 76, "f64": 76, "f8": 76, "fp8": 76, "float8": 76, "i32": 76, "sign": [76, 111], "i64": 76, "u8": 76, "unsign": 76, "uint8": 76, "trt_dla": 76, "torchtrt_dla": 76, "_from": 76, "torchtrt_dla_ec": 76, "torchtrt_safety_ec": 76, "saefti": 76, "trt_dla_ec": 76, "standalon": [76, 82], "certifi": 76, "tf": 76, "torchtrt_linear": 76, "cdhw32": 76, "thirti": 76, "row": [76, 83], "spatial": 76, "31": [76, 89], "subscript": [76, 82], "chw16": 76, "sixteen": 76, "15": [76, 82, 86], "chw2": 76, "chw32": 76, "chw4": 76, "four": [76, 82, 83], "dhwc": 76, "equivi": 76, "channels_last_3d": 76, "dhwc8": 76, "eight": 76, "dla_hwc4": 76, "imag": [76, 91, 94, 98, 102, 107, 111], "roundup": 76, "elements": 76, "dla_linear": 76, "planar": 76, "hwc": 76, "channels_last": 76, "hwc16": 76, "hwc8": 76, "least": [76, 82, 83], "ishapelay": 77, "check_method_op_support": 77, "seriali": 77, "put_binding_nam": 77, "tensorrtcompilespec": [77, 92], "scriptclass": 77, "0x7efd148567b0": 77, "_jit_to_tensorrt": 77, "00": 78, "000": [78, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108], "total": [78, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108], "galleri": [78, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "mem": 78, "torch_compile_advanced_usag": [78, 100], "torch_compile_resnet_exampl": [78, 101], "torch_compile_stable_diffus": [78, 102], "torch_compile_transformers_exampl": [78, 103], "v0": [79, 111], "pytorch_sphinx_them": [80, 87], "conf": [80, 87], "html_theme_opt": 80, "canonical_url": 80, "analytics_id": 80, "logo_onli": 80, "display_vers": 80, "prev_next_buttons_loc": 80, "bottom": 80, "style_external_link": 80, "vcs_pageview_mod": 80, "collapse_navig": 80, "sticky_navig": [80, 84], "navigation_depth": 80, "includehidden": 80, "titles_onli": 80, "canon": 80, "rank": 80, "trail": 80, "slash": 80, "googl": 80, "analyt": 80, "isn": [80, 82, 94], "shown": [80, 82, 89, 113], "sidebar": [80, 86], "button": [80, 82], "icon": [80, 82], "extern": [80, 82, 97, 109], "display_github": 80, "display_gitlab": 80, "gitlab": 80, "bitbucket": 80, "bar": [80, 82], "www": [80, 82, 89, 91, 111], "sphinx": [80, 81, 82, 83, 87, 93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "toctre": 80, "lose": 80, "scroll": [80, 84], "unlimit": 80, "header": [80, 82, 83, 89, 111], "render": 80, "github_url": 80, "bitbucket_url": 80, "gitlab_url": 80, "left": [80, 82], "upon": [80, 100, 103], "rst": [80, 82], "visitor": 80, "revert": 80, "misbuild": 80, "properti": [80, 94], "stick": 80, "screen": 80, "vertic": [80, 82], "too": [80, 82, 83], "sticki": [80, 86], "nav": [80, 86], "At": [81, 93, 99], "django": 81, "payment": 81, "dotpai": 81, "dotpayprovid": 81, "seller_id": 81, "pin": 81, "lock": 81, "lang": 81, "pl": 81, "polish": 81, "gatewai": 81, "transfer": 81, "purchas": 81, "item": [81, 83, 107], "param": 81, "seller": 81, "consult": 81, "ui": 81, "languag": [81, 82, 83, 88, 94, 97, 105, 109, 111], "data_item_1": 81, "emphasi": 82, "hyperlink": 82, "cross": 82, "uri": 82, "web": 82, "anonym": 82, "label": [82, 91, 107, 110, 111], "substitut": 82, "charact": 82, "exceedingli": 82, "ugli": 82, "problem": [82, 106], "problemat": 82, "ext": [82, 83], "autodoc": [82, 83], "demo": [82, 91], "test_py_modul": [82, 86], "my": [82, 105], "role": 82, "pep": 82, "287": 82, "rfc": 82, "2822": 82, "superscript": 82, "gui": 82, "taken": 82, "height": 82, "interfer": 82, "press": 82, "keyboard": 82, "mous": 82, "mmb": 82, "menuselect": 82, "seen": [82, 83], "whitespac": 82, "signific": [82, 94], "strang": 82, "hyphen": 82, "word": [82, 110], "adjust": 82, "width": [82, 110], "browser": 82, "window": 82, "sentenc": [82, 108, 110], "suppli": [82, 99], "258": 82, "equat": 82, "x_": 82, "x_0": 82, "x_1": 82, "x_2": 82, "x_3": 82, "x_4": 82, "nabla": 82, "frac": 82, "theta": 82, "phi": 82, "restructuredtext": [82, 83], "parser": [82, 107], "colon": 82, "indent": 82, "literal_block": 82, "spaces_and_linebreak": 82, "preserv": [82, 88, 91], "markup_process": 82, "Or": 82, "great": [82, 89, 94, 96, 116], "why": [82, 114], "didn": 82, "blank": 82, "align": 82, "permit": 82, "awai": 82, "eric": 82, "orchestra": 82, "leader": 82, "bee": 82, "philosoph": 82, "ipso": 82, "facto": 82, "But": [82, 89, 99, 108], "got": [82, 89], "vi": 82, "entiti": 82, "said": 82, "entir": [82, 116], "ancient": 82, "injuri": 82, "sing": 82, "elk": 82, "bracket": 82, "miss": [82, 89], "brontosaurus": 82, "thin": 82, "thicker": 82, "middl": 82, "That": [82, 89], "mine": 82, "belong": 82, "me": [82, 83], "ann": 82, "begun": 82, "past": 82, "pars": [82, 89], "someurl": 82, "dev0": 82, "e43833d": 82, "caption": [82, 85], "pane": 82, "shell_command": 82, "echo": 82, "did": 82, "window_nam": 82, "session_nam": 82, "shorthand": 82, "some_funct": 82, "highlight": 82, "THE": 82, "heaven": 82, "hexagram": 82, "six": 82, "unbroken": 82, "primal": 82, "light": [82, 115], "spirit": 82, "weak": 82, "essenc": 82, "energi": 82, "unrestrict": 82, "conceiv": 82, "motion": 82, "regard": [82, 116], "basi": 82, "thu": 82, "persist": 82, "dual": 82, "sens": [82, 89], "univers": 82, "world": 82, "men": 82, "express": 82, "deiti": 82, "human": 82, "denot": [82, 94], "holi": 82, "man": [82, 83], "sage": 82, "ruler": 82, "who": 82, "awaken": 82, "utf": [82, 83], "sphinx_rtd_them": [82, 83], "docstr": [82, 83, 90], "dl": 82, "dt": 82, "tag": [82, 111], "tt": 82, "descnam": 82, "descclassnam": 82, "wrote": 82, "anyth": [82, 83, 114], "programm": 82, "myclass": 82, "dothismethod": 82, "flush": 82, "meth": 82, "capit": 82, "flox": 82, "unreferenc": 82, "nonexist": 82, "extrem": 82, "stuff": 82, "mayb": 82, "bold": 82, "ital": 82, "heck": 82, "put": [82, 110], "13": [82, 86], "backlink": 82, "knowledg": 82, "mind": 82, "ey": 82, "thought": 82, "medium": 82, "peopl": 82, "subsect": 82, "interpol": 82, "indirect": 82, "phrase": 82, "docutil": [82, 83], "sourceforg": [82, 83], "ref": 82, "clickabl": 82, "legend": 82, "revis": [82, 83, 98, 102], "revisit": 82, "enhanc": 82, "structuredtext": 82, "wooden": 82, "nickel": 82, "mad": 82, "scientist": 82, "bigger": 82, "bread": 82, "box": [82, 112, 116], "wash": 82, "behind": 82, "ear": 82, "room": 82, "closet": 82, "bathroom": 82, "trash": 82, "sink": 82, "mother": 82, "g_": 82, "mu": 82, "nu": 82, "pi": 82, "t_": 82, "rho_": 82, "servic": 82, "thing1": 82, "thing2": 82, "thing3": 82, "prose": 82, "provok": 82, "mental": 82, "exert": 82, "reader": 82, "discret": 82, "strongli": [82, 108], "advis": 82, "subtitl": 82, "outsid": 82, "often": 82, "besid": 82, "border": 82, "background": [82, 88], "ok": [82, 89], "transmit": 82, "disconnect": 82, "nonetheless": 82, "semant": 82, "blue": [82, 94], "white": 82, "arab": 83, "roman": 83, "upper": 83, "iii": 83, "iv": 83, "classifi": [83, 88, 89, 107, 110], "paragraph": [83, 86], "z": 83, "commonli": 83, "vm": 83, "david": 83, "goodger": 83, "address": [83, 94, 98], "123": 83, "street": 83, "canada": 83, "a1b": 83, "2c3": 83, "contact": 83, "myself": 83, "organ": 83, "humankind": 83, "2012": 83, "03": 83, "19": [83, 86], "53": 83, "0000": 83, "tue": 83, "jan": 83, "progress": 83, "7302": 83, "wish": 83, "redistribut": 83, "reattribut": 83, "sell": 83, "bui": 83, "rent": 83, "leas": 83, "improv": [83, 114], "quot": 83, "excerpt": 83, "collat": 83, "fold": 83, "stapl": 83, "mutil": 83, "anyon": 83, "heart": 83, "bibliograph": 83, "markup": [83, 86], "literal": 83, "yahoo": 83, "oh": 83, "liter": 83, "heh": 83, "child": 83, "beat": 83, "text": [83, 85, 105, 106, 110], "hehe": 83, "kept": 83, "sai": [83, 110], "cackl": 83, "night": 83, "lone": 83, "guangzhou": 83, "destini": 83, "hope": 83, "dream": 83, "forth": 83, "fifth": 83, "sixth": 83, "lorem": [83, 85], "ipsum": [83, 85], "dolor": [83, 85], "sit": [83, 85], "amet": [83, 85], "consectetur": [83, 85], "adipisc": [83, 85], "elit": [83, 85], "donec": [83, 85], "porttitor": [83, 85], "odio": [83, 85], "posuer": [83, 85], "vita": [83, 85], "ornar": [83, 85], "libero": [83, 85], "matti": 83, "loborti": [83, 85], "justo": [83, 85], "vestibulum": [83, 85], "nibh": [83, 85], "aliquet": [83, 85], "feugiat": [83, 85], "sagitti": [83, 85], "nequ": [83, 85], "qui": [83, 85], "eleifend": 83, "dui": [83, 85], "rutrum": [83, 85], "lectu": [83, 85], "suscipit": [83, 85], "letter": [83, 110], "column": 83, "cell": 83, "span": 83, "nam": [83, 85], "mauri": [83, 85], "arcu": [83, 85], "stub": 83, "behav": 84, "area": 84, "interdum": 85, "nec": 85, "finibu": 85, "dictum": 85, "velit": 85, "ut": 85, "eu": 85, "efficitur": 85, "aliquam": 85, "erat": 85, "diam": 85, "gravida": 85, "imperdiet": 85, "tellu": 85, "nisl": 85, "praesent": 85, "eget": 85, "elementum": 85, "rhoncu": 85, "tincidunt": 85, "suspendiss": 85, "volutpat": 85, "scelerisqu": 85, "tristiqu": 85, "aenean": 85, "condimentum": 85, "risu": 85, "accumsan": 85, "laoreet": 85, "maximu": 85, "sapien": 85, "ligula": 85, "fringilla": 85, "commodo": 85, "proin": 85, "et": 85, "pharetra": 85, "etiam": 85, "turpi": 85, "ant": 85, "luctu": 85, "vel": 85, "malesuada": 85, "dignissim": 85, "mi": 85, "nunc": 85, "augu": 85, "sem": 85, "cursu": 85, "nulla": 85, "pellentesqu": 85, "habit": 85, "morbi": 85, "senectu": 85, "netu": 85, "fame": 85, "ac": 85, "egesta": 85, "placerat": 85, "tortor": 85, "iaculi": 85, "venenati": 85, "cra": 85, "puru": 85, "ero": 85, "vehicula": 85, "fusc": 85, "auctor": 85, "phasellu": 85, "est": 85, "viverra": 85, "conval": 85, "faucibu": 85, "vulput": 85, "feli": 85, "sodal": 85, "maecena": 85, "congu": 85, "semper": 85, "enim": 85, "blandit": 85, "sollicitudin": 85, "urna": 85, "orci": 85, "lacu": 85, "quisqu": 85, "facilisi": 85, "hendrerit": 85, "curabitur": 85, "variu": 85, "bibendum": 85, "massa": 85, "magna": 85, "tempu": 85, "metu": 85, "nisi": 85, "pretium": 85, "leo": 85, "euismod": 85, "ultric": 85, "dapibu": 85, "lacinia": 85, "vivamu": 85, "molesti": 85, "hac": 85, "habitass": 85, "platea": 85, "dictumst": 85, "git": 86, "content": [86, 91, 111], "changelog": 86, "math": 86, "14": [86, 95, 103, 111], "17": 86, "18": [86, 89, 98], "submenu": 86, "symlink": 87, "subtre": 87, "_theme": 87, "html_theme": 87, "html_theme_path": 87, "optimiz": 88, "tutori": [88, 91, 93, 94, 96, 98, 99], "beginn": 88, "intro_to_torchscript_tutori": 88, "briefli": 88, "lenet": [88, 89], "lenetfeatextractor": 88, "conv1": [88, 89], "conv2d": [88, 94, 107], "conv2": [88, 89], "lenetclassifi": 88, "fc1": [88, 89], "120": [88, 89], "fc2": [88, 89], "84": [88, 89], "fc3": [88, 89], "feat": [88, 89], "obvious": 88, "pathwai": 88, "input_data": [88, 90], "traced_model": 88, "pick": [88, 113], "script_model": [88, 92], "perspect": 88, "___torch_mangle_10": 88, "129": 88, "___torch_mangle_9": 88, "119": 88, "___torch_mangle_5": 88, "137": 88, "callmethod": 88, "138": 88, "38": 88, "39": 88, "torch_script_modul": [88, 89], "in_tensor": 88, "fly": 88, "lenet_script": [88, 89], "haven": 89, "acquir": 89, "dyanmo": 89, "almost": [89, 116], "trt_lenet_script": 89, "apr": 89, "56": 89, "04": [89, 111], "credit": 89, "stop": 89, "argc": 89, "argv": 89, "cerr": 89, "cout": 89, "even": [89, 98], "cppdoc": 89, "pretti": 89, "fashion": [89, 110], "enable_precis": 89, "And": 89, "convertgraphtotrtengin": 89, "engine_converted_from_jit": 89, "close": [89, 93], "saw": 89, "576": 89, "346": 89, "539": 89, "0464": 89, "0383": 89, "0678": 89, "0932": 89, "1045": 89, "0805": 89, "0435": 89, "0818": 89, "0208": 89, "0358": 89, "cudafloattyp": 89, "0530": 89, "1691": 89, "2802": 89, "1502": 89, "1056": 89, "1549": 89, "input0": [89, 90], "1063": 89, "input1": [89, 90], "input2": 89, "28": 89, "29": 89, "33": 89, "35": 89, "36": 89, "37": 89, "compilegraph": [89, 91], "transform": [89, 91, 95, 97, 99, 103, 105, 106, 107, 108, 109, 111, 115], "laid": 89, "translat": [89, 99], "aren": 89, "techniqu": [89, 91, 106, 114], "checkmethodoperatorsupport": 89, "modular": 89, "ship": [89, 114], "exhaust": 89, "109": 89, "addlay": 89, "yourself": 89, "question": [89, 93], "outself": 89, "flatten_convert": 89, "unwraptoint": 89, "in_shap": 89, "tovec": 89, "out_shap": 89, "shuffl": [89, 91, 107], "addshuffl": 89, "setreshapedimens": 89, "todim": 89, "extens": [89, 116], "ctype": 89, "cdll": 89, "contributor": 89, "upstream": 89, "pr": 89, "usecas": 90, "sole": [90, 91, 116], "individu": 90, "accuraci": [91, 110], "loss": [91, 110], "infrastructur": [91, 111], "streamlin": 91, "expos": [91, 94], "cpp_frontend": 91, "loading_data_recip": 91, "cifar10": [91, 107], "cstddef": 91, "ktrain": 91, "ktest": 91, "un": 91, "cs": 91, "toronto": 91, "edu": 91, "kriz": 91, "cifar": 91, "is_train": 91, "trim": 91, "use_subset": 91, "new_siz": 91, "mode_": 91, "images_": 91, "targets_": 91, "calibration_dataset": 91, "data_dir": 91, "320": 91, "4914": [91, 107], "4822": [91, 107], "4465": [91, 107], "2023": [91, 107], "1994": [91, 107], "2010": [91, 107], "dataloaderopt": 91, "worker": 91, "virtual": 91, "input_shap": [91, 117], "compile_spec": [91, 101, 117], "kf16": [91, 117], "ki8": 91, "vgg16": [91, 107], "testing_dataset": [91, 107], "totensor": [91, 107, 111], "testing_dataload": [91, 107], "num_work": [91, 107], "vgg": [91, 107], "test_ptq_dataloader_calibr": 91, "test_ptq_trt_calibr": 91, "krizhevski": 91, "hinton": 91, "2009": 91, "tini": 91, "simonyan": 91, "zisserman": 91, "2014": 91, "recognit": [91, 110], "arxiv": 91, "preprint": 91, "1409": 91, "1556": 91, "_jit_to_backend": 92, "mobilenet_v2": 92, "pretrain": [92, 96, 98, 101, 104, 110, 111], "gelu": 93, "sy": 93, "approxim": 93, "suppos": 93, "my_mod": 93, "ex_input": [93, 94], "baselin": 93, "my_standard_gelu": 93, "supports_dynamic_shap": 93, "supersed": 93, "converterprior": 93, "vers": 93, "prior": [93, 96, 112, 114], "distinct": 93, "prepend": 93, "candid": 93, "primit": 93, "compiler_ir": 93, "boilerpl": 93, "focu": [93, 98], "interoper": 93, "aten_ops_gelu": 93, "sourceir": 93, "cheap": 93, "unqiu": 93, "op_count": 93, "get_op_count": 93, "nonloc": 93, "elementwis": 93, "source_ir": 93, "lhs_val": 93, "rhs_val": 93, "x_7": 93, "x_8": 93, "79788456080000003": 93, "x_9": 93, "044714999999999998": 93, "x_10": 93, "x_11": 93, "x_12": 93, "x_13": 93, "x_14": 93, "x_15": 93, "my_custom_gelu": 93, "allclos": [93, 98, 99], "my_mod_erf": 93, "my_gelu_erf": 93, "notic": 93, "minut": [93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108], "converter_overload": 93, "jupyt": [93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "ipynb": [93, 94, 95, 96, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108], "cost": [94, 96, 99, 114], "perhap": 94, "overhead": [94, 108, 114], "sake": 94, "circular": 94, "red": 94, "green": 94, "twice": 94, "written": 94, "openai": 94, "formal": 94, "tl": 94, "custom_op": 94, "circ_pad_kernel": 94, "all_pads_0": 94, "all_pads_2": 94, "all_pads_4": 94, "all_pads_6": 94, "orig_dims_0": 94, "orig_dims_1": 94, "orig_dims_2": 94, "orig_dims_3": 94, "y_shape_1": 94, "y_shape_2": 94, "y_shape_3": 94, "x_len": 94, "y_len": 94, "block_siz": 94, "pid": 94, "program_id": 94, "mask_i": 94, "i3": 94, "i2": 94, "i1": 94, "i0": 94, "j0": 94, "j1": 94, "j2": 94, "j3": 94, "load_idx": 94, "mask_x": 94, "launch": [94, 111], "torchtrt_ex": 94, "triton_circular_pad": 94, "mutates_arg": 94, "out_dim": 94, "tolist": 94, "all_pad": 94, "zero": 94, "orig_dim": 94, "blocksiz": 94, "256": [94, 107, 108, 111], "numblock": 94, "tracabl": 94, "prerequisit": 94, "fake": 94, "real": 94, "faketensor": 94, "register_fak": 94, "autograd": 94, "beyond": 94, "register_autograd": 94, "padded_x": 94, "my_model": 94, "2604": 94, "4232": 94, "3041": 94, "0833": 94, "2461": 94, "1270": 94, "2450": 94, "4079": 94, "2887": 94, "2828": 94, "0373": 94, "0332": 94, "3143": 94, "6344": 94, "5638": 94, "1867": 94, "5068": 94, "4363": 94, "7937": 94, "3488": 94, "1350": 94, "7966": 94, "3517": 94, "1379": 94, "5537": 94, "1088": 94, "8950": 94, "0550": 94, "6163": 94, "0109": 94, "5245": 94, "9632": 94, "5686": 94, "3775": 94, "8162": 94, "4216": 94, "4311": 94, "1649": 94, "2091": 94, "3668": 94, "1006": 94, "1447": 94, "0352": 94, "7689": 94, "8131": 94, "_run_on_gpu_0": 94, "_run_on_acc_1": 94, "dry": 94, "50": [94, 110], "count": 94, "__": 94, "were": [94, 99, 114], "aggreg": 94, "stat": 94, "latenc": [94, 108, 114], "abstractli": 94, "pkl": [94, 98], "cupi": 94, "gap": 94, "prealloc": 94, "circularpaddingplugin": 94, "ipluginv2dynamicext": 94, "field_collect": 94, "pluginfieldcollect": 94, "x_shape": 94, "num_output": 94, "plugin_namespac": 94, "plugin_typ": 94, "plugin_vers": 94, "assert": [94, 98, 99], "get_output_datatyp": 94, "input_typ": 94, "get_output_dimens": 94, "output_index": 94, "dimsexpr": 94, "exprbuild": 94, "iexprbuild": 94, "output_dim": 94, "dimensionoper": 94, "configure_plugin": 94, "inp": 94, "dynamicplugintensordesc": 94, "x_dim": 94, "desc": 94, "supports_format_combin": 94, "po": 94, "in_out": 94, "plugintensordesc": 94, "num_input": 94, "enqueu": 94, "input_desc": 94, "output_desc": 94, "in_dtyp": 94, "a_mem": 94, "unownedmemori": 94, "items": 94, "c_mem": 94, "a_ptr": 94, "memorypoint": 94, "c_ptr": 94, "a_d": 94, "memptr": 94, "c_d": 94, "a_t": 94, "as_tensor": 94, "c_t": 94, "cloned_plugin": 94, "__dict__": 94, "circularpaddingplugincr": 94, "iplugincr": 94, "field_nam": 94, "pluginfield": 94, "pluginfieldtyp": 94, "create_plugin": 94, "pluginfieldcollection_": 94, "deserialize_plugin": 94, "pads_dict": 94, "creator": 94, "trt_plugin_registri": 94, "get_plugin_registri": 94, "register_cr": 94, "untyp": 94, "get_trt_tensor": 94, "set_layer_nam": 94, "recal": 94, "intlist": 94, "circular_padding_convert": 94, "retriev": 94, "elsewher": 94, "plugin_registri": 94, "plugin_cr": 94, "get_plugin_cr": 94, "field_config": 94, "eventu": 94, "freez": 94, "_input": 94, "add_plugin_v2": 94, "circular_padding_plugin": 94, "_run_on_acc_0": 94, "grad_fn": 94, "subbackward0": 94, "custom_kernel_plugin": 94, "engine_caching_exampl": [95, 96], "remove_timing_cach": [95, 96], "bertmodel": [95, 103], "random": [95, 96, 98, 99, 108], "seed": [95, 96, 98, 99], "manual_se": [95, 96, 98, 99], "from_pretrain": [95, 98, 102, 103, 105, 106, 108], "uncas": [95, 103, 110], "return_dict": 95, "randint": [95, 103, 108], "compile_bert": 95, "enable_tim": [95, 96], "1st": [95, 96], "measur": [95, 96, 108], "2nd": [95, 96], "3rd": [95, 96], "slower": [95, 96], "messur": [95, 96], "compilation_kwarg": [95, 103], "torch_trt_bert_engine_cach": 95, "30": [95, 96, 98, 99, 101, 103, 113], "synchron": [95, 96, 108], "elapsed_tim": [95, 96], "millisecond": 95, "__name__": [95, 100, 103], "__main__": [95, 100, 103], "engine_caching_bert_exampl": 95, "paid": 96, "upfront": 96, "invalid": 96, "repeatedli": 96, "mitig": 96, "explor": 96, "resnet18": [96, 98, 99, 101, 104], "torch_trt": [96, 98, 99], "_default": 96, "_engine_cach": 96, "flexibl": [96, 116], "histor": 96, "barrier": 96, "reconstruct": 96, "ti": 96, "hash": 96, "magnitud": 96, "torch_compil": [96, 100, 101, 103, 104, 112, 116], "compiled_model": 96, "ms": [96, 108], "dynamo_compil": 96, "example_input": 96, "200": 96, "dynamic_shap": [96, 112], "remot": 96, "systen": 96, "agnost": 96, "implent": 96, "ramenginecach": 96, "held": 96, "engine_cach": 96, "torch_compile_my_cach": 96, "_torch_export_gpt2": [97, 109], "_torch_export_llama2": [97, 109], "straightforward": 98, "especi": 98, "hug": [98, 105, 106], "face": [98, 105, 106], "difficult": 98, "ever": 98, "walk": [98, 99, 105], "lora": [98, 99], "use_python": 98, "mutable_modul": 98, "model2": [98, 99], "expected_output": [98, 99], "refitted_output": [98, 99], "reload": [98, 116], "checkpoint": [98, 107], "civitai": 98, "12597": 98, "moxin": 98, "diffusionpipelin": [98, 102], "no_grad": [98, 105, 106, 107, 108], "model_id": [98, 102], "runwayml": 98, "hous": 98, "forest": 98, "shuimobysim": 98, "wuchangshuo": 98, "qualiti": 98, "worst": 98, "lowr": 98, "cloudi": 98, "watermark": 98, "pipe": [98, 102], "torch_dtyp": [98, 102], "unet": [98, 102], "negative_prompt": 98, "num_inference_step": 98, "without_lora_mut": 98, "jpg": [98, 111], "procedur": 98, "load_lora_weight": 98, "stablediffusionapi": 98, "load_lora_embed": 98, "weight_nam": 98, "safetensor": 98, "adapter_nam": 98, "lora1": 98, "set_adapt": 98, "adapter_weight": 98, "fuse_lora": 98, "unload_lora_weight": 98, "with_lora_mut": 98, "mutable_torchtrt_module_exampl": 98, "expens": 99, "involv": 99, "occasion": [99, 100, 103], "adapt": 99, "infeas": 99, "focus": 99, "mostli": 99, "recogn": 99, "behalf": 99, "init": [99, 107], "sett": 99, "randomli": 99, "exp_program2": 99, "compiled_trt_ep": 99, "new_trt_gm": 99, "accomplish": 99, "gaurente": 99, "attempt": [99, 107, 112], "rebuild": 99, "heurist": 99, "refit_engine_exampl": 99, "x_out": 100, "y_out": 100, "x_y_out": 100, "invoc": 100, "sample_inputs_half": 100, "model_half": 100, "backend_kwarg": 100, "optimized_model_custom": 100, "exit": [100, 103, 111], "2052": [100, 103], "compile_engine_and_inf": [100, 103], "new_input": [101, 103], "new_output": [101, 103], "new_batch_size_input": 101, "new_batch_size_output": 101, "inputs_bs8": 101, "mark_dynam": [101, 112], "outputs_bs8": 101, "No": [101, 112], "inputs_bs12": 101, "outputs_bs12": 101, "compvi": 102, "majest": 102, "castl": 102, "cloud": 102, "majestic_castl": 102, "png": 102, "enable_cudagraph": [104, 114], "out_trt": 104, "set_cudagraphs_mod": [104, 114], "inputs_2": 104, "inputs_3": 104, "out_trt_2": 104, "out_trt_3": 104, "torch_export_cudagraph": 104, "automodelforcausallm": [105, 106, 108], "autotoken": [105, 106], "export_llm": [105, 106, 108], "max_token": [105, 106, 108], "kv_cach": [105, 106], "token": [105, 106, 110], "pad_token_id": 105, "eos_token_id": [105, 106], "attn_implement": [105, 106, 108], "eager": [105, 106, 108], "enjoi": 105, "cute": 105, "dog": 105, "model_input": [105, 106], "return_tensor": [105, 106], "input_id": [105, 106], "regress": [105, 106], "huggingfac": [105, 106, 110], "pyt_gen_token": [105, 106], "gpt2_ep": 105, "max_seq_len": [105, 106, 108], "trt_gen_token": [105, 106], "skip_special_token": [105, 106], "parallel": 105, "paradigm": 105, "torch_export_gpt2": 105, "llama_path": [106, 108], "llama": [106, 108], "7b": [106, 108], "chat": [106, 108], "hf": [106, 108], "llama2_ep": [106, 108], "batch_decod": 106, "clean_up_tokenization_spac": 106, "solv": [106, 111], "smaller": [106, 110], "subproblem": 106, "torch_export_llama2": 106, "argpars": 107, "modelopt": 107, "mtq": 107, "export_torch_mod": 107, "layer_spec": 107, "num_class": 107, "1000": [107, 108, 111], "init_weight": 107, "in_channel": 107, "pool": [107, 117], "maxpool2d": 107, "batchnorm2d": 107, "sequenti": 107, "avgpool": 107, "adaptiveavgpool2d": 107, "4096": 107, "dropout": 107, "_initialize_weight": 107, "kaiming_normal_": 107, "fan_out": 107, "nonlinear": 107, "constant_": 107, "elif": 107, "normal_": 107, "vgg16_cfg": 107, "128": [107, 108], "argumentpars": 107, "add_argu": 107, "ckpt": 107, "parse_arg": 107, "model_state_dict": 107, "device_count": 107, "ordereddict": 107, "new_state_dict": 107, "forget": 107, "training_dataset": 107, "randomcrop": 107, "randomhorizontalflip": 107, "training_dataload": 107, "drop_last": 107, "crit": 107, "crossentropyloss": 107, "calibrate_loop": 107, "pred": 107, "5f": 107, "acc": 107, "2f": 107, "quantize_typ": 107, "quant_cfg": 107, "int8_default_cfg": 107, "fp8_default_cfg": 107, "forward_loop": 107, "qdq": 107, "incomplet": 107, "functionaltensor": 107, "functionaltensormod": 107, "_trace": 107, "_export": 107, "float8_e4m3fn": 107, "class_prob": 107, "class_pr": 107, "test_prob": 107, "test_pr": 107, "test_loss": 107, "test_acc": 107, "vgg16_ptq": 107, "overcom": 108, "throughput": 108, "sometim": [108, 112], "outweigh": 108, "slowdown": 108, "hardwar": [108, 117], "experi": 108, "balanc": 108, "timeit": 108, "time_gener": 108, "output_seq_length": 108, "seq_len": [108, 112], "llm": 108, "input_seq": 108, "start_tim": 108, "default_tim": 108, "inputs_copi": 108, "decod": 108, "logit": 108, "next_token_logit": 108, "next_token": 108, "end_tim": 108, "time_mean_m": 108, "isl": 108, "osl": 108, "warm": 108, "solut": 108, "insight": 108, "weight_streaming_ctx": 108, "weight_stream": 108, "mean_lat": 108, "percentag": 108, "weight_budget_pct": 108, "device_budget": 108, "total_device_budget": 108, "permiss": 108, "equal": 108, "proportion": 108, "streamabl": 108, "streamable_budget": 108, "requested_budget": 108, "get_automatic_weight_streaming_budget": 108, "weight_streaming_exampl": 108, "_rendered_examples_python": 109, "_rendered_examples_jupyt": 109, "acoust": 110, "speech": 110, "quartznet": 110, "contextnet": 110, "subword": 110, "piec": 110, "excit": 110, "se": 110, "audio": 110, "transcrib": 110, "speedup": 110, "obtain": [110, 115], "feedforward": 110, "cnn": 110, "uniformli": 110, "resolut": 110, "highli": [110, 111], "compound": 110, "coeffici": 110, "b0": 110, "corpu": 110, "english": 110, "supervis": 110, "walkthrough": 110, "overal": 110, "adopt": 110, "mobilenetv2": 110, "classif": 110, "imagenet": 110, "imagenett": 110, "qat": 110, "simul": 110, "hand": 111, "consider": 111, "concurr": 111, "grpc": 111, "aforement": 111, "familiar": 111, "resnet50": 111, "torchhub": 111, "docker": 111, "login": 111, "xx": 111, "yy": 111, "mm": 111, "publish": 111, "22": 111, "pwd": 111, "scratch_spac": 111, "nvcr": 111, "py3": 111, "proce": 111, "hub": 111, "_validate_not_a_forked_repo": 111, "suggest": 111, "simplest": 111, "model_repositori": 111, "pbtxt": 111, "pytorch_libtorch": 111, "input__0": 111, "data_typ": 111, "type_fp32": 111, "output__0": 111, "exact": 111, "encourag": 111, "rm": 111, "8000": 111, "8001": 111, "8002": 111, "the_model_repositori": 111, "tritonserv": 111, "spin": 111, "proceed": 111, "flesh": 111, "img1": 111, "hakaimagazin": 111, "wp": 111, "gulf": 111, "bird": 111, "attrdict": 111, "pyindex": 111, "tritoncli": 111, "jump": 111, "firstli": 111, "resiz": 111, "pil": 111, "httpclient": 111, "triton_to_np_dtyp": 111, "rn50_preprocess": 111, "img_path": 111, "img": 111, "centercrop": 111, "485": 111, "456": 111, "406": 111, "229": 111, "transformed_img": 111, "inferenceservercli": 111, "localhost": 111, "secondli": 111, "inferinput": 111, "set_data_from_numpi": 111, "binary_data": 111, "inferrequestedoutput": 111, "class_count": 111, "lastli": 111, "send": 111, "model_nam": 111, "inference_output": 111, "as_numpi": 111, "468750": 111, "90": 111, "523438": 111, "92": 111, "664062": 111, "429688": 111, "136": 111, "234375": 111, "confidence_scor": 111, "classification_index": 111, "eagerli": 112, "swap": 112, "exactli": 112, "_tracer": 112, "queri": 112, "attn_weight": 112, "compiler_dynamic_shap": 112, "inputs_bs2": 112, "mymodul": 113, "linear1": 113, "linear2": 113, "linear3": 113, "40": 113, "__myl_mulsum_myl0_0": 113, "layertyp": 113, "kgen": 113, "__mye116_dconst": 113, "__myln_k_arg__bb1_2": 113, "tacticnam": 113, "__myl_mulsum_0xfa6c1858aea1b13b03f90165d7149ec6": 113, "streamid": 113, "__myl_addresmulsum_myl0_1": 113, "__mye131_dconst": 113, "addmm_constant_0": 113, "addmm_add_broadcast_to_same_shape_lhs_broadcast_constantfloat": 113, "__myln_k_arg__bb1_3": 113, "__myl_addresmulsum_0xb3915d7ebfe48be45b6d49083479e12f": 113, "__myl_addresmulsumadd_myl0_2": 113, "__mye146_dconst": 113, "addmm_2_constant_0": 113, "addmm_2_add_broadcast_to_same_shape_lhs_broadcast_constantfloat": 113, "addmm_1_constant_0": 113, "addmm_1_add_broadcast_to_same_shape_lhs_broadcast_constantfloat": 113, "__myl_addresmulsumadd_0xcdd0085ad25f5f45ac5fafb72acbffd6": 113, "__myl_mulsumaddcas_myl0_0": 113, "__mye112_dconst": 113, "__myl_mulsumaddcas_0xacf8f5dd9be2f3e7bb09cdddeac6c936": 113, "__myl_resmulsumaddcas_myl0_1": 113, "__mye127_dconst": 113, "addmm_1_add_broadcast_to_same_shape_lhs_broadcast_constanthalf": 113, "__myl_resmulsumaddcas_0x5a3b318b5a1c97b7d5110c0291481337": 113, "__myl_resmulsumadd_myl0_2": 113, "__mye142_dconst": 113, "__myl_resmulsumadd_0x3fad91127c640fd6db771aa9cde67db0": 113, "libtorchtrt_runtim": 114, "dl_open": 114, "ld_preload": 114, "load_librari": 114, "wl": 114, "ltorchtrt": 114, "torchtrt_runtime_exampl": 114, "libtorchtrt_plugin": 114, "neglig": 114, "thread": 114, "alert": 114, "switch": 114, "mismatch": 114, "crash": 114, "sacrif": 114, "incur": 114, "intens": 114, "trt_ep": 115, "stai": 115, "trt_t": 115, "ergonom": 116, "deleg": 116, "believ": 116, "amen": 116, "artifact": 116, "pack": 116, "year": 116, "superset": 116, "codebas": 116, "immedi": 116, "traceabl": 116, "scriptabl": 116, "neural": 117, "deconvolut": 117, "scripted_model": 117}, "objects": {"": [[5, 0, 1, "c.STR", "STR"], [9, 0, 1, "c.TORCHTRT_API", "TORCHTRT_API"], [11, 0, 1, "c.TORCHTRT_HIDDEN", "TORCHTRT_HIDDEN"], [7, 0, 1, "c.TORCH_TENSORRT_MAJOR_VERSION", "TORCH_TENSORRT_MAJOR_VERSION"], [8, 0, 1, "c.TORCH_TENSORRT_MINOR_VERSION", "TORCH_TENSORRT_MINOR_VERSION"], [6, 0, 1, "c.TORCH_TENSORRT_PATCH_VERSION", "TORCH_TENSORRT_PATCH_VERSION"], [12, 0, 1, "c.TORCH_TENSORRT_VERSION", "TORCH_TENSORRT_VERSION"], [10, 0, 1, "c.XSTR", "XSTR"], [0, 1, 1, "_CPPv4N14torch_tensorrt8DataTypeE", "torch_tensorrt::DataType"], [0, 2, 1, "_CPPv4N14torch_tensorrt8DataType8DataTypeE5Value", "torch_tensorrt::DataType::DataType"], [0, 2, 1, "_CPPv4N14torch_tensorrt8DataType8DataTypeEN3c1010ScalarTypeE", "torch_tensorrt::DataType::DataType"], [0, 2, 1, "_CPPv4N14torch_tensorrt8DataType8DataTypeEv", "torch_tensorrt::DataType::DataType"], [0, 3, 1, "_CPPv4N14torch_tensorrt8DataType8DataTypeE5Value", "torch_tensorrt::DataType::DataType::t"], [0, 3, 1, "_CPPv4N14torch_tensorrt8DataType8DataTypeEN3c1010ScalarTypeE", "torch_tensorrt::DataType::DataType::t"], [0, 4, 1, "_CPPv4N14torch_tensorrt8DataType5ValueE", "torch_tensorrt::DataType::Value"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kBoolE", "torch_tensorrt::DataType::Value::kBool"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kCharE", "torch_tensorrt::DataType::Value::kChar"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value7kDoubleE", "torch_tensorrt::DataType::Value::kDouble"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value6kFloatE", "torch_tensorrt::DataType::Value::kFloat"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kHalfE", "torch_tensorrt::DataType::Value::kHalf"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value4kIntE", "torch_tensorrt::DataType::Value::kInt"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kLongE", "torch_tensorrt::DataType::Value::kLong"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value8kUnknownE", "torch_tensorrt::DataType::Value::kUnknown"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kBoolE", "torch_tensorrt::DataType::kBool"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kCharE", "torch_tensorrt::DataType::kChar"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value7kDoubleE", "torch_tensorrt::DataType::kDouble"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value6kFloatE", "torch_tensorrt::DataType::kFloat"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kHalfE", "torch_tensorrt::DataType::kHalf"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value4kIntE", "torch_tensorrt::DataType::kInt"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kLongE", "torch_tensorrt::DataType::kLong"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value8kUnknownE", "torch_tensorrt::DataType::kUnknown"], [0, 2, 1, "_CPPv4NK14torch_tensorrt8DataTypecv5ValueEv", "torch_tensorrt::DataType::operator Value"], [0, 2, 1, "_CPPv4N14torch_tensorrt8DataTypecvbEv", "torch_tensorrt::DataType::operator bool"], [0, 2, 1, "_CPPv4NK14torch_tensorrt8DataTypeneE8DataType", "torch_tensorrt::DataType::operator!="], [0, 2, 1, "_CPPv4NK14torch_tensorrt8DataTypeneEN8DataType5ValueE", "torch_tensorrt::DataType::operator!="], [0, 3, 1, "_CPPv4NK14torch_tensorrt8DataTypeneE8DataType", "torch_tensorrt::DataType::operator!=::other"], [0, 3, 1, "_CPPv4NK14torch_tensorrt8DataTypeneEN8DataType5ValueE", "torch_tensorrt::DataType::operator!=::other"], [0, 2, 1, "_CPPv4NK14torch_tensorrt8DataTypeeqE8DataType", "torch_tensorrt::DataType::operator=="], [0, 2, 1, "_CPPv4NK14torch_tensorrt8DataTypeeqEN8DataType5ValueE", "torch_tensorrt::DataType::operator=="], [0, 3, 1, "_CPPv4NK14torch_tensorrt8DataTypeeqE8DataType", "torch_tensorrt::DataType::operator==::other"], [0, 3, 1, "_CPPv4NK14torch_tensorrt8DataTypeeqEN8DataType5ValueE", "torch_tensorrt::DataType::operator==::other"], [46, 1, 1, "_CPPv4N14torch_tensorrt6DeviceE", "torch_tensorrt::Device"], [46, 2, 1, "_CPPv4N14torch_tensorrt6Device6DeviceEv", "torch_tensorrt::Device::Device"], [1, 1, 1, "_CPPv4N14torch_tensorrt6Device10DeviceTypeE", "torch_tensorrt::Device::DeviceType"], [46, 1, 1, "_CPPv4N14torch_tensorrt6Device10DeviceTypeE", "torch_tensorrt::Device::DeviceType"], [1, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeE5Value", "torch_tensorrt::Device::DeviceType::DeviceType"], [1, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeEN3c1010DeviceTypeE", "torch_tensorrt::Device::DeviceType::DeviceType"], [1, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeEv", "torch_tensorrt::Device::DeviceType::DeviceType"], [46, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeE5Value", "torch_tensorrt::Device::DeviceType::DeviceType"], [46, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeEN3c1010DeviceTypeE", "torch_tensorrt::Device::DeviceType::DeviceType"], [46, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeEv", "torch_tensorrt::Device::DeviceType::DeviceType"], [1, 3, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeE5Value", "torch_tensorrt::Device::DeviceType::DeviceType::t"], [1, 3, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeEN3c1010DeviceTypeE", "torch_tensorrt::Device::DeviceType::DeviceType::t"], [46, 3, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeE5Value", "torch_tensorrt::Device::DeviceType::DeviceType::t"], [46, 3, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeEN3c1010DeviceTypeE", "torch_tensorrt::Device::DeviceType::DeviceType::t"], [1, 4, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5ValueE", "torch_tensorrt::Device::DeviceType::Value"], [46, 4, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5ValueE", "torch_tensorrt::Device::DeviceType::Value"], [1, 5, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5Value4kDLAE", "torch_tensorrt::Device::DeviceType::Value::kDLA"], [46, 5, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5Value4kDLAE", "torch_tensorrt::Device::DeviceType::Value::kDLA"], [1, 5, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5Value4kGPUE", "torch_tensorrt::Device::DeviceType::Value::kGPU"], [46, 5, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5Value4kGPUE", "torch_tensorrt::Device::DeviceType::Value::kGPU"], [1, 5, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5Value4kDLAE", "torch_tensorrt::Device::DeviceType::kDLA"], [1, 5, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5Value4kGPUE", "torch_tensorrt::Device::DeviceType::kGPU"], [1, 2, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypecv5ValueEv", "torch_tensorrt::Device::DeviceType::operator Value"], [46, 2, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypecv5ValueEv", "torch_tensorrt::Device::DeviceType::operator Value"], [1, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceTypecvbEv", "torch_tensorrt::Device::DeviceType::operator bool"], [46, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceTypecvbEv", "torch_tensorrt::Device::DeviceType::operator bool"], [1, 2, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeneE10DeviceType", "torch_tensorrt::Device::DeviceType::operator!="], [46, 2, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeneE10DeviceType", "torch_tensorrt::Device::DeviceType::operator!="], [1, 3, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeneE10DeviceType", "torch_tensorrt::Device::DeviceType::operator!=::other"], [46, 3, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeneE10DeviceType", "torch_tensorrt::Device::DeviceType::operator!=::other"], [1, 2, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeeqE10DeviceType", "torch_tensorrt::Device::DeviceType::operator=="], [46, 2, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeeqE10DeviceType", "torch_tensorrt::Device::DeviceType::operator=="], [1, 3, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeeqE10DeviceType", "torch_tensorrt::Device::DeviceType::operator==::other"], [46, 3, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeeqE10DeviceType", "torch_tensorrt::Device::DeviceType::operator==::other"], [46, 6, 1, "_CPPv4N14torch_tensorrt6Device18allow_gpu_fallbackE", "torch_tensorrt::Device::allow_gpu_fallback"], [46, 6, 1, "_CPPv4N14torch_tensorrt6Device11device_typeE", "torch_tensorrt::Device::device_type"], [46, 6, 1, "_CPPv4N14torch_tensorrt6Device8dla_coreE", "torch_tensorrt::Device::dla_core"], [46, 6, 1, "_CPPv4N14torch_tensorrt6Device6gpu_idE", "torch_tensorrt::Device::gpu_id"], [17, 4, 1, "_CPPv4N14torch_tensorrt16EngineCapabilityE", "torch_tensorrt::EngineCapability"], [17, 5, 1, "_CPPv4N14torch_tensorrt16EngineCapability15kDLA_STANDALONEE", "torch_tensorrt::EngineCapability::kDLA_STANDALONE"], [17, 5, 1, "_CPPv4N14torch_tensorrt16EngineCapability7kSAFETYE", "torch_tensorrt::EngineCapability::kSAFETY"], [17, 5, 1, "_CPPv4N14torch_tensorrt16EngineCapability9kSTANDARDE", "torch_tensorrt::EngineCapability::kSTANDARD"], [47, 1, 1, "_CPPv4N14torch_tensorrt11GraphInputsE", "torch_tensorrt::GraphInputs"], [47, 6, 1, "_CPPv4N14torch_tensorrt11GraphInputs15input_signatureE", "torch_tensorrt::GraphInputs::input_signature"], [47, 6, 1, "_CPPv4N14torch_tensorrt11GraphInputs6inputsE", "torch_tensorrt::GraphInputs::inputs"], [48, 1, 1, "_CPPv4N14torch_tensorrt5InputE", "torch_tensorrt::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN2at6TensorE", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEv", "torch_tensorrt::Input::Input"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN2at6TensorE", "torch_tensorrt::Input::Input::tensor"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input5dtypeE", "torch_tensorrt::Input::dtype"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input6formatE", "torch_tensorrt::Input::format"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input9max_shapeE", "torch_tensorrt::Input::max_shape"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input9min_shapeE", "torch_tensorrt::Input::min_shape"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input9opt_shapeE", "torch_tensorrt::Input::opt_shape"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input5shapeE", "torch_tensorrt::Input::shape"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input13tensor_domainE", "torch_tensorrt::Input::tensor_domain"], [2, 1, 1, "_CPPv4N14torch_tensorrt12TensorFormatE", "torch_tensorrt::TensorFormat"], [2, 2, 1, "_CPPv4N14torch_tensorrt12TensorFormat12TensorFormatE5Value", "torch_tensorrt::TensorFormat::TensorFormat"], [2, 2, 1, "_CPPv4N14torch_tensorrt12TensorFormat12TensorFormatEN2at12MemoryFormatE", "torch_tensorrt::TensorFormat::TensorFormat"], [2, 2, 1, "_CPPv4N14torch_tensorrt12TensorFormat12TensorFormatEv", "torch_tensorrt::TensorFormat::TensorFormat"], [2, 3, 1, "_CPPv4N14torch_tensorrt12TensorFormat12TensorFormatE5Value", "torch_tensorrt::TensorFormat::TensorFormat::t"], [2, 3, 1, "_CPPv4N14torch_tensorrt12TensorFormat12TensorFormatEN2at12MemoryFormatE", "torch_tensorrt::TensorFormat::TensorFormat::t"], [2, 4, 1, "_CPPv4N14torch_tensorrt12TensorFormat5ValueE", "torch_tensorrt::TensorFormat::Value"], [2, 5, 1, "_CPPv4N14torch_tensorrt12TensorFormat5Value13kChannelsLastE", "torch_tensorrt::TensorFormat::Value::kChannelsLast"], [2, 5, 1, "_CPPv4N14torch_tensorrt12TensorFormat5Value11kContiguousE", "torch_tensorrt::TensorFormat::Value::kContiguous"], [2, 5, 1, "_CPPv4N14torch_tensorrt12TensorFormat5Value8kUnknownE", "torch_tensorrt::TensorFormat::Value::kUnknown"], [2, 5, 1, "_CPPv4N14torch_tensorrt12TensorFormat5Value13kChannelsLastE", "torch_tensorrt::TensorFormat::kChannelsLast"], [2, 5, 1, "_CPPv4N14torch_tensorrt12TensorFormat5Value11kContiguousE", "torch_tensorrt::TensorFormat::kContiguous"], [2, 5, 1, "_CPPv4N14torch_tensorrt12TensorFormat5Value8kUnknownE", "torch_tensorrt::TensorFormat::kUnknown"], [2, 2, 1, "_CPPv4NK14torch_tensorrt12TensorFormatcv5ValueEv", "torch_tensorrt::TensorFormat::operator Value"], [2, 2, 1, "_CPPv4N14torch_tensorrt12TensorFormatcvbEv", "torch_tensorrt::TensorFormat::operator bool"], [2, 2, 1, "_CPPv4NK14torch_tensorrt12TensorFormatneE12TensorFormat", "torch_tensorrt::TensorFormat::operator!="], [2, 2, 1, "_CPPv4NK14torch_tensorrt12TensorFormatneEN12TensorFormat5ValueE", "torch_tensorrt::TensorFormat::operator!="], [2, 3, 1, "_CPPv4NK14torch_tensorrt12TensorFormatneE12TensorFormat", "torch_tensorrt::TensorFormat::operator!=::other"], [2, 3, 1, "_CPPv4NK14torch_tensorrt12TensorFormatneEN12TensorFormat5ValueE", "torch_tensorrt::TensorFormat::operator!=::other"], [2, 2, 1, "_CPPv4NK14torch_tensorrt12TensorFormateqE12TensorFormat", "torch_tensorrt::TensorFormat::operator=="], [2, 2, 1, "_CPPv4NK14torch_tensorrt12TensorFormateqEN12TensorFormat5ValueE", "torch_tensorrt::TensorFormat::operator=="], [2, 3, 1, "_CPPv4NK14torch_tensorrt12TensorFormateqE12TensorFormat", "torch_tensorrt::TensorFormat::operator==::other"], [2, 3, 1, "_CPPv4NK14torch_tensorrt12TensorFormateqEN12TensorFormat5ValueE", "torch_tensorrt::TensorFormat::operator==::other"], [37, 2, 1, "_CPPv4N14torch_tensorrt15dump_build_infoEv", "torch_tensorrt::dump_build_info"], [35, 2, 1, "_CPPv4N14torch_tensorrt14get_build_infoEv", "torch_tensorrt::get_build_info"], [17, 5, 1, "_CPPv4N14torch_tensorrt16EngineCapability15kDLA_STANDALONEE", "torch_tensorrt::kDLA_STANDALONE"], [17, 5, 1, "_CPPv4N14torch_tensorrt16EngineCapability7kSAFETYE", "torch_tensorrt::kSAFETY"], [17, 5, 1, "_CPPv4N14torch_tensorrt16EngineCapability9kSTANDARDE", "torch_tensorrt::kSTANDARD"], [16, 4, 1, "_CPPv4N14torch_tensorrt7logging5LevelE", "torch_tensorrt::logging::Level"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level6kDEBUGE", "torch_tensorrt::logging::Level::kDEBUG"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level6kERRORE", "torch_tensorrt::logging::Level::kERROR"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level6kGRAPHE", "torch_tensorrt::logging::Level::kGRAPH"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level5kINFOE", "torch_tensorrt::logging::Level::kINFO"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level15kINTERNAL_ERRORE", "torch_tensorrt::logging::Level::kINTERNAL_ERROR"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level8kWARNINGE", "torch_tensorrt::logging::Level::kWARNING"], [24, 2, 1, "_CPPv4N14torch_tensorrt7logging24get_is_colored_output_onEv", "torch_tensorrt::logging::get_is_colored_output_on"], [22, 2, 1, "_CPPv4N14torch_tensorrt7logging18get_logging_prefixEv", "torch_tensorrt::logging::get_logging_prefix"], [23, 2, 1, "_CPPv4N14torch_tensorrt7logging24get_reportable_log_levelEv", "torch_tensorrt::logging::get_reportable_log_level"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level6kDEBUGE", "torch_tensorrt::logging::kDEBUG"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level6kERRORE", "torch_tensorrt::logging::kERROR"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level6kGRAPHE", "torch_tensorrt::logging::kGRAPH"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level5kINFOE", "torch_tensorrt::logging::kINFO"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level15kINTERNAL_ERRORE", "torch_tensorrt::logging::kINTERNAL_ERROR"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level8kWARNINGE", "torch_tensorrt::logging::kWARNING"], [26, 2, 1, "_CPPv4N14torch_tensorrt7logging3logE5LevelNSt6stringE", "torch_tensorrt::logging::log"], [26, 3, 1, "_CPPv4N14torch_tensorrt7logging3logE5LevelNSt6stringE", "torch_tensorrt::logging::log::lvl"], [26, 3, 1, "_CPPv4N14torch_tensorrt7logging3logE5LevelNSt6stringE", "torch_tensorrt::logging::log::msg"], [27, 2, 1, "_CPPv4N14torch_tensorrt7logging24set_is_colored_output_onEb", "torch_tensorrt::logging::set_is_colored_output_on"], [27, 3, 1, "_CPPv4N14torch_tensorrt7logging24set_is_colored_output_onEb", "torch_tensorrt::logging::set_is_colored_output_on::colored_output_on"], [28, 2, 1, "_CPPv4N14torch_tensorrt7logging18set_logging_prefixENSt6stringE", "torch_tensorrt::logging::set_logging_prefix"], [28, 3, 1, "_CPPv4N14torch_tensorrt7logging18set_logging_prefixENSt6stringE", "torch_tensorrt::logging::set_logging_prefix::prefix"], [25, 2, 1, "_CPPv4N14torch_tensorrt7logging24set_reportable_log_levelE5Level", "torch_tensorrt::logging::set_reportable_log_level"], [25, 3, 1, "_CPPv4N14torch_tensorrt7logging24set_reportable_log_levelE5Level", "torch_tensorrt::logging::set_reportable_log_level::lvl"], [3, 1, 1, "_CPPv4I0EN14torch_tensorrt3ptq19Int8CacheCalibratorE", "torch_tensorrt::ptq::Int8CacheCalibrator"], [3, 7, 1, "_CPPv4I0EN14torch_tensorrt3ptq19Int8CacheCalibratorE", "torch_tensorrt::ptq::Int8CacheCalibrator::Algorithm"], [3, 2, 1, "_CPPv4N14torch_tensorrt3ptq19Int8CacheCalibrator19Int8CacheCalibratorERKNSt6stringE", "torch_tensorrt::ptq::Int8CacheCalibrator::Int8CacheCalibrator"], [3, 3, 1, "_CPPv4N14torch_tensorrt3ptq19Int8CacheCalibrator19Int8CacheCalibratorERKNSt6stringE", "torch_tensorrt::ptq::Int8CacheCalibrator::Int8CacheCalibrator::cache_file_path"], [3, 2, 1, "_CPPv4N14torch_tensorrt3ptq19Int8CacheCalibratorcvPN8nvinfer115IInt8CalibratorEEv", "torch_tensorrt::ptq::Int8CacheCalibrator::operator nvinfer1::IInt8Calibrator*"], [4, 1, 1, "_CPPv4I00EN14torch_tensorrt3ptq14Int8CalibratorE", "torch_tensorrt::ptq::Int8Calibrator"], [4, 7, 1, "_CPPv4I00EN14torch_tensorrt3ptq14Int8CalibratorE", "torch_tensorrt::ptq::Int8Calibrator::Algorithm"], [4, 7, 1, "_CPPv4I00EN14torch_tensorrt3ptq14Int8CalibratorE", "torch_tensorrt::ptq::Int8Calibrator::DataLoaderUniquePtr"], [4, 2, 1, "_CPPv4N14torch_tensorrt3ptq14Int8Calibrator14Int8CalibratorE19DataLoaderUniquePtrRKNSt6stringEb", "torch_tensorrt::ptq::Int8Calibrator::Int8Calibrator"], [4, 3, 1, "_CPPv4N14torch_tensorrt3ptq14Int8Calibrator14Int8CalibratorE19DataLoaderUniquePtrRKNSt6stringEb", "torch_tensorrt::ptq::Int8Calibrator::Int8Calibrator::cache_file_path"], [4, 3, 1, "_CPPv4N14torch_tensorrt3ptq14Int8Calibrator14Int8CalibratorE19DataLoaderUniquePtrRKNSt6stringEb", "torch_tensorrt::ptq::Int8Calibrator::Int8Calibrator::dataloader"], [4, 3, 1, "_CPPv4N14torch_tensorrt3ptq14Int8Calibrator14Int8CalibratorE19DataLoaderUniquePtrRKNSt6stringEb", "torch_tensorrt::ptq::Int8Calibrator::Int8Calibrator::use_cache"], [4, 2, 1, "_CPPv4N14torch_tensorrt3ptq14Int8CalibratorcvPN8nvinfer115IInt8CalibratorEEv", "torch_tensorrt::ptq::Int8Calibrator::operator nvinfer1::IInt8Calibrator*"], [29, 2, 1, "_CPPv4I0EN14torch_tensorrt3ptq26make_int8_cache_calibratorE19Int8CacheCalibratorI9AlgorithmERKNSt6stringE", "torch_tensorrt::ptq::make_int8_cache_calibrator"], [29, 7, 1, "_CPPv4I0EN14torch_tensorrt3ptq26make_int8_cache_calibratorE19Int8CacheCalibratorI9AlgorithmERKNSt6stringE", "torch_tensorrt::ptq::make_int8_cache_calibrator::Algorithm"], [29, 3, 1, "_CPPv4I0EN14torch_tensorrt3ptq26make_int8_cache_calibratorE19Int8CacheCalibratorI9AlgorithmERKNSt6stringE", "torch_tensorrt::ptq::make_int8_cache_calibrator::cache_file_path"], [30, 2, 1, "_CPPv4I00EN14torch_tensorrt3ptq20make_int8_calibratorE14Int8CalibratorI9Algorithm10DataLoaderE10DataLoaderRKNSt6stringEb", "torch_tensorrt::ptq::make_int8_calibrator"], [30, 7, 1, "_CPPv4I00EN14torch_tensorrt3ptq20make_int8_calibratorE14Int8CalibratorI9Algorithm10DataLoaderE10DataLoaderRKNSt6stringEb", "torch_tensorrt::ptq::make_int8_calibrator::Algorithm"], [30, 7, 1, "_CPPv4I00EN14torch_tensorrt3ptq20make_int8_calibratorE14Int8CalibratorI9Algorithm10DataLoaderE10DataLoaderRKNSt6stringEb", "torch_tensorrt::ptq::make_int8_calibrator::DataLoader"], [30, 3, 1, "_CPPv4I00EN14torch_tensorrt3ptq20make_int8_calibratorE14Int8CalibratorI9Algorithm10DataLoaderE10DataLoaderRKNSt6stringEb", "torch_tensorrt::ptq::make_int8_calibrator::cache_file_path"], [30, 3, 1, "_CPPv4I00EN14torch_tensorrt3ptq20make_int8_calibratorE14Int8CalibratorI9Algorithm10DataLoaderE10DataLoaderRKNSt6stringEb", "torch_tensorrt::ptq::make_int8_calibrator::dataloader"], [30, 3, 1, "_CPPv4I00EN14torch_tensorrt3ptq20make_int8_calibratorE14Int8CalibratorI9Algorithm10DataLoaderE10DataLoaderRKNSt6stringEb", "torch_tensorrt::ptq::make_int8_calibrator::use_cache"], [36, 2, 1, "_CPPv4N14torch_tensorrt10set_deviceEKi", "torch_tensorrt::set_device"], [36, 3, 1, "_CPPv4N14torch_tensorrt10set_deviceEKi", "torch_tensorrt::set_device::gpu_id"], [49, 1, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpecE", "torch_tensorrt::torchscript::CompileSpec"], [49, 2, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecEN5torch3jit6IValueE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec"], [49, 2, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecENSt6vectorI5InputEE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec"], [49, 2, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecENSt6vectorIN3c108ArrayRefI7int64_tEEEE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec"], [49, 2, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecENSt6vectorINSt6vectorI7int64_tEEEE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec"], [49, 3, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecENSt6vectorIN3c108ArrayRefI7int64_tEEEE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec::fixed_sizes"], [49, 3, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecENSt6vectorINSt6vectorI7int64_tEEEE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec::fixed_sizes"], [49, 3, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecEN5torch3jit6IValueE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec::input_signature"], [49, 3, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecENSt6vectorI5InputEE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec::inputs"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec19allow_shape_tensorsE", "torch_tensorrt::torchscript::CompileSpec::allow_shape_tensors"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec10capabilityE", "torch_tensorrt::torchscript::CompileSpec::capability"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec5debugE", "torch_tensorrt::torchscript::CompileSpec::debug"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec6deviceE", "torch_tensorrt::torchscript::CompileSpec::device"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec12disable_tf32E", "torch_tensorrt::torchscript::CompileSpec::disable_tf32"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec20dla_global_dram_sizeE", "torch_tensorrt::torchscript::CompileSpec::dla_global_dram_size"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec19dla_local_dram_sizeE", "torch_tensorrt::torchscript::CompileSpec::dla_local_dram_size"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec13dla_sram_sizeE", "torch_tensorrt::torchscript::CompileSpec::dla_sram_size"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec18enabled_precisionsE", "torch_tensorrt::torchscript::CompileSpec::enabled_precisions"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec12graph_inputsE", "torch_tensorrt::torchscript::CompileSpec::graph_inputs"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec14min_block_sizeE", "torch_tensorrt::torchscript::CompileSpec::min_block_size"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec20num_avg_timing_itersE", "torch_tensorrt::torchscript::CompileSpec::num_avg_timing_iters"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec14ptq_calibratorE", "torch_tensorrt::torchscript::CompileSpec::ptq_calibrator"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec5refitE", "torch_tensorrt::torchscript::CompileSpec::refit"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec24require_full_compilationE", "torch_tensorrt::torchscript::CompileSpec::require_full_compilation"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec14sparse_weightsE", "torch_tensorrt::torchscript::CompileSpec::sparse_weights"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec22torch_executed_modulesE", "torch_tensorrt::torchscript::CompileSpec::torch_executed_modules"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec18torch_executed_opsE", "torch_tensorrt::torchscript::CompileSpec::torch_executed_ops"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec24truncate_long_and_doubleE", "torch_tensorrt::torchscript::CompileSpec::truncate_long_and_double"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec14workspace_sizeE", "torch_tensorrt::torchscript::CompileSpec::workspace_size"], [31, 2, 1, "_CPPv4N14torch_tensorrt11torchscript29check_method_operator_supportERKN5torch3jit6ModuleENSt6stringE", "torch_tensorrt::torchscript::check_method_operator_support"], [31, 3, 1, "_CPPv4N14torch_tensorrt11torchscript29check_method_operator_supportERKN5torch3jit6ModuleENSt6stringE", "torch_tensorrt::torchscript::check_method_operator_support::method_name"], [31, 3, 1, "_CPPv4N14torch_tensorrt11torchscript29check_method_operator_supportERKN5torch3jit6ModuleENSt6stringE", "torch_tensorrt::torchscript::check_method_operator_support::module"], [32, 2, 1, "_CPPv4N14torch_tensorrt11torchscript7compileERKN5torch3jit6ModuleE11CompileSpec", "torch_tensorrt::torchscript::compile"], [32, 3, 1, "_CPPv4N14torch_tensorrt11torchscript7compileERKN5torch3jit6ModuleE11CompileSpec", "torch_tensorrt::torchscript::compile::info"], [32, 3, 1, "_CPPv4N14torch_tensorrt11torchscript7compileERKN5torch3jit6ModuleE11CompileSpec", "torch_tensorrt::torchscript::compile::module"], [34, 2, 1, "_CPPv4N14torch_tensorrt11torchscript28convert_method_to_trt_engineERKN5torch3jit6ModuleENSt6stringE11CompileSpec", "torch_tensorrt::torchscript::convert_method_to_trt_engine"], [34, 3, 1, "_CPPv4N14torch_tensorrt11torchscript28convert_method_to_trt_engineERKN5torch3jit6ModuleENSt6stringE11CompileSpec", "torch_tensorrt::torchscript::convert_method_to_trt_engine::info"], [34, 3, 1, "_CPPv4N14torch_tensorrt11torchscript28convert_method_to_trt_engineERKN5torch3jit6ModuleENSt6stringE11CompileSpec", "torch_tensorrt::torchscript::convert_method_to_trt_engine::method_name"], [34, 3, 1, "_CPPv4N14torch_tensorrt11torchscript28convert_method_to_trt_engineERKN5torch3jit6ModuleENSt6stringE11CompileSpec", "torch_tensorrt::torchscript::convert_method_to_trt_engine::module"], [33, 2, 1, "_CPPv4N14torch_tensorrt11torchscript26embed_engine_in_new_moduleERKNSt6stringE6DeviceRKNSt6vectorINSt6stringEEERKNSt6vectorINSt6stringEEE", "torch_tensorrt::torchscript::embed_engine_in_new_module"], [33, 3, 1, "_CPPv4N14torch_tensorrt11torchscript26embed_engine_in_new_moduleERKNSt6stringE6DeviceRKNSt6vectorINSt6stringEEERKNSt6vectorINSt6stringEEE", "torch_tensorrt::torchscript::embed_engine_in_new_module::device"], [33, 3, 1, "_CPPv4N14torch_tensorrt11torchscript26embed_engine_in_new_moduleERKNSt6stringE6DeviceRKNSt6vectorINSt6stringEEERKNSt6vectorINSt6stringEEE", "torch_tensorrt::torchscript::embed_engine_in_new_module::engine"], [33, 3, 1, "_CPPv4N14torch_tensorrt11torchscript26embed_engine_in_new_moduleERKNSt6stringE6DeviceRKNSt6vectorINSt6stringEEERKNSt6vectorINSt6stringEEE", "torch_tensorrt::torchscript::embed_engine_in_new_module::input_binding_names"], [33, 3, 1, "_CPPv4N14torch_tensorrt11torchscript26embed_engine_in_new_moduleERKNSt6stringE6DeviceRKNSt6vectorINSt6stringEEERKNSt6vectorINSt6stringEEE", "torch_tensorrt::torchscript::embed_engine_in_new_module::output_binding_names"], [76, 8, 0, "-", "torch_tensorrt"]], "torch_tensorrt": [[76, 9, 1, "", "Device"], [76, 9, 1, "", "DeviceType"], [76, 9, 1, "", "EngineCapability"], [76, 9, 1, "", "Input"], [76, 9, 1, "", "MutableTorchTensorRTModule"], [76, 12, 1, "", "compile"], [76, 12, 1, "", "convert_method_to_trt_engine"], [76, 9, 1, "", "dtype"], [115, 8, 0, "-", "dynamo"], [72, 8, 0, "-", "fx"], [76, 12, 1, "", "load"], [73, 8, 0, "-", "logging"], [76, 9, 1, "", "memory_format"], [75, 8, 0, "-", "runtime"], [76, 12, 1, "", "save"], [77, 8, 0, "-", "ts"]], "torch_tensorrt.Device": [[76, 10, 1, "", "__init__"], [76, 11, 1, "", "device_type"], [76, 11, 1, "", "dla_core"], [76, 11, 1, "", "gpu_id"]], "torch_tensorrt.DeviceType": [[76, 11, 1, "", "DLA"], [76, 11, 1, "", "GPU"], [76, 11, 1, "", "UNKNOWN"], [76, 10, 1, "", "to"], [76, 10, 1, "", "try_from"], [76, 10, 1, "", "try_to"]], "torch_tensorrt.EngineCapability": [[76, 11, 1, "", "DLA_STANDALONE"], [76, 11, 1, "", "SAFETY"], [76, 11, 1, "", "STANDARD"], [76, 10, 1, "", "to"], [76, 10, 1, "", "try_from"], [76, 10, 1, "", "try_to"]], "torch_tensorrt.Input": [[76, 10, 1, "", "__init__"], [76, 11, 1, "", "dtype"], [76, 10, 1, "", "example_tensor"], [76, 11, 1, "", "format"], [76, 10, 1, "", "from_tensor"], [76, 10, 1, "", "from_tensors"]], "torch_tensorrt.MutableTorchTensorRTModule": [[76, 10, 1, "", "__init__"], [76, 10, 1, "", "compile"], [76, 10, 1, "", "refit_gm"]], "torch_tensorrt.dtype": [[76, 11, 1, "", "b"], [76, 11, 1, "", "bf16"], [76, 11, 1, "", "f16"], [76, 11, 1, "", "f32"], [76, 11, 1, "", "f64"], [76, 11, 1, "", "f8"], [76, 11, 1, "", "i32"], [76, 11, 1, "", "i64"], [76, 11, 1, "", "i8"], [76, 10, 1, "", "to"], [76, 10, 1, "", "try_from"], [76, 10, 1, "", "try_to"], [76, 11, 1, "", "u8"], [76, 11, 1, "", "unknown"]], "torch_tensorrt.dynamo": [[71, 9, 1, "", "CompilationSettings"], [71, 12, 1, "", "compile"], [71, 12, 1, "", "export"], [71, 12, 1, "", "refit_module_weights"], [71, 12, 1, "", "trace"]], "torch_tensorrt.fx": [[72, 9, 1, "", "InputTensorSpec"], [72, 9, 1, "", "TRTInterpreter"], [72, 9, 1, "", "TRTInterpreterResult"], [72, 9, 1, "", "TRTModule"], [72, 12, 1, "", "compile"]], "torch_tensorrt.logging": [[73, 9, 1, "", "debug"], [73, 9, 1, "", "errors"], [73, 9, 1, "", "graphs"], [73, 9, 1, "", "info"], [73, 9, 1, "", "internal_errors"], [73, 9, 1, "", "warnings"]], "torch_tensorrt.memory_format": [[76, 11, 1, "", "cdhw32"], [76, 11, 1, "", "chw16"], [76, 11, 1, "", "chw2"], [76, 11, 1, "", "chw32"], [76, 11, 1, "", "chw4"], [76, 11, 1, "", "dhwc"], [76, 11, 1, "", "dhwc8"], [76, 11, 1, "", "dla_hwc4"], [76, 11, 1, "", "dla_linear"], [76, 11, 1, "", "hwc"], [76, 11, 1, "", "hwc16"], [76, 11, 1, "", "hwc8"], [76, 11, 1, "", "linear"], [76, 10, 1, "", "to"], [76, 10, 1, "", "try_from"], [76, 10, 1, "", "try_to"]], "torch_tensorrt.runtime": [[75, 9, 1, "", "PythonTorchTensorRTModule"], [75, 9, 1, "", "TorchTensorRTModule"], [75, 12, 1, "", "set_multi_device_safe_mode"]], "torch_tensorrt.runtime.PythonTorchTensorRTModule": [[75, 10, 1, "", "__init__"], [75, 10, 1, "", "cudagraphs_validate_shapes"], [75, 10, 1, "", "disable_profiling"], [75, 10, 1, "", "enable_profiling"], [75, 10, 1, "", "forward"], [75, 10, 1, "", "get_layer_info"]], "torch_tensorrt.runtime.TorchTensorRTModule": [[75, 10, 1, "", "__init__"], [75, 10, 1, "", "forward"], [75, 10, 1, "", "get_extra_state"], [75, 10, 1, "", "set_extra_state"]], "torch_tensorrt.ts": [[77, 12, 1, "", "TensorRTCompileSpec"], [77, 12, 1, "", "check_method_op_support"], [77, 12, 1, "", "compile"], [77, 12, 1, "", "convert_method_to_trt_engine"], [77, 12, 1, "", "embed_engine_in_new_module"], [74, 8, 0, "-", "ptq"]], "torch_tensorrt.ts.ptq": [[74, 9, 1, "", "CacheCalibrator"], [74, 9, 1, "", "CalibrationAlgo"], [74, 9, 1, "", "DataLoaderCalibrator"]], "torch_tensorrt.ts.ptq.CalibrationAlgo": [[74, 11, 1, "", "ENTROPY_CALIBRATION"], [74, 11, 1, "", "ENTROPY_CALIBRATION_2"], [74, 11, 1, "", "LEGACY_CALIBRATION"], [74, 11, 1, "", "MINMAX_CALIBRATION"]]}, "objtypes": {"0": "c:macro", "1": "cpp:class", "2": "cpp:function", "3": "cpp:functionParam", "4": "cpp:enum", "5": "cpp:enumerator", "6": "cpp:member", "7": "cpp:templateParam", "8": "py:module", "9": "py:class", "10": "py:method", "11": "py:attribute", "12": "py:function"}, "objnames": {"0": ["c", "macro", "C macro"], "1": ["cpp", "class", "C++ class"], "2": ["cpp", "function", "C++ function"], "3": ["cpp", "functionParam", "C++ function parameter"], "4": ["cpp", "enum", "C++ enum"], "5": ["cpp", "enumerator", "C++ enumerator"], "6": ["cpp", "member", "C++ member"], "7": ["cpp", "templateParam", "C++ template parameter"], "8": ["py", "module", "Python module"], "9": ["py", "class", "Python class"], "10": ["py", "method", "Python method"], "11": ["py", "attribute", "Python attribute"], "12": ["py", "function", "Python function"]}, "titleterms": {"class": [0, 1, 2, 3, 4, 20, 21, 38, 40, 41, 50, 71, 72, 74, 75, 76], "datatyp": 0, "document": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 16, 17, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 46, 47, 48, 49, 61, 69, 85, 86], "devic": [1, 46, 114], "devicetyp": 1, "nest": [1, 46], "relationship": [1, 3, 4, 46, 48], "tensorformat": 2, "templat": [3, 4, 29, 30], "int8cachecalibr": 3, "inherit": [3, 4, 48], "base": [3, 4, 48, 80], "type": [3, 4, 46, 48, 54], "int8calibr": 4, "defin": [5, 6, 7, 8, 9, 10, 11, 12, 19, 50, 107], "str": 5, "torch_tensorrt_patch_vers": 6, "torch_tensorrt_major_vers": 7, "torch_tensorrt_minor_vers": 8, "torchtrt_api": 9, "xstr": 10, "torchtrt_hidden": 11, "torch_tensorrt_vers": 12, "directori": [13, 14, 15, 51], "cpp": [13, 18, 19, 20, 21, 56], "subdirectori": [13, 14], "includ": [14, 18, 19, 20, 21], "torch_tensorrt": [15, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 45, 67, 71, 72, 73, 74, 75, 76, 77, 101, 103, 104, 116], "file": [15, 18, 19, 20, 21, 42, 43, 44, 45, 50, 51], "enum": [16, 17, 38, 39, 50, 74, 76], "level": [16, 80, 82, 83], "enginecap": 17, "log": [18, 22, 23, 24, 25, 26, 27, 28, 39, 42, 73], "h": [18, 19, 20, 21, 42, 43, 44, 45, 56], "content": [18, 19, 20, 21, 38, 39, 40, 41, 80, 81, 82, 83, 84, 85], "definit": [18, 19, 20, 21, 83, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108], "By": [18, 19], "namespac": [18, 19, 20, 21, 38, 39, 40, 41, 50], "macro": [19, 43], "ptq": [20, 29, 30, 40, 44, 74, 91, 107], "function": [22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 50, 61, 71, 72, 75, 76, 77, 107], "get_logging_prefix": 22, "get_reportable_log_level": 23, "get_is_colored_output_on": 24, "set_reportable_log_level": 25, "set_is_colored_output_on": 27, "set_logging_prefix": 28, "make_int8_cache_calibr": 29, "make_int8_calibr": 30, "torchscript": [31, 32, 33, 34, 41, 60, 66, 69, 88, 89, 92, 115, 116], "check_method_operator_support": 31, "compil": [32, 57, 59, 63, 64, 66, 68, 69, 89, 96, 99, 100, 101, 102, 103, 104, 105, 106, 108, 110, 112, 113, 115, 116], "embed_engine_in_new_modul": 33, "convert_method_to_trt_engin": 34, "get_build_info": 35, "set_devic": 36, "dump_build_info": 37, "program": [42, 43, 44, 45, 63, 99, 114], "list": [42, 43, 44, 45, 83], "struct": [46, 47, 48, 49, 50], "graphinput": 47, "input": [48, 101, 103], "compilespec": 49, "torch": [50, 61, 63, 64, 65, 66, 68, 69, 89, 90, 92, 93, 94, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 109, 110, 111, 112, 113, 114, 115, 116], "tensorrt": [50, 58, 61, 63, 64, 65, 66, 69, 89, 90, 92, 93, 94, 98, 99, 105, 106, 107, 109, 110, 111, 112, 113, 114, 115, 116], "c": [50, 61, 66, 68, 69, 89, 91, 110], "api": [50, 51, 61, 66, 69], "hierarchi": 50, "full": [50, 51], "torchtrtc": [52, 89], "convers": [53, 57, 59, 60], "phase": [53, 55, 56, 57, 58, 59], "node": 53, "evalu": [53, 54, 70], "convert": [53, 54, 60, 65, 70, 89, 93], "write": [54, 60, 62, 93, 94], "dynamo": [54, 62, 69, 71, 105, 106, 115, 116], "implement": [54, 93], "registr": 54, "capabl": 54, "valid": 54, "contract": [54, 60], "exampl": [54, 62, 82, 84], "convolut": 54, "oper": [54, 64, 70, 89, 94], "decomposit": 54, "addmm": [54, 55], "lower": [55, 57, 59, 62], "pass": [55, 62], "us": [55, 61, 89, 90, 92, 93, 94, 100, 101, 102, 103, 104, 105, 106, 107, 110, 112], "eliminatecommonsubexpress": 55, "elimin": 55, "dead": 55, "code": [55, 69, 82], "except": 55, "Or": 55, "pattern": 55, "redund": 55, "guard": 55, "freez": 55, "modul": [55, 88, 89, 98, 116], "fuse": 55, "branch": 55, "linear": 55, "flatten": 55, "graph": [55, 58, 116], "tupl": 55, "fallback": [55, 56], "peephol": 55, "optim": [55, 68, 111], "remov": 55, "contigu": 55, "dropout": 55, "To": 55, "unpack": 55, "logsoftmax": 55, "unrol": 55, "loop": [55, 107], "replac": [55, 82], "tile": 55, "repeat": 55, "partit": [56, 57, 59], "partitoninfo": 56, "segmentedblock": 56, "shape_analysi": 56, "automat": [56, 108], "depend": [56, 66, 97, 109], "awar": [56, 110], "runtim": [57, 58, 59, 75, 114], "background": [58, 60], "engin": [58, 65, 94, 95, 96], "executor": 58, "op": [58, 65, 94], "construct": 58, "result": 58, "serial": [58, 64, 68], "deseri": 58, "abi": [58, 66], "version": [58, 66], "format": [58, 116], "system": [59, 66], "overview": [59, 67], "what": 60, "guarante": 60, "respons": 60, "context": [60, 80, 108], "arg": [60, 81], "weight": [60, 99, 107, 108], "other": 60, "advic": 60, "link": [61, 82], "develop": 61, "avail": 61, "layer": 61, "expect": 61, "dimens": 61, "python": [61, 66, 68, 69, 88, 90, 91], "sometim": 61, "easier": 61, "read": 61, "pytorch": [61, 65, 69, 92, 94, 105, 106, 110], "native_op": 61, "ir": [61, 115, 116], "aten": 62, "basic": 62, "requir": 62, "regist": [62, 89], "export": [63, 68, 104, 112], "customiz": [63, 64], "set": [63, 64, 98, 100, 104, 111], "under": [63, 89, 112], "hood": [63, 89, 112], "trace": 63, "backend": [64, 101, 102, 103, 105, 106], "kei": 64, "featur": 64, "custom": [64, 89, 93, 94, 96, 100, 112], "usag": [64, 99, 100], "after": 64, "model": [64, 65, 69, 94, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 113, 115], "perform": 64, "coverag": 64, "feasibl": 64, "dynam": [64, 101, 110, 112], "shape": [64, 101, 110, 112], "support": [64, 70], "recompil": [64, 101], "condit": 64, "fx": [65, 69, 72, 110, 116], "frontend": [65, 66, 69, 92, 110, 116], "user": [65, 69], "guid": [65, 69], "acc": 65, "tracer": 65, "fx2trt": 65, "how": [65, 80, 91], "add": 65, "miss": 65, "instal": [66, 87], "precompil": 66, "binari": 66, "specif": 66, "cuda": [66, 100, 103], "nightli": 66, "build": [66, 67, 80, 111], "onli": 66, "from": [66, 92], "sourc": 66, "linux": 66, "packag": [66, 114], "addit": 66, "option": [66, 68, 80, 81, 83, 101, 103, 108, 116], "distribut": 66, "No": 66, "librari": [66, 114], "standalon": 66, "releas": 66, "debug": 66, "pre": [66, 107], "cxx11": 66, "choos": 66, "right": 66, "window": 66, "step": [66, 68, 111], "advanc": [66, 99, 100], "setup": 66, "troubleshoot": 66, "altern": 66, "cmake": 66, "nativ": 66, "aarch64": 66, "jetson": 66, "prerequisit": [66, 67], "environ": 66, "cli": [66, 69], "jetpack": 67, "6": [67, 84], "1": [67, 68, 84, 111], "quick": 68, "start": [68, 69], "2": [68, 84, 85, 111], "deploi": [68, 107, 110, 114], "deploy": 68, "In": [69, 99], "framework": 69, "infer": [69, 101, 102, 103, 104, 107, 111], "nvidia": 69, "gpu": 69, "get": 69, "tutori": [69, 109], "zoo": [69, 97, 109], "contributor": 69, "indic": 69, "legaci": [69, 110, 116], "further": 69, "inform": 69, "current": 70, "through": 70, "ts": [74, 77, 116], "submodul": 76, "comput": 78, "time": [78, 116], "changelog": 79, "configur": 80, "project": 80, "wide": 80, "html": 80, "theme": [80, 86], "toc": 80, "page": 80, "tabl": [80, 81, 82, 83, 84, 85], "mod": 81, "test_py_modul": 81, "gener": [81, 105, 106], "index": 81, "paramet": 81, "data": 81, "paragraph": [82, 85], "markup": 82, "inlin": 82, "math": 82, "meta": 82, "block": 82, "liter": 82, "line": 82, "quot": 82, "doctest": 82, "emphas": 82, "number": [82, 83], "sidebar": 82, "ch": 82, "ien": 82, "The": [82, 89], "creativ": 82, "A": 82, "refer": 82, "footnot": 82, "citat": [82, 91], "glossari": 82, "target": 82, "direct": 82, "center": 82, "text": 82, "imag": [82, 83], "figur": 82, "admonit": 82, "And": 82, "wai": 82, "topic": 82, "rubric": 82, "titl": 82, "compound": 82, "download": [82, 87], "enumer": 83, "field": 83, "bullet": 83, "second": 83, "But": 83, "deeper": 83, "down": 83, "rabbit": 83, "hole": 83, "hlist": 83, "grid": 83, "giant": 83, "can": 83, "have": 83, "caption": [83, 86], "like": 83, "thi": [83, 86], "one": 83, "long": [84, 86], "sticki": 84, "nav": 84, "menu": [84, 86], "3": [84, 111], "4": 84, "5": 84, "7": 84, "8": 84, "9": 84, "10": 84, "11": 84, "12": 84, "13": 84, "14": 84, "15": 84, "16": 84, "17": 84, "18": 84, "19": 84, "20": 84, "submenu": 84, "subsubmenu": 84, "structur": 85, "element": 85, "section": 85, "subsect": 85, "subsubsect": 85, "demo": 86, "an": 86, "incred": 86, "via": 87, "git": 87, "creat": [88, 91], "work": [88, 89], "save": [88, 98, 115], "disk": 88, "quickstart": 89, "unsupport": 89, "post": 91, "train": [91, 107, 110], "quantiz": [91, 107, 110], "your": [91, 111], "own": 91, "applic": 91, "directli": 92, "overload": 93, "metadata": 93, "our": [93, 94], "kernel": 94, "within": 94, "test": 94, "wrap": 94, "insert": 94, "cach": [95, 96, 99], "bert": [95, 103, 110], "jit": [96, 112], "aot": [96, 112], "mutabl": 98, "initi": 98, "make": [98, 99], "modif": 98, "stabl": [98, 102], "diffus": [98, 102], "huggingfac": 98, "refit": 99, "new": 99, "standard": 99, "workflow": 99, "import": [99, 100, 101, 102, 103, 104, 105, 106, 107, 108], "refitt": 99, "pretrain": 99, "map": 99, "place": 99, "default": [100, 104], "cleanup": [100, 103], "driver": [100, 103], "error": [100, 103], "note": [100, 103], "resnet": 101, "argument": [101, 103], "avoid": 101, "specifi": 101, "befor": 101, "trt": 101, "cudagraph": [104, 114], "integr": 104, "gpt2": 105, "output": [105, 106], "decod": [105, 106], "sentenc": [105, 106], "llama2": 106, "load": [107, 115], "dataset": 107, "loss": 107, "calibr": 107, "tune": 107, "fp8": 107, "stream": 108, "run": 108, "budget": 108, "size": 108, "manag": 108, "notebook": 110, "citrinet": 110, "efficientnet": 110, "mask": 110, "languag": 110, "mlm": 110, "hug": 110, "face": 110, "transform": 110, "acceler": 110, "serv": [110, 111], "resnet50": 110, "lenet": 110, "deep": 110, "learn": 110, "object": 110, "detect": 110, "ssd": 110, "int8": 110, "triton": 111, "up": 111, "server": 111, "client": 111, "queri": 111, "constraint": 112, "mix": 113, "precis": 113, "libtorchtrt": 114, "so": 114, "plugin": 114, "multi": 114, "safe": 114, "mode": 114, "exportedprogram": 115, "b": 115, "explain": 116, "just": 116, "accept": 116, "return": 116, "ahead": 116, "dla": 117}, "envversion": {"sphinx.domains.c": 2, "sphinx.domains.changeset": 1, "sphinx.domains.citation": 1, "sphinx.domains.cpp": 6, "sphinx.domains.index": 1, "sphinx.domains.javascript": 2, "sphinx.domains.math": 2, "sphinx.domains.python": 3, "sphinx.domains.rst": 2, "sphinx.domains.std": 2, "nbsphinx": 4, "sphinx.ext.intersphinx": 1, "sphinx.ext.todo": 2, "sphinx.ext.viewcode": 1, "sphinx": 56}}) \ No newline at end of file +Search.setIndex({"docnames": ["_cpp_api/classtorch__tensorrt_1_1DataType", "_cpp_api/classtorch__tensorrt_1_1Device_1_1DeviceType", "_cpp_api/classtorch__tensorrt_1_1TensorFormat", "_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8CacheCalibrator", "_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8Calibrator", "_cpp_api/define_macros_8h_1a18d295a837ac71add5578860b55e5502", "_cpp_api/define_macros_8h_1a282fd3c0b1c3a215148ae372070e1268", "_cpp_api/define_macros_8h_1a31398a6d4d27e28817afb0f0139e909e", "_cpp_api/define_macros_8h_1a35703561b26b1a9d2738ad7d58b27827", "_cpp_api/define_macros_8h_1abd1465eb38256d3f22cc1426b23d516b", "_cpp_api/define_macros_8h_1abe87b341f562fd1cf40b7672e4d759da", "_cpp_api/define_macros_8h_1ad19939408f7be171a74a89928b36eb59", "_cpp_api/define_macros_8h_1adad592a7b1b7eed529cdf6acd584c883", "_cpp_api/dir_cpp", "_cpp_api/dir_cpp_include", "_cpp_api/dir_cpp_include_torch_tensorrt", "_cpp_api/enum_namespacetorch__tensorrt_1_1logging_1a130f65408ad8cbaee060f05e8db69558", "_cpp_api/enum_namespacetorch__tensorrt_1a3fbe5d72e4fc624dbd038853079620eb", "_cpp_api/file_cpp_include_torch_tensorrt_logging.h", "_cpp_api/file_cpp_include_torch_tensorrt_macros.h", "_cpp_api/file_cpp_include_torch_tensorrt_ptq.h", "_cpp_api/file_cpp_include_torch_tensorrt_torch_tensorrt.h", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0593f776f469c20469e2f729fc7861a3", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0c012cb374addd90eb1f42eaec570650", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a56e110feaaba2c3fd44bd201fd21a76a", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a7cb50492421ea9de4e3db895819df6f2", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ac46ac0901cb97e3ae6e93b45f24e90b8", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ad2efd47b6c3689e58ccc595680579ae5", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1af8f3443813315af7901903d25dd495cc", "_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a226e3c83379d1012cde8578c1c86b16c", "_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a6186e305f47c1d94b6130ef6c7f7e178", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a5b405fd3bf3c8fc2e2a54cbbab979797", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a6e19490a08fb1553c9dd347a5ae79db9", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a81f9783517335dda877d8cfcf38987c9", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1ae8d56472106eeef37fbe51ff7f40c9b2", "_cpp_api/function_namespacetorch__tensorrt_1ac4ab8313ae72c2c899ea31548b528528", "_cpp_api/function_namespacetorch__tensorrt_1ad1acd06eaeaffbbcf6e7ebf426891384", "_cpp_api/function_namespacetorch__tensorrt_1ad6a4ee8ca6c8f6e5519eb1128ec7f4a1", "_cpp_api/namespace_torch_tensorrt", "_cpp_api/namespace_torch_tensorrt__logging", "_cpp_api/namespace_torch_tensorrt__ptq", "_cpp_api/namespace_torch_tensorrt__torchscript", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_logging.h", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_macros.h", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_ptq.h", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_torch_tensorrt.h", "_cpp_api/structtorch__tensorrt_1_1Device", "_cpp_api/structtorch__tensorrt_1_1GraphInputs", "_cpp_api/structtorch__tensorrt_1_1Input", "_cpp_api/structtorch__tensorrt_1_1torchscript_1_1CompileSpec", "_cpp_api/torch_tensort_cpp", "_cpp_api/unabridged_orphan", "cli/torchtrtc", "contributors/conversion", "contributors/dynamo_converters", "contributors/lowering", "contributors/partitioning", "contributors/phases", "contributors/runtime", "contributors/system_overview", "contributors/ts_converters", "contributors/useful_links", "contributors/writing_dynamo_aten_lowering_passes", "dynamo/dynamo_export", "dynamo/torch_compile", "fx/getting_started_with_fx_path", "getting_started/installation", "getting_started/jetpack", "getting_started/quick_start", "index", "indices/supported_ops", "py_api/dynamo", "py_api/fx", "py_api/logging", "py_api/ptq", "py_api/runtime", "py_api/torch_tensorrt", "py_api/ts", "sg_execution_times", "src/pytorch-sphinx-theme/docs/changelog", "src/pytorch-sphinx-theme/docs/configuring", "src/pytorch-sphinx-theme/docs/demo/api", "src/pytorch-sphinx-theme/docs/demo/demo", "src/pytorch-sphinx-theme/docs/demo/lists_tables", "src/pytorch-sphinx-theme/docs/demo/long", "src/pytorch-sphinx-theme/docs/demo/structure", "src/pytorch-sphinx-theme/docs/index", "src/pytorch-sphinx-theme/docs/installing", "ts/creating_torchscript_module_in_python", "ts/getting_started_with_cpp_api", "ts/getting_started_with_python_api", "ts/ptq", "ts/torchscript_frontend_from_pytorch", "tutorials/_rendered_examples/dynamo/converter_overloading", "tutorials/_rendered_examples/dynamo/cross_runtime_compilation_for_windows", "tutorials/_rendered_examples/dynamo/custom_kernel_plugins", "tutorials/_rendered_examples/dynamo/engine_caching_bert_example", "tutorials/_rendered_examples/dynamo/engine_caching_example", "tutorials/_rendered_examples/dynamo/index", "tutorials/_rendered_examples/dynamo/mutable_torchtrt_module_example", "tutorials/_rendered_examples/dynamo/refit_engine_example", "tutorials/_rendered_examples/dynamo/torch_compile_advanced_usage", "tutorials/_rendered_examples/dynamo/torch_compile_resnet_example", "tutorials/_rendered_examples/dynamo/torch_compile_stable_diffusion", "tutorials/_rendered_examples/dynamo/torch_compile_transformers_example", "tutorials/_rendered_examples/dynamo/torch_export_cudagraphs", "tutorials/_rendered_examples/dynamo/torch_export_gpt2", "tutorials/_rendered_examples/dynamo/torch_export_llama2", "tutorials/_rendered_examples/dynamo/vgg16_ptq", "tutorials/_rendered_examples/dynamo/weight_streaming_example", "tutorials/_rendered_examples/index", "tutorials/notebooks", "tutorials/serving_torch_tensorrt_with_triton", "user_guide/dynamic_shapes", "user_guide/mixed_precision", "user_guide/runtime", "user_guide/saving_models", "user_guide/torch_tensorrt_explained", "user_guide/using_dla"], "filenames": ["_cpp_api/classtorch__tensorrt_1_1DataType.rst", "_cpp_api/classtorch__tensorrt_1_1Device_1_1DeviceType.rst", "_cpp_api/classtorch__tensorrt_1_1TensorFormat.rst", "_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8CacheCalibrator.rst", "_cpp_api/classtorch__tensorrt_1_1ptq_1_1Int8Calibrator.rst", "_cpp_api/define_macros_8h_1a18d295a837ac71add5578860b55e5502.rst", "_cpp_api/define_macros_8h_1a282fd3c0b1c3a215148ae372070e1268.rst", "_cpp_api/define_macros_8h_1a31398a6d4d27e28817afb0f0139e909e.rst", "_cpp_api/define_macros_8h_1a35703561b26b1a9d2738ad7d58b27827.rst", "_cpp_api/define_macros_8h_1abd1465eb38256d3f22cc1426b23d516b.rst", "_cpp_api/define_macros_8h_1abe87b341f562fd1cf40b7672e4d759da.rst", "_cpp_api/define_macros_8h_1ad19939408f7be171a74a89928b36eb59.rst", "_cpp_api/define_macros_8h_1adad592a7b1b7eed529cdf6acd584c883.rst", "_cpp_api/dir_cpp.rst", "_cpp_api/dir_cpp_include.rst", "_cpp_api/dir_cpp_include_torch_tensorrt.rst", "_cpp_api/enum_namespacetorch__tensorrt_1_1logging_1a130f65408ad8cbaee060f05e8db69558.rst", "_cpp_api/enum_namespacetorch__tensorrt_1a3fbe5d72e4fc624dbd038853079620eb.rst", "_cpp_api/file_cpp_include_torch_tensorrt_logging.h.rst", "_cpp_api/file_cpp_include_torch_tensorrt_macros.h.rst", "_cpp_api/file_cpp_include_torch_tensorrt_ptq.h.rst", "_cpp_api/file_cpp_include_torch_tensorrt_torch_tensorrt.h.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0593f776f469c20469e2f729fc7861a3.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a0c012cb374addd90eb1f42eaec570650.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a56e110feaaba2c3fd44bd201fd21a76a.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1a7cb50492421ea9de4e3db895819df6f2.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ac46ac0901cb97e3ae6e93b45f24e90b8.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1ad2efd47b6c3689e58ccc595680579ae5.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1logging_1af8f3443813315af7901903d25dd495cc.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a226e3c83379d1012cde8578c1c86b16c.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1ptq_1a6186e305f47c1d94b6130ef6c7f7e178.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a5b405fd3bf3c8fc2e2a54cbbab979797.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a6e19490a08fb1553c9dd347a5ae79db9.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1a81f9783517335dda877d8cfcf38987c9.rst", "_cpp_api/function_namespacetorch__tensorrt_1_1torchscript_1ae8d56472106eeef37fbe51ff7f40c9b2.rst", "_cpp_api/function_namespacetorch__tensorrt_1ac4ab8313ae72c2c899ea31548b528528.rst", "_cpp_api/function_namespacetorch__tensorrt_1ad1acd06eaeaffbbcf6e7ebf426891384.rst", "_cpp_api/function_namespacetorch__tensorrt_1ad6a4ee8ca6c8f6e5519eb1128ec7f4a1.rst", "_cpp_api/namespace_torch_tensorrt.rst", "_cpp_api/namespace_torch_tensorrt__logging.rst", "_cpp_api/namespace_torch_tensorrt__ptq.rst", "_cpp_api/namespace_torch_tensorrt__torchscript.rst", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_logging.h.rst", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_macros.h.rst", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_ptq.h.rst", "_cpp_api/program_listing_file_cpp_include_torch_tensorrt_torch_tensorrt.h.rst", "_cpp_api/structtorch__tensorrt_1_1Device.rst", "_cpp_api/structtorch__tensorrt_1_1GraphInputs.rst", "_cpp_api/structtorch__tensorrt_1_1Input.rst", "_cpp_api/structtorch__tensorrt_1_1torchscript_1_1CompileSpec.rst", "_cpp_api/torch_tensort_cpp.rst", "_cpp_api/unabridged_orphan.rst", "cli/torchtrtc.rst", "contributors/conversion.rst", "contributors/dynamo_converters.rst", "contributors/lowering.rst", "contributors/partitioning.rst", "contributors/phases.rst", "contributors/runtime.rst", "contributors/system_overview.rst", "contributors/ts_converters.rst", "contributors/useful_links.rst", "contributors/writing_dynamo_aten_lowering_passes.rst", "dynamo/dynamo_export.rst", "dynamo/torch_compile.rst", "fx/getting_started_with_fx_path.rst", "getting_started/installation.rst", "getting_started/jetpack.rst", "getting_started/quick_start.rst", "index.rst", "indices/supported_ops.rst", "py_api/dynamo.rst", "py_api/fx.rst", "py_api/logging.rst", "py_api/ptq.rst", "py_api/runtime.rst", "py_api/torch_tensorrt.rst", "py_api/ts.rst", "sg_execution_times.rst", "src/pytorch-sphinx-theme/docs/changelog.rst", "src/pytorch-sphinx-theme/docs/configuring.rst", "src/pytorch-sphinx-theme/docs/demo/api.rst", "src/pytorch-sphinx-theme/docs/demo/demo.rst", "src/pytorch-sphinx-theme/docs/demo/lists_tables.rst", "src/pytorch-sphinx-theme/docs/demo/long.rst", "src/pytorch-sphinx-theme/docs/demo/structure.rst", "src/pytorch-sphinx-theme/docs/index.rst", "src/pytorch-sphinx-theme/docs/installing.rst", "ts/creating_torchscript_module_in_python.rst", "ts/getting_started_with_cpp_api.rst", "ts/getting_started_with_python_api.rst", "ts/ptq.rst", "ts/torchscript_frontend_from_pytorch.rst", "tutorials/_rendered_examples/dynamo/converter_overloading.rst", "tutorials/_rendered_examples/dynamo/cross_runtime_compilation_for_windows.rst", "tutorials/_rendered_examples/dynamo/custom_kernel_plugins.rst", "tutorials/_rendered_examples/dynamo/engine_caching_bert_example.rst", "tutorials/_rendered_examples/dynamo/engine_caching_example.rst", "tutorials/_rendered_examples/dynamo/index.rst", "tutorials/_rendered_examples/dynamo/mutable_torchtrt_module_example.rst", "tutorials/_rendered_examples/dynamo/refit_engine_example.rst", "tutorials/_rendered_examples/dynamo/torch_compile_advanced_usage.rst", "tutorials/_rendered_examples/dynamo/torch_compile_resnet_example.rst", "tutorials/_rendered_examples/dynamo/torch_compile_stable_diffusion.rst", "tutorials/_rendered_examples/dynamo/torch_compile_transformers_example.rst", "tutorials/_rendered_examples/dynamo/torch_export_cudagraphs.rst", "tutorials/_rendered_examples/dynamo/torch_export_gpt2.rst", "tutorials/_rendered_examples/dynamo/torch_export_llama2.rst", "tutorials/_rendered_examples/dynamo/vgg16_ptq.rst", "tutorials/_rendered_examples/dynamo/weight_streaming_example.rst", "tutorials/_rendered_examples/index.rst", "tutorials/notebooks.rst", "tutorials/serving_torch_tensorrt_with_triton.rst", "user_guide/dynamic_shapes.rst", "user_guide/mixed_precision.rst", "user_guide/runtime.rst", "user_guide/saving_models.rst", "user_guide/torch_tensorrt_explained.rst", "user_guide/using_dla.rst"], "titles": ["Class DataType", "Class Device::DeviceType", "Class TensorFormat", "Template Class Int8CacheCalibrator", "Template Class Int8Calibrator", "Define STR", "Define TORCH_TENSORRT_PATCH_VERSION", "Define TORCH_TENSORRT_MAJOR_VERSION", "Define TORCH_TENSORRT_MINOR_VERSION", "Define TORCHTRT_API", "Define XSTR", "Define TORCHTRT_HIDDEN", "Define TORCH_TENSORRT_VERSION", "Directory cpp", "Directory include", "Directory torch_tensorrt", "Enum Level", "Enum EngineCapability", "File logging.h", "File macros.h", "File ptq.h", "File torch_tensorrt.h", "Function torch_tensorrt::logging::get_logging_prefix", "Function torch_tensorrt::logging::get_reportable_log_level", "Function torch_tensorrt::logging::get_is_colored_output_on", "Function torch_tensorrt::logging::set_reportable_log_level", "Function torch_tensorrt::logging::log", "Function torch_tensorrt::logging::set_is_colored_output_on", "Function torch_tensorrt::logging::set_logging_prefix", "Template Function torch_tensorrt::ptq::make_int8_cache_calibrator", "Template Function torch_tensorrt::ptq::make_int8_calibrator", "Function torch_tensorrt::torchscript::check_method_operator_support", "Function torch_tensorrt::torchscript::compile", "Function torch_tensorrt::torchscript::embed_engine_in_new_module", "Function torch_tensorrt::torchscript::convert_method_to_trt_engine", "Function torch_tensorrt::get_build_info", "Function torch_tensorrt::set_device", "Function torch_tensorrt::dump_build_info", "Namespace torch_tensorrt", "Namespace torch_tensorrt::logging", "Namespace torch_tensorrt::ptq", "Namespace torch_tensorrt::torchscript", "Program Listing for File logging.h", "Program Listing for File macros.h", "Program Listing for File ptq.h", "Program Listing for File torch_tensorrt.h", "Struct Device", "Struct GraphInputs", "Struct Input", "Struct CompileSpec", "Torch-TensorRT C++ API", "Full API", "torchtrtc", "Conversion Phase", "Writing Dynamo Converters", "Lowering Phase", "Partitioning Phase", "Compiler Phases", "Runtime Phase", "System Overview", "Writing TorchScript Converters", "Useful Links for Torch-TensorRT Development", "Writing Dynamo ATen Lowering Passes", "Compiling Exported Programs with Torch-TensorRT", "TensorRT Backend for torch.compile", "Torch-TensorRT (FX Frontend) User Guide", "Installation", "Overview", "Quick Start", "Torch-TensorRT", "Operators Supported", "torch_tensorrt.dynamo", "torch_tensorrt.fx", "torch_tensorrt.logging", "torch_tensorrt.ts.ptq", "torch_tensorrt.runtime", "torch_tensorrt", "torch_tensorrt.ts", "Computation times", "Changelog", "Configuration", "5. :mod:`test_py_module`", "3. Paragraph Level Markup", "4. Lists & Tables", "1. Long Sticky Nav", "1. Structural Elements", "<no title>", "Installation", "Creating a TorchScript Module", "Using Torch-TensorRT in C++", "Using Torch-TensorRT in Python", "Post Training Quantization (PTQ)", "Using Torch-TensorRT TorchScript Frontend Directly From PyTorch", "Overloading Torch-TensorRT Converters with Custom Converters", "Cross runtime compilation for windows example", "Using Custom Kernels within TensorRT Engines with Torch-TensorRT", "Engine Caching (BERT)", "Engine Caching", "Dependencies", "Mutable Torch TensorRT Module", "Refitting Torch-TensorRT Programs with New Weights", "Torch Compile Advanced Usage", "Compiling ResNet with dynamic shapes using the torch.compile backend", "Compiling Stable Diffusion model using the torch.compile backend", "Compiling BERT using the torch.compile backend", "Torch Export with Cudagraphs", "Compiling GPT2 using the dynamo backend", "Compiling Llama2 using the dynamo backend", "Deploy Quantized Models using Torch-TensorRT", "Weight Streaming", "Torch-TensorRT Tutorials", "Legacy notebooks", "Serving a Torch-TensorRT model with Triton", "Dynamic shapes with Torch-TensorRT", "Compile Mixed Precision models with Torch-TensorRT", "Deploying Torch-TensorRT Programs", "Saving models compiled with Torch-TensorRT", "Torch-TensorRT Explained", "DLA"], "terms": {"defin": [0, 1, 2, 3, 4, 33, 43, 46, 47, 48, 49, 51, 52, 54, 65, 68, 75, 76, 80, 88, 89, 90, 91, 93, 95, 97, 101, 104, 105, 106, 107, 111], "file": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 46, 47, 48, 49, 52, 54, 56, 58, 59, 64, 65, 66, 67, 68, 71, 72, 74, 76, 77, 78, 80, 81, 83, 87, 89, 91, 94, 112, 113, 116], "torch_tensorrt": [0, 1, 2, 14, 16, 17, 42, 43, 44, 46, 47, 48, 49, 50, 51, 52, 54, 56, 62, 63, 64, 65, 68, 69, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 103, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 118], "h": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 15, 46, 47, 48, 49, 50, 51, 52, 55, 68, 76, 89, 91], "support": [0, 1, 2, 27, 31, 46, 48, 49, 52, 54, 56, 61, 63, 65, 67, 68, 69, 72, 75, 76, 77, 80, 81, 88, 89, 90, 93, 95, 100, 102, 104, 106, 107, 108, 109, 112, 114, 117, 118], "data": [0, 2, 3, 4, 29, 30, 44, 46, 48, 49, 52, 53, 56, 57, 59, 60, 64, 65, 70, 71, 72, 74, 76, 77, 82, 86, 90, 91, 95, 97, 108, 109, 111], "type": [0, 1, 2, 30, 49, 50, 52, 53, 56, 58, 60, 62, 63, 64, 65, 71, 72, 74, 75, 76, 77, 82, 89, 90, 91, 93, 94, 95, 97, 108, 109, 111, 114, 116], "can": [0, 1, 4, 29, 30, 34, 46, 47, 48, 49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 62, 63, 64, 65, 66, 67, 68, 71, 74, 75, 76, 77, 80, 82, 88, 89, 90, 91, 92, 93, 94, 95, 97, 99, 100, 101, 104, 105, 108, 109, 111, 112, 113, 114, 115, 116, 117], "us": [0, 1, 2, 3, 4, 29, 30, 32, 34, 36, 43, 44, 45, 46, 48, 49, 52, 53, 54, 56, 58, 59, 60, 62, 63, 64, 65, 66, 67, 68, 69, 71, 72, 74, 75, 76, 77, 78, 80, 81, 82, 83, 88, 91, 94, 97, 98, 99, 100, 109, 110, 112, 114, 115, 116, 117, 118], "tensorrt": [0, 1, 3, 4, 29, 30, 31, 32, 33, 34, 37, 44, 45, 46, 48, 49, 52, 53, 54, 55, 56, 57, 59, 60, 62, 67, 68, 71, 72, 74, 75, 76, 77, 88, 91, 94, 97, 98, 101, 102, 103, 104, 105, 109], "engin": [0, 1, 17, 32, 33, 34, 45, 46, 48, 49, 52, 53, 56, 57, 59, 62, 63, 64, 69, 71, 72, 75, 76, 77, 80, 89, 90, 91, 92, 93, 98, 100, 102, 104, 109, 110, 113, 115, 117, 118], "thi": [0, 1, 2, 29, 30, 42, 43, 44, 45, 46, 47, 48, 49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 62, 63, 64, 65, 66, 67, 68, 71, 72, 75, 76, 77, 80, 81, 82, 84, 85, 88, 89, 91, 92, 93, 95, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 111, 112, 113, 114, 115, 116, 117], "compat": [0, 1, 46, 55, 58, 64, 65, 71, 75, 76, 77, 117], "c10": [0, 1, 45, 46, 48, 49, 89, 91], "check": [0, 1, 31, 46, 52, 55, 60, 65, 67, 71, 75, 77, 89, 95, 99, 100, 112, 115], "trt": [0, 1, 3, 4, 46, 48, 53, 55, 58, 60, 62, 64, 65, 67, 68, 70, 71, 75, 76, 89, 93, 95, 104, 106, 107, 109, 113, 115, 116], "so": [0, 44, 52, 53, 54, 55, 58, 59, 60, 62, 64, 65, 66, 67, 72, 75, 76, 81, 82, 83, 89, 91, 93, 95, 97, 101, 102, 104, 106, 107, 113], "should": [0, 3, 4, 29, 45, 49, 52, 53, 54, 55, 56, 57, 59, 60, 63, 64, 65, 67, 71, 75, 76, 77, 80, 82, 85, 91, 93, 95, 96, 97, 100, 105, 112], "reason": [0, 65, 88, 93, 95, 97, 117], "you": [0, 1, 2, 29, 30, 46, 48, 49, 52, 53, 54, 55, 56, 58, 59, 60, 63, 65, 66, 67, 68, 71, 75, 76, 77, 80, 82, 83, 84, 88, 89, 90, 91, 92, 93, 95, 97, 98, 99, 100, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117], "need": [0, 1, 2, 25, 29, 43, 46, 53, 54, 55, 60, 65, 66, 67, 71, 72, 75, 76, 82, 89, 90, 91, 93, 95, 96, 97, 99, 100, 111, 112, 113, 115], "explictli": 0, "public": [0, 1, 2, 3, 4, 44, 45, 46, 47, 48, 49, 83, 91], "enum": [0, 1, 2, 42, 45, 46, 71, 77, 91, 93], "valu": [0, 1, 2, 16, 17, 45, 46, 48, 53, 56, 58, 60, 63, 70, 71, 74, 76, 80, 89, 99, 101, 102, 104, 109, 111], "underli": [0, 1, 2, 46, 60], "In": [0, 1, 2, 46, 53, 54, 56, 57, 58, 59, 60, 64, 65, 66, 75, 76, 82, 83, 85, 90, 91, 93, 95, 99, 111, 112, 113, 114, 115, 116], "case": [0, 1, 2, 46, 49, 53, 54, 56, 58, 60, 62, 64, 65, 66, 67, 75, 76, 91, 93, 95, 99, 100, 113, 114, 115], "itself": [0, 1, 2, 46, 52, 55, 92, 93, 112], "interfac": [0, 1, 2, 46, 58, 59, 60, 64, 69, 91], "vs": [0, 1, 2, 46, 55, 66, 71, 76, 77, 92], "normal": [0, 1, 2, 46, 65, 82, 88, 89, 91, 93, 99, 100, 105, 108, 112, 118], "instatin": [0, 1, 2, 46], "ex": [0, 1, 2, 33, 46, 67, 77, 83, 85], "kfloat": [0, 45, 49], "enumer": [0, 1, 2, 16, 17, 46], "klong": [0, 45], "int64": [0, 76, 77, 109], "kdoubl": [0, 45], "fp64": [0, 76], "fp32": [0, 48, 49, 52, 64, 65, 71, 76, 77, 91, 106, 107, 111, 112, 114], "khalf": [0, 45, 89], "fp16": [0, 48, 49, 52, 64, 65, 71, 72, 76, 89, 90, 99, 103, 106, 107, 109, 114, 118], "kchar": [0, 45], "int8": [0, 44, 48, 49, 52, 64, 71, 76, 77, 91, 108, 118], "kint": [0, 45], "int": [0, 3, 4, 36, 44, 45, 49, 52, 54, 56, 63, 64, 70, 71, 72, 76, 77, 80, 89, 95, 108, 109], "kbool": [0, 45], "bool": [0, 1, 2, 3, 4, 24, 27, 30, 31, 42, 44, 45, 46, 49, 55, 60, 64, 70, 71, 72, 74, 75, 76, 77, 80, 89, 91, 94, 95], "kunknown": [0, 2, 45], "sentinel": [0, 2, 76], "function": [0, 1, 2, 3, 4, 46, 48, 49, 54, 55, 56, 58, 60, 62, 64, 65, 66, 88, 89, 91, 92, 93, 95, 100, 101, 104, 105, 106, 107, 111, 112, 113, 115, 117, 118], "default": [0, 1, 2, 3, 4, 16, 29, 30, 33, 43, 45, 46, 48, 49, 52, 54, 56, 62, 64, 65, 66, 71, 72, 75, 76, 77, 80, 81, 82, 89, 90, 91, 92, 93, 94, 95, 97, 108, 113, 115, 116, 117], "construct": [0, 1, 2, 3, 4, 46, 48, 49, 53, 54, 55, 57, 59, 60, 65, 74, 75, 76, 82, 83, 89, 91, 93, 95, 97, 113], "new": [0, 1, 2, 3, 4, 32, 33, 46, 48, 49, 56, 58, 59, 60, 62, 64, 65, 68, 69, 71, 77, 82, 89, 97, 98, 99, 102, 104, 105, 110, 112, 115], "object": [0, 1, 2, 3, 4, 46, 48, 49, 52, 58, 60, 62, 63, 64, 71, 75, 76, 77, 91, 92, 93, 113, 116], "inlin": [0, 1, 2, 3, 4, 29, 30, 44, 46, 48, 55, 83, 86, 89], "constexpr": [0, 1, 2, 45, 46, 95], "t": [0, 1, 2, 45, 46, 55, 60, 65, 66, 70, 76, 80, 82, 83, 88, 89, 91, 93, 95, 108, 112, 113], "constructor": [0, 2, 46, 48, 49, 58, 88], "from": [0, 1, 2, 3, 4, 29, 30, 44, 46, 48, 49, 52, 53, 55, 56, 57, 58, 59, 60, 63, 64, 65, 67, 69, 71, 72, 75, 76, 77, 78, 80, 81, 82, 83, 88, 89, 91, 93, 94, 95, 96, 97, 99, 100, 103, 104, 106, 107, 108, 109, 111, 112, 115, 116, 117], "torchtrt_api": [0, 2, 19, 22, 23, 24, 25, 26, 27, 28, 31, 32, 33, 34, 35, 36, 37, 42, 43, 44, 45, 48, 49, 50], "scalartyp": [0, 45, 70], "torch": [0, 1, 2, 4, 20, 21, 29, 30, 31, 32, 33, 34, 37, 44, 45, 46, 47, 48, 49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 62, 67, 71, 72, 74, 75, 76, 77, 78, 88, 91, 94, 96, 97, 98, 109, 118], "paramet": [0, 1, 2, 3, 4, 25, 26, 27, 29, 30, 31, 32, 33, 34, 36, 46, 48, 49, 53, 54, 55, 60, 64, 65, 71, 72, 74, 75, 76, 77, 86, 88, 89, 100, 106, 107], "oper": [0, 1, 2, 3, 4, 31, 44, 45, 46, 49, 52, 53, 55, 56, 57, 58, 59, 60, 62, 63, 65, 69, 71, 76, 77, 90, 91, 93, 100, 102, 104, 117, 118], "const": [0, 1, 2, 3, 4, 29, 30, 31, 32, 33, 34, 36, 44, 45, 46, 55, 60, 70, 89, 91], "get": [0, 1, 2, 3, 4, 23, 35, 44, 46, 55, 56, 60, 62, 63, 65, 67, 75, 76, 89, 91, 93, 97, 106, 107, 109, 111, 112], "return": [0, 1, 2, 3, 4, 23, 24, 29, 30, 31, 32, 33, 34, 35, 42, 43, 44, 45, 46, 54, 55, 56, 57, 58, 59, 60, 62, 64, 65, 71, 72, 75, 76, 77, 88, 89, 90, 91, 93, 95, 97, 100, 101, 108, 109, 112, 113, 114], "explicit": [0, 1, 2, 3, 4, 45, 46, 55, 65, 72, 75, 82, 91, 117], "delet": [0, 1, 2, 45, 46, 55], "other": [0, 1, 2, 45, 46, 52, 53, 55, 58, 62, 64, 65, 66, 70, 71, 75, 76, 81, 82, 89, 90, 93, 115], "comparis": [0, 2], "true": [0, 1, 2, 4, 46, 49, 55, 56, 60, 62, 64, 65, 70, 71, 72, 75, 76, 77, 80, 83, 89, 91, 92, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 112, 114, 115, 118], "fals": [0, 1, 2, 3, 4, 44, 45, 46, 49, 54, 62, 64, 65, 70, 71, 72, 75, 76, 77, 80, 81, 82, 83, 89, 91, 92, 93, 94, 95, 96, 97, 99, 100, 101, 103, 104, 105, 106, 107, 108, 109, 115], "struct": [1, 21, 38, 41, 45, 54, 91], "onli": [1, 3, 4, 16, 29, 44, 46, 48, 52, 54, 55, 56, 59, 60, 64, 65, 67, 68, 71, 72, 75, 76, 82, 91, 93, 94, 95, 99, 100, 107, 109, 114, 115, 118], "applic": [1, 29, 46, 52, 55, 59, 64, 71, 75, 76, 89, 90, 92, 115, 118], "kcuda": [1, 46, 56, 89], "which": [1, 2, 29, 32, 34, 46, 49, 53, 54, 55, 56, 57, 58, 59, 60, 62, 63, 64, 65, 66, 71, 72, 74, 75, 76, 77, 80, 82, 83, 88, 89, 90, 91, 92, 93, 94, 95, 97, 101, 102, 105, 106, 107, 111, 112, 113, 114, 115, 116, 117], "map": [1, 46, 53, 54, 55, 57, 59, 60, 65, 76, 89, 91, 92, 97, 101, 111, 112], "kgpu": [1, 45, 46], "To": [1, 46, 52, 54, 56, 64, 66, 71, 80, 88, 89, 90, 92, 95, 100, 106, 107, 112], "datatyp": [1, 21, 38, 45, 46, 48, 49, 50, 71, 76, 77, 90, 95, 112, 114], "target": [1, 33, 45, 46, 48, 49, 52, 54, 56, 58, 59, 64, 65, 66, 69, 71, 75, 76, 77, 90, 91, 92, 93, 95, 100, 117, 118], "gpu": [1, 32, 34, 36, 45, 46, 52, 64, 65, 71, 75, 76, 77, 89, 91, 92, 95, 106, 107, 109, 112, 115, 117, 118], "run": [1, 34, 46, 49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 64, 65, 66, 67, 68, 71, 72, 75, 76, 77, 82, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 111, 112, 113, 114, 115, 116, 117, 118], "kdla": [1, 45, 46, 118], "dla": [1, 45, 46, 49, 52, 64, 69, 71, 76, 77], "intern": [1, 16, 46, 60, 63, 73, 75, 82, 89], "note": [1, 46, 48, 54, 60, 62, 65, 66, 67, 75, 76, 80, 82, 89, 95, 100, 113, 118], "The": [1, 46, 48, 49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 62, 63, 64, 65, 66, 71, 75, 76, 77, 80, 83, 88, 90, 91, 92, 93, 94, 95, 96, 97, 99, 100, 102, 105, 106, 109, 111, 112, 113, 116, 117], "valid": [1, 46, 56, 60, 62, 71, 75, 76, 93], "kcpu": [1, 46], "comparison": [1, 46], "an": [2, 3, 4, 48, 49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 62, 64, 65, 66, 68, 71, 72, 74, 75, 76, 77, 80, 82, 83, 88, 89, 90, 91, 93, 95, 97, 100, 101, 105, 106, 107, 109, 111, 112, 113, 115, 116, 117], "memeori": 2, "layout": [2, 48, 70, 71, 76, 77], "store": [2, 4, 49, 52, 53, 58, 60, 64, 65, 71, 75, 76, 77, 88, 89, 95, 97, 100], "tensor": [2, 33, 44, 45, 48, 49, 52, 53, 54, 55, 56, 58, 60, 62, 63, 64, 65, 70, 71, 72, 75, 76, 77, 88, 89, 90, 91, 93, 95, 101, 109, 111], "kcontigu": [2, 45, 48], "contigu": [2, 48, 49, 52, 71, 76, 77], "nchw": [2, 71, 76, 77], "linear": [2, 56, 70, 76, 88, 95, 108, 114], "kchannelslast": [2, 45], "channel": [2, 76, 81], "last": [2, 55, 65, 76, 108], "nhwc": [2, 52], "memoryformat": [2, 45], "ptq": [3, 4, 15, 18, 19, 38, 50, 51, 52, 69, 71, 76, 77], "privat": [3, 4, 44, 45, 91], "algorithm": [3, 4, 29, 30, 44, 65, 74, 91, 107], "typenam": [3, 4, 29, 30, 44], "gener": [3, 4, 29, 52, 55, 58, 59, 60, 62, 64, 65, 66, 71, 72, 80, 82, 83, 86, 88, 89, 91, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 108, 109, 110, 115], "int8calibr": [3, 20, 30, 40, 44, 50], "implement": [3, 4, 55, 56, 58, 63, 65, 75, 81, 89, 91, 95, 97, 115], "specifi": [3, 4, 33, 52, 54, 60, 64, 65, 66, 71, 76, 77, 80, 82, 90, 92, 109, 112, 113, 114, 116, 117], "calibr": [3, 4, 29, 30, 44, 49, 52, 71, 74, 76, 77, 89, 91], "read": [3, 4, 29, 30, 44, 80, 82, 91], "nvinfer1": [3, 4, 29, 30, 44, 45, 49, 60, 91], "iint8calibr": [3, 4, 29, 30, 44, 45, 49, 71, 76, 77, 91], "iint8entropycalibrator2": [3, 4, 29, 30, 44, 91], "std": [3, 4, 22, 26, 28, 29, 30, 31, 33, 34, 35, 42, 44, 45, 47, 48, 49, 56, 89, 91, 112, 118], "string": [3, 4, 18, 20, 21, 22, 26, 28, 29, 30, 31, 33, 34, 35, 42, 44, 45, 49, 54, 56, 58, 60, 64, 71, 76, 80, 89, 91], "cache_file_path": [3, 4, 29, 30, 44], "8": [3, 52, 55, 63, 64, 66, 75, 76, 82, 83, 86, 89, 94, 95, 102, 105, 112, 113], "cach": [3, 4, 29, 30, 44, 52, 64, 65, 69, 71, 72, 74, 76, 89, 91, 98, 110, 115], "getbatchs": [3, 4, 44], "noexceptoverrid": [3, 4], "batch": [3, 4, 44, 64, 65, 72, 75, 91, 97, 102, 104, 108, 109, 112, 113, 118], "size": [3, 4, 44, 48, 49, 52, 55, 56, 64, 65, 70, 71, 72, 76, 77, 80, 89, 91, 95, 97, 102, 104, 108, 111, 113], "next": [3, 4, 53, 54, 58, 63, 72, 76, 80, 82, 83, 91, 93, 101, 105, 108, 112], "alwai": [3, 4, 27, 52, 76, 82, 100, 109], "1": [3, 4, 33, 44, 45, 48, 49, 52, 54, 55, 56, 58, 60, 62, 63, 64, 65, 66, 70, 71, 72, 74, 75, 76, 77, 79, 80, 82, 83, 86, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 99, 100, 102, 104, 105, 106, 107, 108, 109, 111, 113, 114, 116, 118], "due": [3, 4, 66, 81, 82, 108], "issu": [3, 4, 64, 71, 76, 89, 101, 104], "getbatch": [3, 4, 44], "void": [3, 4, 25, 26, 27, 28, 36, 37, 42, 44, 45], "bind": [3, 4, 33, 44, 75, 77, 82], "char": [3, 4, 44, 52, 89], "name": [3, 4, 31, 33, 34, 44, 54, 56, 58, 60, 65, 66, 67, 72, 74, 75, 76, 77, 82, 83, 88, 89, 92, 93, 95, 100, 105, 108, 112, 114], "nbbind": [3, 4, 44], "Not": 3, "arrai": [3, 4, 33, 53, 54, 76, 77, 93, 95, 109], "pointer": [3, 4, 91], "fed": [3, 4, 48], "buffer": [3, 4, 65, 95], "each": [3, 4, 49, 53, 55, 56, 58, 60, 64, 65, 66, 71, 72, 75, 80, 82, 89, 93, 100, 107, 115], "input": [3, 4, 21, 29, 33, 38, 44, 45, 47, 49, 50, 52, 53, 54, 55, 56, 58, 60, 62, 63, 64, 65, 68, 70, 71, 72, 73, 75, 76, 77, 83, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 99, 100, 101, 105, 106, 107, 108, 109, 111, 112, 113, 114, 115, 116, 117, 118], "number": [3, 4, 49, 52, 54, 55, 56, 60, 63, 64, 65, 71, 72, 76, 77, 80, 89, 90, 95, 100, 102, 104, 109, 111, 117], "readcalibrationcach": [3, 4, 44], "size_t": [3, 4, 44, 91], "length": [3, 4, 44, 65, 70, 83, 109], "how": [3, 4, 66, 67, 82, 84, 86, 88, 92, 93, 95, 97, 99, 101, 108, 109, 111, 112, 113, 115], "enabl": [3, 4, 24, 49, 52, 54, 56, 57, 59, 64, 65, 66, 71, 72, 74, 75, 76, 77, 80, 97, 99, 100, 102, 104, 105, 106, 107, 109, 114, 115], "use_cach": [3, 4, 30, 44, 74, 91, 106, 107, 109], "set": [3, 4, 16, 21, 25, 27, 29, 32, 34, 36, 45, 46, 48, 49, 52, 53, 54, 55, 56, 57, 58, 59, 65, 66, 71, 72, 75, 76, 77, 80, 84, 87, 88, 89, 90, 91, 93, 95, 100, 106, 108, 109, 111, 113, 114, 115, 117, 118], "writecalibrationcach": [3, 4, 44], "write": [3, 4, 29, 30, 44, 65, 69, 82, 89, 91, 112], "provid": [3, 4, 49, 52, 54, 56, 58, 60, 62, 64, 65, 66, 68, 71, 72, 75, 76, 77, 82, 89, 90, 91, 92, 93, 97, 98, 100, 101, 105, 109, 110, 112, 113, 115, 116, 117], "cast": [3, 4, 55, 64, 71, 106, 107, 114], "convienc": [3, 4, 49], "convert": [3, 4, 31, 32, 34, 52, 55, 56, 57, 59, 63, 64, 69, 71, 76, 77, 90, 92, 95, 98, 102, 104, 109, 110, 111, 115], "easili": [3, 4, 99], "assign": [3, 4, 81], "ptq_calibr": [3, 4, 45, 49, 91], "field": [3, 4, 63, 72, 76, 91], "compilespec": [3, 4, 21, 32, 34, 41, 45, 50, 56, 77, 89, 91, 118], "dataloaderuniqueptr": [4, 44], "libtorch": [4, 37, 60, 66, 68, 89, 91, 117], "dataload": [4, 29, 30, 44, 49, 74, 91, 108], "unique_ptr": [4, 30], "unqiue_ptr": 4, "A": [4, 29, 30, 32, 33, 47, 48, 54, 55, 56, 60, 65, 66, 71, 72, 76, 77, 83, 91, 103, 112], "uniqu": [4, 90], "what": [4, 54, 55, 65, 68, 76, 82, 88, 89, 90, 106, 107, 117], "make_data_load": [4, 91], "factori": [4, 29, 30, 64, 71, 91], "path": [4, 13, 14, 15, 29, 30, 52, 64, 65, 66, 67, 71, 74, 76, 88, 89, 91, 94, 97, 105, 108, 112, 117], "find": [4, 65, 66, 67, 89, 95, 109], "whether": [4, 52, 54, 64, 65, 71, 72, 76, 81, 91, 102, 104, 115], "exist": [4, 31, 32, 34, 54, 63, 64, 65, 67, 71, 74, 76, 77, 91, 97, 111], "There": [4, 53, 54, 59, 60, 62, 63, 65, 66, 83, 88, 91, 100, 111, 112, 113, 115], "consum": [4, 53, 88], "macro": [5, 6, 7, 8, 9, 10, 11, 12, 15, 18, 20, 21, 42, 44, 45, 50, 51], "x": [5, 10, 33, 43, 55, 56, 66, 67, 68, 75, 77, 83, 88, 89, 93, 95, 97, 101, 105, 108, 109, 113, 114, 116], "includ": [13, 15, 16, 35, 37, 42, 43, 44, 45, 51, 52, 54, 56, 57, 58, 59, 62, 64, 65, 66, 67, 68, 71, 72, 75, 76, 80, 82, 88, 89, 91, 95, 115], "parent": [14, 15, 18, 19, 20, 21], "cpp": [14, 15, 42, 43, 44, 45, 51, 55, 59, 66, 89, 91], "log": [15, 16, 19, 20, 38, 44, 50, 51, 55, 60, 64, 65, 69, 70, 71, 72, 76, 93, 102, 104, 114], "emum": [16, 17], "messag": [16, 25, 26, 52, 73], "sever": [16, 26, 73], "kinternal_error": [16, 42], "print": [16, 31, 44, 62, 64, 67, 71, 77, 82, 89, 92, 93, 94, 95, 96, 97, 99, 100, 102, 104, 106, 107, 108, 109, 112], "error": [16, 49, 52, 53, 55, 59, 64, 65, 71, 73, 76, 77, 82, 89, 113], "kerror": [16, 42], "all": [16, 42, 43, 44, 45, 49, 52, 54, 55, 56, 58, 62, 64, 65, 66, 67, 71, 73, 75, 76, 78, 82, 83, 88, 89, 90, 91, 93, 95, 106, 107, 110, 111, 112, 114, 115, 117], "kwarn": [16, 42], "warn": [16, 44, 52, 60, 73, 75], "kinfo": [16, 42, 44], "info": [16, 32, 34, 45, 52, 60, 73, 75, 76, 114], "kdebug": [16, 42, 44], "debug": [16, 27, 45, 49, 52, 60, 62, 64, 71, 73, 75, 76, 77, 92, 94, 95, 96, 97, 99, 100, 101, 102, 104, 108, 114], "kgraph": [16, 42, 55], "everyth": [16, 64, 71, 76], "intermedi": [16, 49, 52, 54, 64, 71, 73, 76, 77, 88, 114, 117], "graph": [16, 31, 32, 34, 45, 49, 52, 53, 54, 56, 57, 59, 60, 62, 63, 64, 65, 71, 72, 73, 76, 77, 88, 89, 93, 95, 97, 99, 100, 102, 104, 105, 111, 113, 115], "lower": [16, 54, 63, 65, 69, 71, 72, 73, 76, 83, 95, 97, 102, 104, 109, 111, 117], "phase": [16, 60, 63, 89, 93, 100, 113, 117], "select": [17, 29, 30, 34, 49, 52, 58, 64, 65, 66, 70, 71, 76, 77, 81, 84, 90, 91, 95, 117], "capabl": [17, 45, 49, 52, 58, 71, 76, 77, 92, 93, 94], "kstandard": [17, 45, 49], "ksafeti": [17, 45], "kdla_standalon": [17, 45], "directori": [18, 19, 20, 21, 42, 43, 44, 45, 50, 66, 67, 71, 91, 97], "program": [18, 19, 20, 21, 29, 51, 52, 57, 58, 59, 69, 71, 88, 97, 98, 106, 107, 110, 113], "list": [18, 19, 20, 21, 31, 49, 51, 53, 56, 58, 60, 62, 63, 65, 68, 70, 71, 72, 75, 76, 77, 86, 89, 90, 93, 95, 112], "torchscript": [19, 21, 38, 43, 45, 49, 50, 52, 56, 57, 58, 59, 63, 68, 71, 72, 74, 75, 76, 77, 90, 111, 113, 118], "str": [19, 43, 44, 50, 54, 64, 65, 70, 71, 74, 75, 76, 77, 93, 94, 95, 97, 108], "torch_tensorrt_major_vers": [19, 43, 50], "torch_tensorrt_minor_vers": [19, 43, 50], "torch_tensorrt_patch_vers": [19, 43, 50], "torch_tensorrt_vers": [19, 43, 50], "torchtrt_hidden": [19, 43, 50], "xstr": [19, 43, 50], "nvinfer": [20, 44], "fstream": [20, 44], "iostream": [20, 21, 44, 45, 89], "iter": [20, 44, 49, 52, 53, 64, 71, 74, 76, 77, 96, 97, 108, 109], "memori": [20, 21, 44, 45, 55, 60, 71, 76, 77, 89, 90, 95, 97, 106, 107, 109], "sstream": [20, 44], "vector": [20, 21, 33, 44, 45, 47, 48, 49, 56, 58, 76, 89, 91, 118], "templat": [20, 40, 44, 45, 50, 80, 89], "int8cachecalibr": [20, 29, 40, 44, 50], "cuda_runtim": [21, 45], "custom_class": [21, 45], "devic": [21, 33, 36, 38, 45, 49, 50, 52, 58, 64, 70, 71, 72, 74, 75, 76, 77, 90, 91, 92, 95, 99, 103, 106, 107, 109, 111, 118], "graphinput": [21, 38, 45, 49, 50], "devicetyp": [21, 38, 45, 46, 50, 75, 76, 77, 91, 92, 95, 118], "tensorformat": [21, 38, 45, 48, 50, 76, 95], "level": [23, 25, 26, 39, 42, 44, 50, 54, 55, 56, 59, 64, 65, 71, 76, 77, 86, 88, 93, 95, 112, 117], "current": [23, 54, 56, 58, 60, 62, 63, 64, 65, 66, 67, 71, 72, 75, 76, 77, 80, 93, 95, 99, 106, 107, 108, 109, 115], "report": [23, 44, 75], "Is": [24, 76], "color": [24, 27, 82], "output": [24, 27, 33, 49, 52, 53, 54, 55, 56, 58, 60, 62, 63, 64, 65, 66, 71, 73, 75, 76, 77, 80, 82, 83, 89, 93, 95, 97, 99, 100, 103, 109, 111, 112, 113, 114, 116], "lvl": [25, 26, 42], "inform": [25, 33, 35, 37, 48, 52, 53, 56, 58, 62, 64, 65, 66, 71, 72, 73, 76, 82, 88, 89, 91, 92, 95, 97, 109, 113], "ad": [25, 52, 53, 54, 56, 62, 65, 66, 95, 99], "abov": [25, 54, 56, 62, 65, 66, 73, 81, 82, 89, 95, 102, 104, 114, 116], "msg": [26, 42], "add": [26, 53, 54, 55, 56, 60, 63, 66, 70, 80, 82, 87, 89, 90, 93, 95], "global": [26, 52, 64, 71, 76, 89], "colored_output_on": [27, 42], "prefix": [27, 28, 42, 82], "help": [27, 52, 53, 60, 64, 65, 89, 94, 97, 108, 109, 111, 115], "when": [27, 44, 45, 46, 52, 53, 55, 56, 57, 58, 59, 60, 64, 65, 66, 71, 75, 76, 77, 80, 82, 84, 88, 89, 91, 93, 95, 97, 99, 100, 109, 111, 113, 115], "termin": [27, 52, 89], "If": [27, 33, 53, 54, 55, 56, 62, 63, 64, 65, 66, 68, 71, 72, 76, 80, 82, 89, 90, 91, 93, 95, 97, 100, 101, 105, 109, 112, 113, 114, 115, 117, 118], "build": [29, 30, 35, 49, 52, 53, 57, 59, 60, 63, 64, 65, 71, 75, 76, 81, 86, 89, 91, 93, 95, 102, 104, 109, 113], "post": [29, 30, 49, 52, 63, 69, 89, 97], "train": [29, 30, 49, 52, 69, 70, 89, 90, 97, 109], "quantiz": [29, 30, 52, 64, 69, 74, 76, 89, 98, 110], "creat": [29, 30, 33, 52, 53, 54, 56, 58, 60, 65, 69, 76, 77, 82, 89, 93, 95, 100, 109, 112], "previous": [29, 33, 89, 97, 100], "therefor": [29, 58, 65, 66, 75, 82, 89, 111, 115], "have": [29, 33, 44, 52, 53, 54, 55, 56, 60, 62, 63, 64, 65, 66, 67, 71, 72, 74, 75, 76, 77, 82, 88, 89, 90, 91, 95, 98, 102, 104, 108, 110, 111, 112, 113], "requir": [29, 49, 52, 53, 54, 55, 63, 64, 65, 66, 67, 71, 76, 77, 80, 89, 91, 93, 94, 95, 98, 108, 109, 110, 112, 113, 115], "dataset": [29, 74, 91, 111], "save": [29, 44, 52, 58, 64, 65, 68, 69, 71, 75, 76, 77, 89, 90, 94, 96, 97, 100, 103, 109, 111, 112, 115, 117], "later": [29, 71, 89, 100, 116, 117], "differ": [29, 55, 56, 59, 64, 65, 66, 76, 80, 88, 93, 95, 97, 99, 106, 109, 111, 115, 117], "scratch": [29, 97, 100], "depend": [29, 35, 53, 59, 64, 65, 67, 68, 71, 89, 90, 109, 112, 115], "howev": [29, 66, 80, 81, 89, 93, 95, 97, 112, 113, 117], "network": [29, 30, 54, 60, 65, 76, 89, 91, 93, 95, 109, 111, 112, 118], "also": [29, 53, 54, 60, 62, 64, 66, 68, 80, 82, 83, 89, 90, 91, 97, 105, 108, 111], "recalibr": 29, "its": [29, 53, 56, 58, 60, 66, 75, 76, 82, 95, 108, 112, 115, 117], "structur": [29, 46, 49, 56, 59, 60, 64, 71, 76, 80, 82, 86, 88, 95, 112], "chang": [29, 55, 56, 59, 62, 64, 65, 75, 76, 77, 80, 91, 93, 97, 99, 100, 112, 115, 117], "respons": [29, 54, 58, 82, 115], "ensur": [29, 54, 55, 56, 62, 64, 66, 67, 71, 75, 106, 107], "By": [29, 30, 51, 56, 64, 66, 71, 80, 88, 97, 113], "entropi": [29, 30, 91], "v2": [29, 30, 82], "perform": [29, 30, 54, 62, 63, 71, 75, 76, 91, 95, 105, 109, 111, 112, 114, 115, 116, 117], "recommend": [29, 30, 65, 66, 76, 82, 89, 95, 112, 113], "feed": [29, 30, 89], "forward": [29, 30, 32, 33, 56, 58, 60, 64, 68, 71, 75, 76, 77, 88, 89, 90, 91, 92, 93, 95, 101, 108, 113, 114], "overrid": [29, 30, 44, 54, 65, 91], "minmax": [29, 30, 91], "recomend": [29, 30], "nlp": [29, 30, 91], "task": [29, 30, 65, 91, 111], "call": [29, 30, 32, 49, 54, 55, 58, 60, 65, 71, 72, 75, 76, 77, 82, 88, 89, 92, 93, 95, 97, 99, 101, 104, 111, 113, 115, 117], "make_int8_calibr": [29, 40, 44, 50, 91], "class": [29, 30, 44, 45, 46, 51, 58, 60, 64, 65, 73, 77, 82, 83, 88, 89, 90, 91, 93, 95, 97, 101, 108, 111, 113, 114], "e": [29, 30, 52, 55, 60, 65, 66, 67, 68, 72, 76, 88, 89, 91, 95, 97, 100], "g": [29, 30, 52, 55, 65, 66, 67, 72, 76, 82, 91, 95, 100], "iint8minmaxcalibr": [29, 30, 91], "calibration_cache_fil": [29, 30, 91], "move": [30, 44, 55, 58, 77, 89, 91, 93, 106, 107], "calibration_dataload": [30, 91], "contain": [30, 31, 52, 53, 54, 55, 56, 60, 65, 66, 72, 75, 76, 82, 83, 88, 89, 91, 95, 97, 112, 115], "jit": [31, 32, 33, 34, 45, 47, 49, 52, 53, 55, 56, 57, 58, 59, 60, 61, 64, 68, 69, 71, 75, 76, 77, 88, 89, 90, 92, 95, 100, 112, 116, 117], "modul": [31, 32, 33, 34, 45, 49, 52, 56, 57, 58, 59, 60, 64, 65, 66, 67, 68, 69, 71, 72, 74, 75, 76, 77, 81, 82, 83, 90, 91, 92, 93, 94, 95, 98, 100, 101, 108, 110, 111, 113, 114, 116, 118], "method_nam": [31, 34, 45, 52, 76, 77, 89], "see": [31, 55, 56, 58, 62, 64, 65, 66, 76, 77, 82, 88, 89, 90, 93, 95, 97, 100, 101], "fulli": [31, 52, 55, 64, 71, 75, 76, 77, 89, 91, 95, 118], "compil": [31, 34, 41, 45, 49, 50, 52, 54, 55, 56, 58, 60, 62, 65, 71, 72, 73, 75, 76, 77, 78, 80, 88, 90, 91, 92, 93, 95, 96, 98, 99, 108, 110, 112, 115, 118], "take": [31, 32, 33, 34, 53, 54, 57, 58, 59, 60, 62, 65, 71, 72, 75, 76, 77, 80, 82, 89, 91, 92, 93, 95, 101, 111, 113], "method": [31, 32, 33, 34, 48, 52, 55, 60, 66, 71, 76, 77, 82, 88, 89, 92, 97, 111], "pure": [31, 71, 76], "Will": 31, "out": [31, 44, 53, 55, 56, 57, 59, 60, 64, 66, 71, 76, 77, 82, 89, 95, 99, 108, 109, 112, 113], "unsupport": [31, 49, 54, 64, 76, 95, 117], "script": [31, 55, 56, 68, 76, 77, 88, 89, 90, 92, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 115, 117], "nvidia": [32, 34, 42, 43, 44, 45, 52, 61, 64, 65, 66, 67, 71, 76, 77, 89, 101, 104, 112, 117, 118], "configur": [32, 34, 48, 62, 64, 66, 71, 75, 76, 77, 86, 89, 91, 95, 109, 112, 113], "equival": [32, 57, 59, 60, 71, 76, 77, 88, 89, 91, 93, 95, 102, 104], "specif": [32, 49, 54, 55, 57, 59, 62, 64, 71, 76, 77, 82, 93, 109, 117], "traget": 32, "input_binding_nam": [33, 45, 75, 77], "output_binding_nam": [33, 45, 75, 77], "emb": [33, 52, 63, 77, 83], "pre": [33, 55, 74, 77, 91, 97, 109, 115], "built": [33, 52, 58, 59, 64, 66, 71, 75, 76, 77, 97, 100], "serial": [33, 34, 52, 57, 59, 66, 71, 75, 76, 77, 89, 95, 97, 117], "regist": [33, 54, 58, 60, 65, 75, 77, 93, 95], "execut": [33, 49, 52, 55, 57, 58, 59, 63, 64, 65, 66, 69, 71, 72, 75, 76, 77, 78, 88, 89, 91, 93, 95, 112], "must": [33, 48, 49, 52, 54, 55, 56, 60, 62, 65, 66, 71, 72, 76, 77, 82, 83, 89, 97, 113, 115, 117], "follow": [33, 52, 54, 56, 58, 62, 63, 64, 65, 66, 77, 80, 82, 83, 87, 88, 89, 91, 93, 95, 97, 98, 102, 106, 107, 110, 111, 112, 113, 114, 115], "format": [33, 45, 48, 49, 52, 70, 71, 76, 77, 82, 83, 90, 95, 97, 108, 111, 112, 114, 116], "symbol": [33, 65, 66, 77, 82, 115], "index": [33, 61, 62, 66, 67, 69, 70, 77, 80, 86, 91, 95], "0": [33, 43, 44, 45, 49, 52, 54, 56, 59, 60, 62, 64, 65, 66, 67, 69, 70, 71, 72, 74, 75, 76, 77, 78, 79, 81, 82, 89, 91, 92, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 112, 113, 114, 118], "2": [33, 43, 54, 56, 60, 63, 64, 65, 66, 67, 69, 70, 71, 74, 75, 76, 77, 80, 82, 83, 86, 88, 89, 91, 93, 95, 96, 97, 99, 100, 101, 102, 104, 106, 107, 108, 109, 113, 116], "y": [33, 56, 77, 83, 93, 95, 101], "compilesepc": 33, "order": [33, 49, 54, 56, 60, 62, 65, 66, 71, 72, 75, 76, 77, 89, 90, 93, 97, 114], "pass": [33, 53, 54, 56, 57, 58, 59, 60, 63, 64, 65, 66, 69, 73, 74, 75, 76, 77, 88, 89, 91, 93, 95, 97, 100], "origin": [33, 65, 72, 76, 95, 97, 99, 117], "pytorch": [33, 48, 49, 52, 54, 55, 56, 57, 58, 59, 60, 63, 64, 66, 67, 68, 71, 74, 75, 76, 77, 88, 89, 90, 91, 93, 97, 99, 100, 108, 112, 113, 114, 115, 116, 117], "assum": [33, 75, 92, 95, 98, 110], "convent": 33, "below": [33, 56, 60, 62, 63, 64, 65, 66, 67, 82, 89, 90, 97, 103, 112], "equivil": 34, "librari": [35, 42, 43, 44, 45, 52, 54, 57, 58, 59, 60, 76, 89, 95, 98, 110], "version": [35, 37, 59, 62, 64, 65, 67, 71, 75, 76, 80, 83, 95, 111, 112, 116], "gpu_id": [36, 45, 46, 52, 75, 76, 77, 91, 92, 95, 118], "id": [36, 45, 52, 76, 80, 81, 85, 118], "cudasetdevic": 36, "dump": [37, 52, 95], "base": [37, 50, 58, 63, 64, 66, 71, 72, 76, 82, 88, 90, 91, 96, 100, 104, 111, 117], "stdout": [37, 75], "enginecap": [38, 45, 49, 50, 64, 71, 75, 76, 77, 92, 95], "dump_build_info": [38, 45, 50], "get_build_info": [38, 45, 50], "set_devic": [38, 45, 50, 115], "get_is_colored_output_on": [39, 42, 50], "get_logging_prefix": [39, 42, 50], "get_reportable_log_level": [39, 42, 50], "set_is_colored_output_on": [39, 42, 50], "set_logging_prefix": [39, 42, 50], "set_reportable_log_level": [39, 42, 50], "make_int8_cache_calibr": [40, 44, 50, 91], "check_method_operator_support": [41, 45, 50], "convert_method_to_trt_engin": [41, 45, 50, 76, 77, 89, 92], "embed_engine_in_new_modul": [41, 45, 50, 77], "document": [42, 43, 44, 45, 50, 59, 80, 82, 83, 87, 88, 89, 91, 92, 112, 113, 115], "copyright": [42, 43, 44, 45, 83, 89], "c": [42, 43, 44, 45, 52, 59, 64, 67, 70, 71, 72, 75, 76, 83, 90, 95, 99, 112, 115, 118], "corpor": [42, 43, 44, 45], "right": [42, 43, 44, 45, 55, 59, 60, 82], "reserv": [42, 43, 44, 45, 106, 107], "licens": [42, 43, 44, 45, 89], "under": [42, 43, 44, 45, 59, 65, 82, 93, 102, 117], "bsd": [42, 43, 44, 45], "style": [42, 43, 44, 45, 64, 68, 80, 82, 83], "found": [42, 43, 44, 45, 63, 66, 75, 82, 89, 91, 93, 95, 97, 115], "root": [42, 43, 44, 45, 66, 80, 91, 108], "sourc": [42, 43, 44, 45, 54, 59, 64, 65, 67, 71, 72, 73, 74, 75, 76, 77, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110], "tree": [42, 43, 44, 45, 80, 91, 108, 115], "pragma": [42, 43, 44, 45, 91], "onc": [42, 43, 44, 45, 53, 55, 56, 58, 64, 65, 66, 67, 76, 91, 95, 107, 109, 112, 115], "namespac": [42, 43, 44, 45, 51, 55, 69, 76, 91, 95], "ar": [42, 46, 49, 52, 53, 54, 55, 56, 58, 59, 60, 62, 63, 64, 65, 66, 71, 74, 75, 76, 77, 80, 82, 83, 84, 88, 89, 91, 92, 93, 95, 96, 97, 99, 100, 102, 106, 107, 109, 111, 112, 113, 114, 115, 116, 117], "ones": [42, 56, 57, 59, 66, 82, 89, 93, 95, 117], "necessari": [42, 62, 64, 66, 75, 93, 100, 115], "user": [42, 48, 54, 56, 57, 58, 59, 62, 63, 64, 66, 67, 71, 82, 83, 89, 90, 91, 93, 97, 100, 109, 112, 113, 114, 115, 117], "dont": 42, "know": [42, 60, 80, 82, 93, 95], "we": [42, 44, 53, 54, 55, 56, 57, 58, 59, 60, 62, 63, 64, 65, 72, 75, 80, 82, 88, 89, 91, 93, 95, 97, 98, 99, 100, 101, 102, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 116, 117], "want": [42, 56, 65, 66, 67, 68, 72, 88, 89, 91, 92, 93, 95, 100, 101, 112], "use_cmake_generated_export_head": 43, "torch_tensorrt_export": 43, "els": [43, 44, 48, 77, 82, 83, 94, 96, 97, 108], "__gnuc__": 43, "__attribute__": 43, "__visibility__": 43, "hidden": [43, 80], "endif": [43, 44, 45], "doe": [43, 44, 55, 56, 60, 62, 65, 66, 76, 82, 91, 95, 102, 104], "gaurd": 43, "someth": [43, 55, 82, 112], "6": [43, 55, 56, 58, 66, 70, 82, 86, 88, 89, 94, 95], "setup": [43, 67, 91, 112], "alias": 43, "eas": 43, "ts": [43, 52, 56, 68, 69, 76, 88, 89, 90, 92, 113, 116], "torchtrt": [43, 56, 94, 95, 108], "ifndef": [44, 45], "doxygen_should_skip_thi": [44, 45], "get_batch_impl": 44, "element_typ": 44, "super": [44, 88, 93, 95, 101, 108, 113, 114], "batchtyp": 44, "dataloader_": 44, "cache_file_path_": 44, "use_cache_": 44, "auto": [44, 56, 60, 64, 68, 71, 82, 83, 89, 91, 106, 107, 109, 118], "batched_data_": 44, "push_back": [44, 56], "it_": 44, "begin": [44, 65, 66, 82, 101, 105], "noexcept": [44, 91], "hack": 44, "explict": 44, "work": [44, 55, 59, 60, 64, 65, 68, 71, 74, 75, 76, 82, 83, 91, 95, 100, 101, 105, 109, 113], "here": [44, 53, 54, 56, 58, 63, 64, 65, 66, 68, 80, 82, 83, 88, 89, 91, 93, 95, 98, 105, 106, 107, 108, 110, 112, 113, 115, 116], "explic": 44, "just": [44, 45, 55, 56, 64, 65, 69, 73, 75, 82, 84, 88, 89, 90, 92, 95, 97, 99, 111, 115], "still": [44, 56, 65, 66, 91, 93, 101, 117], "static_cast": 44, "option": [44, 48, 52, 56, 57, 59, 62, 63, 64, 65, 71, 75, 76, 77, 82, 86, 91, 93, 95, 96, 97, 101, 103, 114, 115, 116, 118], "batch_siz": [44, 91, 108], "end": [44, 52, 60, 62, 70, 71, 76, 77, 82, 89, 91, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "statu": [44, 83], "reset": [44, 96, 97, 101, 104, 115], "incas": 44, "go": [44, 55, 56, 65, 68, 88, 89, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 111, 112, 117], "again": [44, 58, 60, 82, 95, 99], "stringstream": 44, "ss": 44, "cache_": 44, "clear": 44, "ifstream": 44, "io": [44, 67, 112], "binari": [44, 91], "noskipw": 44, "good": [44, 60, 65, 82, 97], "copi": [44, 60, 65, 67, 70, 74, 83, 109, 112], "istream_iter": 44, "back_insert": 44, "nullptr": [44, 45, 49], "ofstream": [44, 89], "cache_fil": [44, 74, 91], "reinterpret_cast": 44, "cache_size_": 44, "int8_t": 45, "arrayref": [45, 48, 49], "friend": 45, "ostream": 45, "os": [45, 67, 97], "dtype": [45, 48, 49, 52, 63, 64, 65, 70, 71, 72, 75, 76, 77, 90, 95, 96, 102, 104, 105, 109, 111, 113, 114], "device_typ": [45, 46, 76, 91, 92, 118], "int64_t": [45, 46, 48, 49, 91, 118], "core": [45, 52, 55, 56, 59, 64, 71, 76, 89, 93, 117, 118], "agx": 45, "platform": [45, 52, 59, 64, 66, 67, 71, 94, 112, 118], "xavier": [45, 118], "dla_cor": [45, 46, 52, 76, 91, 92, 118], "allow_gpu_fallback": [45, 46, 71, 76, 77, 91, 92, 118], "customclasshold": [45, 48], "min_shap": [45, 48, 63, 65, 71, 76, 77, 90, 102, 105, 111, 113], "opt_shap": [45, 48, 63, 71, 76, 77, 90, 102, 105, 111, 113], "max_shap": [45, 48, 63, 65, 71, 76, 77, 90, 102, 105, 111, 113], "shape": [45, 47, 48, 49, 52, 56, 60, 63, 65, 69, 70, 71, 72, 75, 76, 77, 78, 90, 93, 95, 98, 105, 108, 109, 110, 112, 115, 118], "doubl": [45, 48, 49, 52, 63, 71, 76, 77, 82, 115], "tensor_domain": [45, 48, 76], "input_is_dynam": 45, "ivalu": [45, 47, 49, 53, 58, 60, 89], "input_signatur": [45, 47, 49, 77, 90], "nest": [45, 49, 50, 82, 83], "full": [45, 49, 52, 60, 64, 71, 73, 76, 89, 91, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 112, 115, 118], "spec": [45, 48, 49, 52, 73, 76, 77, 92, 97], "flatten": [45, 47, 70, 88, 89, 108], "fixed_s": [45, 49], "reflect": [45, 76], "builderconfig": 45, "graph_input": [45, 49], "enabled_precis": [45, 49, 63, 64, 71, 75, 76, 77, 89, 90, 91, 92, 95, 96, 97, 99, 100, 101, 102, 103, 104, 106, 107, 108, 109, 112, 114, 118], "disable_tf32": [45, 49, 64, 71, 75, 76, 77, 91, 95, 106, 107], "sparse_weight": [45, 49, 64, 65, 71, 75, 76, 77, 95], "refit": [45, 49, 64, 69, 71, 76, 77, 92, 95, 97, 98, 99, 110], "truncate_long_and_doubl": [45, 49, 63, 64, 77, 103], "allow_shape_tensor": [45, 49, 77], "uint64_t": [45, 49], "num_avg_timing_it": [45, 49, 64, 71, 75, 76, 77, 92, 95], "workspace_s": [45, 49, 52, 64, 71, 75, 76, 77, 95, 100, 102, 104], "dla_sram_s": [45, 49, 52, 64, 71, 75, 76, 77, 95], "1048576": [45, 49, 64, 71, 75, 76, 77, 95], "dla_local_dram_s": [45, 49, 52, 64, 71, 75, 76, 77, 95], "1073741824": [45, 49, 64, 71, 75, 76, 77, 95], "dla_global_dram_s": [45, 49, 52, 64, 71, 75, 76, 77, 95], "536870912": [45, 49, 64, 71, 75, 76, 77, 95], "require_full_compil": [45, 49, 64, 71, 75, 76, 77, 95], "min_block_s": [45, 49, 56, 63, 64, 71, 75, 76, 77, 93, 94, 95, 96, 97, 100, 101, 102, 104, 108], "3": [45, 49, 52, 55, 56, 58, 63, 64, 65, 67, 68, 70, 71, 74, 76, 77, 82, 83, 86, 88, 89, 91, 92, 94, 95, 96, 97, 99, 100, 102, 105, 106, 107, 108, 109, 111, 113, 116, 118], "torch_executed_op": [45, 49, 56, 63, 64, 71, 75, 76, 77, 95, 100, 101, 102, 104], "torch_executed_modul": [45, 49, 56, 71, 76, 77], "member": [46, 47, 48, 49], "hold": [46, 47, 48, 53, 60, 76, 91], "relat": [46, 82, 101, 104], "let": [46, 52, 55, 60, 65, 71, 76, 77, 80, 82, 111, 112, 117], "layer": [46, 49, 52, 53, 55, 60, 62, 64, 65, 71, 75, 76, 77, 89, 91, 93, 95, 106, 107, 108, 111, 112, 113, 114, 117, 118], "thei": [46, 52, 53, 54, 55, 58, 60, 64, 65, 71, 74, 75, 76, 80, 82, 90, 93, 97], "complex": [47, 49, 64, 66, 88, 90, 99, 107], "either": [47, 48, 52, 60, 62, 71, 76, 77, 80, 82, 88, 89, 90, 93, 94, 95, 97, 116], "one": [47, 54, 55, 60, 64, 65, 67, 71, 75, 76, 82, 88, 89, 90, 93, 95, 101, 104, 106, 107, 112], "rang": [48, 49, 52, 65, 76, 95, 96, 97, 102, 109, 111, 113], "optim": [48, 52, 63, 64, 65, 69, 71, 72, 74, 76, 88, 89, 90, 100, 102, 103, 104, 109, 111, 113, 117], "profil": [48, 72, 75, 114], "singl": [48, 52, 55, 56, 65, 76, 82, 88, 89, 91, 109, 115], "repres": [48, 49, 54, 60, 65, 68, 82], "signifi": [48, 55], "static": [48, 49, 53, 60, 63, 64, 71, 76, 77, 80, 89, 108, 113], "three": [48, 57, 59, 65, 72, 76, 82, 83, 111, 112], "min": [48, 52, 60, 70, 76, 97, 102, 113], "optimin": 48, "max": [48, 52, 60, 70, 76, 80, 97, 102, 108, 113], "allow": [48, 49, 52, 53, 54, 55, 56, 62, 64, 65, 66, 71, 76, 77, 80, 93, 95, 97, 100, 102, 104, 109, 115], "argument": [48, 52, 54, 55, 58, 60, 62, 64, 65, 71, 75, 76, 77, 82, 83, 89, 90, 93, 94, 95, 113], "expect": [48, 54, 55, 60, 76, 89, 90, 111], "tradit": [48, 71, 76, 77, 91], "convect": 48, "produc": [48, 53, 54, 58, 60, 63, 76, 82, 89, 111], "low": [48, 65, 93, 99], "high": [48, 55, 56, 80, 93, 95, 117], "weight": [48, 49, 52, 53, 64, 65, 69, 70, 71, 76, 77, 82, 89, 97, 98, 99, 103, 110, 111], "first": [48, 53, 54, 55, 65, 68, 82, 83, 89, 90, 91, 93, 95, 97, 99, 101, 112, 113, 116, 117], "calcul": [48, 53, 56, 89, 95, 109], "detect": [48, 58, 76], "float32": [48, 49, 52, 63, 64, 65, 71, 76, 77, 95, 99, 103, 106, 107, 109, 113, 114], "dynam": [48, 49, 63, 65, 69, 71, 72, 76, 77, 78, 93, 97, 98, 101, 103, 104, 107, 109, 110, 115], "opt": [48, 66, 75, 76, 105], "minimum": [48, 49, 52, 56, 63, 64, 71, 76, 77, 95, 109], "maximum": [48, 49, 52, 64, 65, 71, 72, 76, 77, 102, 104, 109, 112], "accept": [48, 52, 54, 58, 60, 66, 76, 89, 90, 101, 116], "exampl": [48, 56, 58, 59, 60, 65, 66, 71, 73, 75, 76, 77, 78, 80, 81, 83, 86, 88, 89, 90, 91, 93, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 115, 116], "s": [48, 49, 53, 56, 58, 60, 63, 65, 66, 67, 69, 71, 72, 75, 76, 80, 82, 83, 88, 89, 91, 93, 95, 97, 109, 111, 112, 113, 115, 116], "cannot": [48, 55, 56, 65, 66, 71, 75, 76, 77, 81, 88, 94, 95], "through": [48, 53, 54, 55, 56, 58, 64, 65, 71, 73, 74, 82, 89, 90, 95, 99, 100, 111, 117], "altern": [48, 56, 62, 63, 76, 90, 93, 105, 111, 116], "refer": [48, 54, 57, 59, 65, 81, 86, 89, 91, 95, 108, 112, 113, 116], "given": [48, 49, 52, 54, 55, 65, 71, 72, 74, 76, 77, 88, 89, 90, 92, 93, 113], "kernel": [48, 49, 52, 60, 64, 65, 69, 71, 76, 77, 93, 98, 110, 114, 115], "ani": [48, 52, 53, 54, 60, 62, 64, 65, 70, 71, 74, 75, 76, 77, 80, 82, 89, 90, 91, 93, 95, 102, 113], "event": [48, 64, 96, 97], "place": [48, 55, 62, 65, 82, 83, 84, 91, 95, 108], "variabl": [48, 65, 75, 76], "dimens": [48, 55, 65, 72, 76, 102, 111, 113, 114], "domain": [48, 76, 83, 91], "convien": 49, "fix": [49, 65, 82, 95, 115, 118], "describ": [49, 56, 60, 76, 88, 92, 112], "entri": [49, 60, 97], "okai": 49, "ha": [49, 53, 54, 55, 56, 57, 59, 60, 62, 64, 65, 66, 67, 71, 72, 76, 82, 83, 88, 89, 91, 93, 94, 97, 100, 108, 111, 113, 117], "flaten": 49, "precis": [49, 52, 63, 64, 65, 69, 71, 76, 89, 90, 91, 102, 104, 106, 107, 109, 118], "dure": [49, 52, 54, 56, 60, 63, 64, 71, 74, 76, 91, 93, 106, 107, 109, 111, 113, 115], "prevent": [49, 52, 54, 56], "tf32": [49, 52, 64, 71], "comput": [49, 64, 65, 66, 67, 71, 75, 82, 91, 94, 98, 110, 111], "inner": [49, 83, 111], "product": [49, 67, 76], "round": [49, 71, 76, 77, 95], "10": [49, 66, 67, 71, 72, 76, 77, 86, 88, 89, 91, 108, 109, 111, 112, 113, 114], "bit": [49, 60, 65, 66, 71, 76, 77, 89], "mantissa": [49, 71, 76, 77], "befor": [49, 54, 55, 56, 59, 60, 65, 71, 76, 77, 89, 112, 113], "multipli": [49, 71, 76, 77], "accumul": [49, 64, 71, 76, 77, 106, 107], "sum": [49, 65, 70, 71, 76, 77, 95, 108], "23": [49, 55, 71, 76, 77, 83], "behavior": [49, 56, 65, 71, 76, 77, 93, 106, 107, 113, 115, 116], "sparsiti": [49, 52, 65, 71, 76, 77], "conv": [49, 52, 89, 95], "fc": [49, 52, 55], "truncat": [49, 52, 63, 64, 71, 76, 77], "long": [49, 52, 53, 63, 76, 82, 83], "float": [49, 52, 63, 64, 70, 76, 88, 89, 90, 91, 92, 95, 96, 97, 100, 101, 104, 105, 114], "ishap": 49, "restrict": [49, 64, 71, 76, 77, 113], "cuda": [49, 58, 63, 65, 67, 68, 71, 72, 75, 76, 89, 90, 91, 92, 93, 94, 95, 96, 97, 99, 100, 102, 103, 105, 106, 107, 108, 109, 112, 113, 114, 115, 116], "safeti": [49, 52, 76], "averag": [49, 52, 64, 71, 76, 77, 95], "time": [49, 52, 53, 54, 55, 56, 57, 58, 59, 60, 64, 65, 66, 68, 69, 71, 72, 75, 76, 77, 80, 82, 89, 91, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "workspac": [49, 52, 64, 65, 66, 71, 72, 76, 77, 95, 101, 102, 104], "fast": [49, 52, 64, 68, 71, 76, 77], "softwar": [49, 52, 64, 71, 76, 77, 82], "manag": [49, 52, 53, 55, 57, 59, 60, 64, 66, 67, 71, 73, 75, 76, 77, 89, 105, 115], "ram": [49, 52, 64, 71, 76, 77], "commun": [49, 52, 64, 71, 76, 77, 89], "within": [49, 52, 57, 59, 64, 69, 71, 75, 76, 77, 80, 82, 98, 106, 107, 110, 111], "host": [49, 52, 64, 66, 71, 76, 77, 95, 109, 112], "share": [49, 52, 64, 66, 71, 75, 76, 77, 97], "across": [49, 52, 55, 56, 64, 71, 76, 77, 80], "metadata": [49, 52, 54, 58, 60, 64, 71, 76, 77, 80, 100, 113, 114], "quantizatiom": 49, "instead": [49, 52, 53, 54, 55, 66, 71, 75, 76, 89, 93, 100, 108, 115], "potenti": [49, 71, 76, 85], "subgraph": [49, 52, 53, 54, 55, 60, 62, 89, 95, 97, 117], "aten": [49, 54, 55, 56, 60, 61, 64, 69, 70, 71, 76, 77, 89, 93, 101, 117], "thrown": [49, 71, 76, 77], "empti": [49, 71, 72, 76, 77, 83, 88, 95], "torch_tensorrtnamespac": 50, "loggingenum": 50, "levelnamespac": 50, "ptqtemplat": 50, "int8cachecalibratortempl": 50, "int8calibratornamespac": 50, "torchscriptstruct": 50, "compilespecstruct": 50, "deviceclass": 50, "devicetypestruct": 50, "graphinputsstruct": 50, "inputclass": 50, "datatypeclass": 50, "tensorformatenum": 50, "cppdirectori": 50, "includedirectori": 50, "torch_tensorrtfil": 50, "hfile": 50, "relationship": 50, "inherit": [50, 65, 71, 91], "subdirectori": 51, "definit": [51, 54, 60, 82], "cli": [52, 90], "It": [52, 54, 55, 56, 57, 59, 60, 65, 66, 69, 76, 80, 82, 94, 95, 109, 111, 115, 117], "serv": [52, 58, 65, 69, 71, 76], "easi": [52, 53, 55, 89, 91], "wai": [52, 64, 65, 66, 88, 89, 91, 93, 95, 97, 100, 111, 115, 116], "command": [52, 64, 66, 82, 83, 88, 89, 112], "line": [52, 66, 83, 89, 99], "quickli": [52, 89, 91], "part": [52, 56, 59, 65, 75, 80, 81, 82, 95, 97], "deploy": [52, 75, 89, 90, 91, 111, 112, 115, 118], "pipelin": [52, 89, 99, 103, 118], "basic": [52, 56, 65, 83, 112], "featur": [52, 56, 65, 66, 89, 91, 92, 103, 108, 109, 111, 117], "though": [52, 59, 60, 88, 89, 117], "alreadi": [52, 53, 54, 55, 89, 91, 93, 95, 98, 110, 113], "two": [52, 55, 60, 62, 64, 65, 66, 76, 82, 83, 87, 88, 90, 91, 93, 97, 112, 113], "embed": [52, 54, 58, 70, 77, 82, 118], "plan": [52, 59, 63, 64, 71], "after": [52, 53, 55, 56, 62, 65, 71, 75, 76, 88, 89, 90, 101, 104, 112, 115], "link": [52, 53, 62, 69, 80, 81, 86, 89, 95, 115], "against": [52, 89, 93], "libtorchtrt": [52, 66, 89], "python": [52, 56, 59, 62, 64, 65, 67, 71, 72, 75, 76, 77, 82, 83, 89, 92, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 115, 118], "import": [52, 55, 56, 63, 64, 65, 66, 67, 68, 75, 80, 82, 88, 89, 90, 92, 93, 95, 96, 97, 99, 112, 113, 115, 116], "packag": [52, 55, 64, 67, 89], "aspect": 52, "ident": [52, 62, 71, 76, 100], "standard": [52, 58, 66, 69, 71, 75, 76, 77, 82, 92, 93, 95, 99, 111, 115], "load": [52, 56, 58, 64, 65, 68, 71, 74, 75, 76, 77, 89, 90, 91, 92, 94, 95, 96, 97, 99, 100, 109, 111, 112, 115, 117], "like": [52, 53, 55, 58, 60, 65, 66, 68, 76, 81, 82, 88, 89, 90, 91, 93, 95, 97, 99, 100, 109, 112, 115], "would": [52, 54, 60, 64, 65, 66, 67, 75, 89, 90, 92, 93, 95, 112, 115], "input_file_path": [52, 118], "output_file_path": [52, 118], "input_spec": [52, 65, 72], "displai": [52, 62, 64, 73, 80, 115], "menu": [52, 80, 82], "verbios": 52, "v": [52, 67, 83, 108, 112], "verbos": [52, 64, 65, 71, 72, 83, 102, 104], "about": [52, 53, 58, 60, 66, 75, 80, 89, 112, 113], "process": [52, 56, 64, 76, 81, 82, 88, 91, 92, 100, 101, 105, 111, 112, 115], "onto": [52, 58], "consol": 52, "w": [52, 66, 76], "disabl": [52, 64, 66, 71, 75, 80, 81, 93, 97, 109, 115], "i": [52, 55, 60, 66, 68, 70, 82, 83, 88, 89, 91, 95, 96, 97, 106, 108], "debugg": [52, 71, 76, 77], "fallback": [52, 57, 59, 60, 100, 118], "model": [52, 56, 58, 63, 68, 71, 72, 73, 74, 76, 78, 88, 89, 90, 91, 92, 96, 97, 99, 113, 115, 117], "throw": [52, 55, 76, 89], "spars": [52, 54, 64, 70, 71], "p": [52, 70, 89, 112, 118], "repeat": [52, 70], "f32": [52, 71, 75, 76, 95], "half": [52, 64, 76, 82, 89, 90, 91, 92, 95, 101, 102, 106, 107, 109, 112, 114, 118], "float16": [52, 76, 95, 99, 103, 114], "f16": [52, 76, 89, 118], "i8": [52, 76], "d": [52, 67, 76, 82, 83, 89, 118], "multi": [52, 75], "dlacor": 52, "avail": [52, 54, 60, 62, 64, 65, 66, 67, 71, 75, 76, 80, 95, 109, 111, 117, 118], "dla_standalon": [52, 76], "file_path": [52, 76, 94, 116], "teo": 52, "op_nam": 52, "op": [52, 53, 54, 55, 56, 57, 59, 60, 62, 63, 64, 75, 76, 89, 93, 101, 115, 117], "partial": [52, 82], "tem": 52, "module_nam": 52, "mod": [52, 56, 65, 71, 86, 89, 91, 114], "mb": [52, 78], "num_op": 52, "block": [52, 53, 55, 56, 64, 71, 86, 117], "treat": 52, "num": 52, "avg": 52, "num_it": 52, "sram": 52, "local": [52, 55, 66, 67, 80, 89], "dram": 52, "atol": 52, "absolut": [52, 66], "toler": 52, "threshold": 52, "numer": [52, 65, 83], "deviat": 52, "1e": [52, 99, 100], "rtol": 52, "rel": [52, 56], "5": [52, 56, 58, 59, 64, 65, 66, 67, 71, 75, 76, 82, 83, 86, 88, 89, 93, 95, 99, 101, 109, 112], "skip": 52, "complianc": 52, "64bit": [52, 94], "32bit": 52, "custom": [52, 62, 63, 65, 66, 69, 98, 106, 107, 110], "dll": 52, "n": [52, 60, 62, 76, 89, 91, 93, 95, 96], "min_n": 52, "min_c": 52, "min_h": 52, "min_w": 52, "opt_n": 52, "opt_c": 52, "opt_h": 52, "opt_w": 52, "max_n": 52, "max_c": 52, "max_h": 52, "max_w": 52, "32": [52, 76, 88, 89, 90, 91, 106, 107, 108, 118], "flag": [52, 56, 57, 59, 64, 66, 71, 74, 76, 90, 105, 106, 107, 115, 116], "forc": [52, 63, 65, 71, 76, 77, 80], "posit": [52, 54, 65, 76, 80], "test": [52, 56, 59, 65, 66, 67, 71, 76, 82, 83, 91, 108, 111, 112], "ssd_trace": 52, "pt": [52, 65, 89, 106, 107, 112], "ssd_trt": 52, "300": [52, 92], "512": [52, 71, 76, 77, 108, 111], "1024": [52, 71, 76, 77, 106, 111], "simplifi": [53, 95], "form": [53, 75, 76, 82, 90, 112], "up": [53, 55, 56, 57, 58, 59, 62, 65, 66, 71, 76, 82, 88, 93, 95, 97, 100, 101, 104, 109, 111], "context": [53, 57, 58, 59, 64, 73, 75, 93, 105, 115], "inetworkdefinit": [53, 54], "record": [53, 88, 96, 97, 105, 115], "togeth": [53, 60, 89], "start": [53, 56, 65, 70, 74, 76, 83, 89, 92, 95, 96, 97, 111], "look": [53, 54, 55, 68, 71, 76, 88, 91, 92, 93, 97, 112, 113], "assembl": [53, 62, 89], "resourc": [53, 91, 95], "coupl": [53, 59, 65, 115], "state": [53, 54, 60, 62, 75, 89, 93, 99], "been": [53, 60, 64, 66, 67, 74, 83, 89, 94, 97, 100, 117], "evaluated_value_map": [53, 60], "stage": [53, 65], "arg": [53, 54, 62, 65, 71, 74, 75, 76, 86, 89, 93, 94, 95, 97, 108, 111], "itensor": [53, 54, 60, 65, 89, 93, 95], "value_tensor_map": [53, 60], "typic": [53, 60, 76, 112], "abl": [53, 55, 60, 62, 65, 91, 92, 95, 100], "system": [53, 60, 62, 64, 69, 71, 75, 76, 77, 93, 94, 95, 97, 100, 117], "registri": [53, 54, 89, 95], "enter": [53, 76], "recurs": 53, "resolv": [53, 55, 57, 59, 101, 104], "until": [53, 56, 59, 60, 66, 71, 76, 117], "final": [53, 56, 57, 59, 66, 93, 95, 101, 104, 111], "some": [53, 54, 55, 56, 57, 58, 59, 60, 62, 63, 65, 66, 76, 81, 82, 89, 91, 93, 95, 97, 113, 117], "These": [53, 54, 56, 58, 62, 64, 66, 71, 74, 75, 76, 80, 82, 91, 93, 112, 117], "those": [53, 54, 62, 64, 82], "do": [53, 54, 55, 56, 60, 63, 65, 81, 83, 88, 89, 90, 91, 93, 95, 106, 107, 118], "theori": [53, 82], "kind": [53, 65], "common": [53, 55, 65, 72, 82, 93, 97], "prim": [53, 55, 56, 58, 70, 88, 89], "constant": [53, 54, 55, 56, 89, 95], "emit": 53, "listconstruct": [53, 56, 58, 89], "make": [53, 54, 65, 66, 67, 71, 76, 82, 84, 89, 90, 91, 95, 97, 111, 112, 118], "associ": [53, 60, 89, 97, 115], "where": [53, 54, 55, 60, 62, 64, 65, 71, 75, 76, 77, 83, 89, 91, 93, 100], "result": [53, 55, 56, 66, 68, 71, 73, 75, 76, 77, 80, 88, 90, 94, 95, 99, 100, 109, 112, 114, 117], "done": [53, 56, 59, 95, 100, 112, 116], "mai": [53, 54, 56, 58, 59, 65, 66, 71, 75, 76, 77, 82, 83, 88, 89, 90, 91, 93, 95, 100, 101, 104, 109, 112, 115], "For": [53, 56, 62, 63, 64, 65, 66, 68, 72, 76, 80, 82, 83, 88, 89, 91, 92, 93, 95, 99, 101, 108, 111, 112, 115, 116], "more": [53, 64, 65, 66, 67, 69, 71, 76, 80, 83, 88, 89, 90, 91, 92, 95, 97, 99, 102, 104, 112, 115], "writing_convert": [53, 89], "locat": [54, 62, 66, 91, 93, 95], "py": [54, 55, 59, 62, 65, 66, 67, 78, 80, 82, 87, 88, 89, 91, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 113], "convers": [54, 55, 56, 58, 63, 64, 65, 71, 76, 77, 89, 93, 95, 111, 113], "decror": 54, "dynamo_tensorrt_convert": [54, 93, 95], "signatur": [54, 77], "leaky_relu": [54, 70], "def": [54, 62, 65, 82, 88, 90, 93, 95, 96, 97, 101, 108, 109, 112, 113, 114], "leaky_relu_convert": 54, "ctx": [54, 60, 89, 93, 95, 109], "conversionctx": [54, 60, 89, 93], "tupl": [54, 58, 63, 65, 71, 72, 75, 76, 77, 90, 93, 95, 97, 100, 113, 114], "kwarg": [54, 65, 71, 74, 75, 76, 93, 95, 111], "dict": [54, 71, 75, 76, 77, 93, 95, 97], "union": [54, 60, 64, 71, 75, 76, 77, 89, 93], "sequenc": [54, 62, 65, 71, 72, 75, 76, 77, 82, 93, 95, 109, 111], "decor": [54, 62, 65, 93], "kei": [54, 82, 88, 97, 112, 113], "node": [54, 55, 56, 57, 59, 60, 62, 64, 65, 71, 72, 89, 93, 95, 108, 111, 113], "capability_valid": [54, 93], "lambda": [54, 60, 82, 89, 93, 112], "fx": [54, 62, 63, 71, 75, 76, 89, 90, 93, 95, 100, 116], "determin": [54, 55, 64, 65, 76, 93, 109, 113, 115], "properli": [54, 66], "handl": [54, 55, 56, 58, 64, 65, 75, 76, 95], "partition": [54, 71, 76, 95], "sure": [54, 66, 67, 89, 90, 112, 118], "prioriti": [54, 93], "develop": [54, 65, 66, 67, 69, 82, 83, 89, 93, 95], "bodi": [54, 82, 83], "nativ": [54, 59, 61, 89, 93, 95, 100], "numpi": [54, 76, 95, 96, 97, 99, 100, 109, 112], "frozen": 54, "attribut": [54, 55, 56, 58, 65, 76, 82, 89], "previou": [54, 80, 101], "correspond": [54, 60, 65, 66, 75, 76, 93, 97, 99, 108, 115], "edg": [54, 82], "well": [54, 63, 66, 69, 73, 75, 82, 89, 91, 93, 97, 105, 116], "being": [54, 65, 66, 71, 89, 93, 95, 100], "truth": 54, "http": [54, 61, 64, 66, 67, 80, 82, 88, 89, 91, 93, 95, 99, 101, 104, 108, 111, 112, 113, 115], "github": [54, 61, 64, 66, 67, 80, 89, 91, 101, 104, 108, 112, 115], "com": [54, 61, 64, 66, 67, 89, 91, 99, 101, 104, 108, 112, 115], "blob": [54, 61, 66, 80, 91, 97], "main": [54, 55, 56, 57, 58, 59, 60, 63, 65, 66, 80, 82, 84, 89, 93, 95, 106, 108], "src": [54, 58, 61, 70], "native_funct": [54, 61], "yaml": [54, 61], "sinc": [54, 55, 64, 65, 67, 75, 82, 88, 89, 91, 93, 96, 97, 100], "mani": [54, 56, 64, 65, 80, 82, 83, 93, 97, 100, 117], "composit": [54, 89], "raw": [54, 80, 93], "impl": [54, 93], "subpackag": 54, "chain": [54, 60], "primarili": [54, 59, 66, 89, 93], "manipul": [54, 62, 76], "net": [54, 60, 82, 83, 89, 95], "addit": [54, 55, 64, 65, 75, 76, 89, 93, 95, 97, 100, 111, 113], "call_modul": 54, "call_funct": [54, 62, 65], "eg": [54, 112, 114], "aten_": 54, "_leaky_relu": 54, "opoverloadpacket": 54, "while": [54, 56, 66, 75, 91, 93, 99, 109, 111, 112, 115, 117], "opoverload": 54, "particular": [54, 64, 97], "collect": [54, 56, 64, 71, 76, 77, 89, 90, 108], "trtinterpret": [54, 65, 72], "along": [54, 76], "match": [54, 55, 93, 100], "special": [54, 56], "account": [54, 112], "illustr": [54, 65, 102, 106, 107, 111], "scale_grad_by_freq": [54, 70], "embedding_param_valid": 54, "establish": 54, "subset": [54, 64, 71, 76, 91, 111], "converter_util": [54, 95], "enforce_tensor_typ": 54, "dictionari": [54, 76, 77, 92, 101], "between": [54, 55, 56, 60, 66, 76, 82, 83, 91, 97, 99, 106, 109], "possibl": [54, 66, 82, 93, 95, 97, 111, 112], "prefer": [54, 64, 66, 89], "keyword": [54, 62, 71, 75, 76, 77, 93, 101, 104], "both": [54, 56, 64, 66, 69, 71, 72, 75, 76, 80, 82, 88, 91, 93, 95, 97], "enforc": [54, 89], "situat": 54, "partit": [54, 55, 63, 64, 71, 76, 93, 117], "greater": [54, 71, 73, 76], "than": [54, 55, 64, 66, 71, 76, 81, 82, 93, 96, 97, 99, 109, 111, 115], "3d": [54, 65], "autocast": 54, "therebi": [54, 58, 95, 111], "limit": [54, 55, 73, 81, 91, 94, 97, 98, 109, 110, 117], "author": [54, 83], "conv_nod": 54, "7": [54, 56, 58, 59, 75, 76, 86, 89, 95, 101, 102, 104, 108, 113], "ignor": [54, 71, 75, 76, 95], "misc": [54, 95], "trttensor": 54, "np": [54, 93, 95, 96, 97, 99, 100, 109, 112], "ndarrai": [54, 95], "aten_ops_convolut": 54, "conversioncontext": [54, 93, 95], "side": [54, 55, 80, 89, 93], "effect": [54, 55, 64, 65, 71, 80, 89, 91, 93, 95, 111], "term": [54, 76, 82, 83, 91, 93, 95, 111], "getitem": 54, "categor": 54, "modif": [54, 62, 76], "op_evalu": 54, "capbility_valid": 54, "opcod": 54, "decompos": 54, "suboper": 54, "separ": [54, 56, 57, 59, 66], "Such": 54, "via": [54, 64, 65, 67, 69, 71, 75, 76, 77, 80, 86, 90, 91, 101, 102, 104, 106, 107, 111, 113, 115, 116, 117], "register_torch_trt_decomposit": 54, "addmm_replac": 54, "replac": [54, 56, 62, 66, 67, 74, 95, 108, 117], "input_": 54, "mat1": 54, "mat2": [54, 70], "beta": [54, 65, 70, 77], "alpha": [54, 65, 70, 83], "mul": [54, 56, 70, 93], "matmul": [54, 55, 64, 70, 71, 89, 106, 107, 113], "modifi": [54, 56, 62, 65, 83, 99, 113], "edit": [54, 66, 80], "torch_enabled_decomposit": 54, "torch_disabled_decomposit": 54, "disjoint": 54, "preced": [54, 82], "over": [54, 57, 59, 65, 82, 108, 109, 112, 117], "much": [54, 60, 80, 82, 91], "significantli": [54, 55, 80, 97], "easier": [54, 57, 59, 60, 65, 71, 75, 76, 89, 91, 95, 99], "tri": 54, "made": [55, 57, 59, 76, 82], "represent": [55, 60, 65, 88, 111, 117], "instanc": [55, 62, 64, 66, 71, 74, 75, 88, 89, 93, 111, 115], "idea": [55, 82, 93], "reduc": [55, 56, 57, 59, 65, 71, 76, 91, 95, 97, 111, 115], "actual": [55, 58, 60, 65, 88, 89, 95], "aim": [55, 117], "closer": 55, "scope": [55, 95, 101, 104], "csrc": [55, 61], "common_subexpression_elimin": 55, "subexpress": 55, "dead_code_elimin": 55, "exception_elimin": 55, "wa": [55, 58, 62, 64, 65, 71, 75, 76, 82, 89, 93, 94, 117], "1013": 55, "ne": [55, 70], "1012": 55, "24": [55, 67], "lib": [55, 66, 67, 89], "python3": [55, 66, 89], "site": [55, 66, 82, 89], "nn": [55, 61, 65, 71, 72, 75, 76, 77, 88, 89, 90, 93, 95, 101, 108, 113, 114, 117], "batchnorm": 55, "248": 55, "11": [55, 66, 82, 86, 89, 112], "block0": 55, "raiseexcept": 55, "249": 55, "12": [55, 56, 67, 82, 86, 88, 89, 102, 112, 113], "block1": 55, "guard_elimin": 55, "whose": [55, 65, 102], "freeze_modul": 55, "propag": 55, "fuse_addmm_branch": 55, "variant": [55, 115], "caught": 55, "ret": 55, "622": 55, "self": [55, 58, 60, 70, 75, 76, 88, 89, 90, 93, 95, 97, 101, 108, 111, 113, 114, 118], "bia": [55, 70, 89, 108], "x9": 55, "3677": 55, "output0": [55, 114], "add_": [55, 70, 89, 93], "fuse_linear": 55, "back": [55, 56, 58, 59, 75, 76, 82, 88, 89, 95, 117], "fuse_flatten_linear": 55, "implicitli": [55, 76], "connect": [55, 71, 76, 77, 82, 99, 112, 118], "higher": [55, 64, 71, 76, 80, 82, 88, 109], "1d": 55, "lower_graph": 55, "access": [55, 60, 65, 80, 89, 92, 117], "rather": 55, "getattr": [55, 58, 88, 89], "trainabl": 55, "remain": [55, 76, 91, 117], "lower_tupl": 55, "lowersimpletupl": 55, "tupleconstruct": [55, 58], "tupleunpack": 55, "leav": [55, 62, 64, 71], "statement": [55, 82, 93], "loweralltupl": 55, "_all_": 55, "rais": [55, 65, 76, 94], "onnx": 55, "module_fallback": 55, "consist": [55, 65, 82, 95, 115, 117], "pair": [55, 60, 66, 82, 91, 111], "delimit": 55, "around": [55, 58, 60, 64, 66, 71, 75, 82, 85, 88, 95], "second": [55, 65, 82, 90, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "mark": [55, 56, 80, 97], "notatemoduleforfallback": 55, "marknodesforfallback": 55, "tell": [55, 56, 57, 58, 59, 60, 82, 117], "them": [55, 56, 58, 63, 64, 65, 66, 71, 75, 80, 89, 95, 97, 107, 111, 113, 117], "peephole_optimz": 55, "intent": [55, 82], "catch": [55, 76, 89], "small": [55, 95, 96, 112], "might": [55, 66, 80, 100, 113], "interest": [55, 82], "now": [55, 56, 59, 60, 65, 66, 76, 82, 89, 92, 93, 95, 97, 100, 109, 114, 115], "expand": [55, 70], "simpli": [55, 101, 111], "remove_contigu": 55, "remove_dropout": 55, "infer": [55, 64, 65, 71, 76, 77, 89, 91, 94, 100, 101, 109, 111, 113, 115, 116, 117], "remove_to": 55, "unpack_addmm": 55, "reus": [55, 65, 91, 97], "dedic": [55, 83], "unpack_log_softmax": 55, "softmax": [55, 65, 70, 108], "loop_unrol": 55, "suffici": [55, 66, 76], "short": [55, 64, 71, 82, 83, 100], "tile_to_repeat": 55, "instruct": [56, 57, 59, 65, 66, 89, 112], "criteria": [56, 57, 59, 64], "lack": [56, 57, 59, 65, 95, 109], "explicitli": [56, 57, 59, 66, 77, 90, 91, 92, 106, 107, 114], "On": 56, "segment": [56, 63, 95, 102, 104, 111], "verifi": [56, 71, 93, 95, 100], "Then": [56, 91, 92, 100], "roughli": 56, "analysi": 56, "everi": [56, 72, 75, 76, 89, 115], "complet": [56, 63, 71, 76, 88, 89], "mean": [56, 60, 64, 65, 70, 71, 72, 101, 109, 112, 117], "trace": [56, 65, 71, 75, 77, 88, 89, 113, 116, 117], "tensorlist": [56, 60], "figur": [56, 83, 85], "our": [56, 59, 63, 88, 89, 112], "stitch": [56, 89], "altogeth": [56, 80], "brief": 56, "descript": [56, 83, 94, 108], "partitioninfo": 56, "api": [56, 59, 60, 62, 63, 64, 65, 75, 76, 77, 81, 89, 90, 91, 92, 95, 101, 102, 105, 109, 111, 112, 113, 115, 116], "maintain": [56, 58, 60, 76, 99, 117], "code": [56, 59, 62, 64, 65, 66, 81, 83, 88, 89, 91, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 113], "mymodel": [56, 63, 68, 90, 95, 113, 116], "ts_model": [56, 89], "trt_model": [56, 92, 95, 102, 106, 107, 108, 109, 112, 116], "off": [56, 58, 105], "consecut": [56, 63], "satisfi": [56, 62, 65], "forced_fallback_op": 56, "randn": [56, 63, 68, 71, 76, 77, 89, 92, 93, 97, 102, 105, 113, 114, 116], "224": [56, 63, 68, 71, 72, 76, 77, 89, 94, 97, 99, 100, 102, 105, 111, 112, 113, 116], "trt_ts_modul": [56, 90], "input_s": 56, "inputrang": 56, "cfg": [56, 89], "relu": [56, 70, 88, 89, 101, 108], "trt_mod": [56, 68, 89, 91, 118], "consid": [56, 77, 89, 95, 114], "segmentmodelwithdependencyawar": 56, "test_segment": 56, "20": [56, 67, 86, 100, 102, 104], "x_lgamma": 56, "lgamma": 56, "y_lgamma": 56, "div": [56, 70], "div_lgamma": 56, "27": [56, 89], "cat": [56, 66, 67, 70, 108, 109], "greedi": [56, 106, 107, 109], "strategi": [56, 76], "travers": [56, 57, 59, 64], "gather": 56, "same": [56, 58, 62, 64, 65, 66, 71, 76, 80, 82, 88, 89, 92, 94, 95, 97, 100, 102, 104, 112, 113, 115, 116], "encount": [56, 64, 66, 93, 101, 104], "4": [56, 58, 63, 64, 65, 66, 70, 76, 78, 80, 82, 83, 86, 89, 95, 101, 103, 104, 105, 108, 113], "suboptim": 56, "arithmet": 56, "split": [56, 65, 70], "own": [56, 60, 64, 66, 71, 82, 89, 97, 108, 112], "could": [56, 64, 65, 95, 102, 104, 115], "rewrit": [56, 62], "portion": [56, 82, 95, 103], "without": [56, 60, 68, 71, 80, 82, 89, 91, 95, 96, 97, 100, 115], "reorder": 56, "seri": 56, "cleanli": 56, "approach": [56, 97], "achiev": [56, 111], "hit": 56, "larger": [56, 71, 76, 80, 109, 111], "boundari": [56, 74, 76], "guarante": [56, 75], "trigger": [56, 64, 65, 76, 89, 97, 99, 100, 117], "appear": [56, 82], "adjac": [56, 71, 76, 82], "As": [56, 65, 66, 76, 89, 93, 95, 97, 100, 117], "clean": [56, 62, 82, 101, 104], "step": [56, 65, 67, 70, 76, 91, 95, 100, 111], "consolid": [56, 88], "further": [56, 64, 65, 115, 117], "merg": 56, "identifi": 56, "do_not_merg": 56, "combin": [56, 64, 65], "condit": [56, 82, 117], "loop": [56, 64, 65, 106, 107], "ir": [57, 59, 60, 63, 64, 68, 71, 76, 88, 89, 90, 98, 101, 102, 104, 105, 110, 113], "larg": [57, 59, 80, 82, 89, 91, 100, 109, 111], "opset": [57, 59, 93], "compon": [57, 59, 66, 67, 74, 88, 115, 117], "evalu": [57, 58, 59, 108], "deploi": [57, 59, 69, 89, 91, 98, 110, 112], "instanti": [57, 58, 59, 60, 89, 103], "wrap": [57, 58, 59, 65, 82, 85, 89, 92, 101, 104], "extend": [57, 59, 60, 70, 89, 97, 111], "providi": [57, 59], "stand": [58, 82], "interpret": [58, 65, 82], "execute_engin": [58, 75, 89], "stack": [58, 70, 91, 108, 117], "machin": [58, 66, 91, 94, 112], "pop": 58, "push": 58, "element": [58, 65, 82, 83, 86], "realiz": 58, "abstract": [58, 60, 83, 93], "__torch__": [58, 88, 89], "portabl": [58, 66, 77], "serializ": [58, 64, 88, 117], "instnanti": 58, "whatev": [58, 65, 95], "self_1": [58, 89], "torchvis": [58, 91, 92, 94, 97, 99, 100, 102, 105, 108, 112], "resnet": [58, 69, 78, 94, 98, 99, 110, 111, 112], "___torch_mangle_4847": 58, "resnet_trt": 58, "input_0": [58, 89], "__torch___torchvision_models_resnet____torch_mangle_4847_resnet_trt_engin": 58, "listunpack": [58, 89], "multipl": [58, 66, 71, 75, 76, 82, 83, 91, 109, 112, 115], "repack": 58, "ssd": 58, "ssd300_trt": 58, "__torch___pytorch_detection_ssd_src_model_ssd300_trt_engin": 58, "holder": [58, 84], "torchbind": 58, "pickler": 58, "seril": 58, "zip": [58, 66, 99, 100, 110], "depickl": 58, "encod": [58, 111], "sm": 58, "correct": [58, 66, 80, 98, 99, 100, 108, 110], "bazel": [59, 66, 67], "linux": [59, 64, 67, 71, 89, 94], "x86_64": [59, 66], "aarch64": 59, "gcc": [59, 89], "untest": 59, "try": [59, 76, 82, 83, 89, 92, 95, 97, 117], "older": 59, "repositori": [59, 66, 80, 87, 112], "notebook": [59, 69, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110], "doc": [59, 61, 66, 67, 80, 81, 82, 87, 93, 95, 113], "docsrc": 59, "third_parti": [59, 66], "toolchain": [59, 66, 67], "unstabl": 59, "subject": [59, 62, 117], "matur": 59, "most": [59, 65, 66, 72, 95, 100, 112, 115, 117], "hood": [59, 102, 117], "major": [59, 65, 76], "top": [59, 80, 84], "coordin": [59, 76], "ingest": 59, "flow": [60, 65, 82, 88, 111], "ilay": 60, "analogu": 60, "goal": [60, 64, 97], "registernodeconversionpattern": [60, 89], "helper": [60, 93], "pattern": [60, 76, 89, 109], "schema": [60, 89, 93, 95], "caus": [60, 64, 80, 101, 102, 104, 109, 115], "acthardtanh": 60, "torchtrt_unus": 60, "hardtanh": [60, 70], "scalar": [60, 70], "min_val": [60, 70], "max_val": [60, 70], "unwraptodoubl": 60, "new_lay": 60, "addactiv": 60, "activationtyp": [60, 65], "kclip": 60, "torchtrt_check": 60, "unabl": [60, 89, 95], "setalpha": 60, "setbeta": 60, "setnam": [60, 89], "util": [60, 62, 74, 77, 89, 91, 101, 104, 106, 107, 108, 109, 111, 112, 117], "node_info": [60, 89], "c_str": [60, 89], "out_tensor": [60, 89], "associatevalueandtensor": [60, 89], "getoutput": [60, 89], "log_debug": 60, "getdimens": [60, 89], "accord": [60, 64, 77, 94], "unwrap": 60, "tool": [60, 64, 65, 66, 89, 93, 97, 111], "don": [60, 65, 80, 82, 83, 91, 93, 108, 112, 113], "annot": [60, 89], "your": [60, 63, 64, 66, 67, 68, 75, 80, 82, 83, 87, 88, 89, 90, 92, 97, 113, 115], "Its": [60, 82], "track": [60, 91], "sort": [60, 70, 92], "live": [60, 82], "directli": [60, 62, 63, 66, 69, 74, 76, 91, 93, 95, 101, 116], "associatevalueandivalu": 60, "inspect": [60, 88, 89], "dataflow": [60, 89], "mechan": [60, 64, 65, 95, 100, 111], "safe": [60, 64, 71, 75, 76, 77], "unsur": 60, "deep": [60, 64, 69, 80, 91, 95, 118], "straight": 60, "chanc": 60, "none": [60, 64, 65, 70, 71, 72, 74, 75, 76, 77, 80, 82, 93, 95, 97, 101, 108, 109], "wrapper": [60, 65, 116], "similar": [60, 63, 64, 65, 66, 89, 92, 95, 106, 107], "tocustomclass": 60, "tensorcontain": 60, "istensor": 60, "iscustomclass": 60, "lot": [60, 63], "singular": 60, "becaus": [60, 65, 66, 72, 88, 89, 93, 95, 96, 97, 109, 114], "alloc": 60, "freed": 60, "destructor": 60, "destroi": [60, 83], "realli": 60, "think": [60, 82], "becom": [60, 66, 99], "benefit": [60, 89, 97, 109], "deal": [60, 97], "quit": [60, 66, 89, 111], "effici": 60, "batch_norm": [60, 70], "fusion": [60, 62, 65], "deeplearn": [61, 65, 67], "sdk": [61, 67, 117], "matrix": 61, "html": [61, 66, 67, 82, 88, 91, 93, 95, 113], "c_api": 61, "python_api": 61, "org": [61, 66, 80, 82, 88, 89, 91, 93, 95, 113, 115], "stabl": [61, 67, 69, 77, 78, 80, 98, 110, 113], "master": [61, 66, 91, 115], "overview": [61, 69, 101, 105], "md": 61, "appli": [62, 63, 91, 100], "desir": [62, 71, 83, 91, 97], "coalesc": 62, "insert": [62, 64, 71, 89, 91, 93, 97, 100], "graphmodul": [62, 63, 71, 72, 76, 89, 90, 95, 100, 116, 117], "caller": 62, "invok": [62, 64, 65, 88, 89, 115], "lint": 62, "recompil": [62, 71, 76, 93, 97, 100, 104, 113, 117], "repair": 62, "disallow": 62, "repair_input_as_output": 62, "gm": [62, 71], "sample_input": [62, 65, 101], "scenario": [62, 64, 99, 109], "clone": [62, 66, 70, 95], "modified_graph": 62, "extract": [62, 89, 111], "placehold": [62, 93], "isinst": [62, 65, 95, 108], "issubclass": 62, "direct": [62, 86, 100, 115], "len": [62, 70, 95], "direct_output": 62, "inserting_aft": 62, "cloned_placehold": 62, "replace_input_with": 62, "date": [62, 83, 117], "eliminate_dead_cod": 62, "logger": [62, 73], "f": [62, 64, 65, 67, 76, 82, 88, 93, 94, 95, 108, 109], "__init__": [62, 75, 76, 82, 88, 93, 95, 97, 101, 108, 113, 114], "pass_manag": 62, "passmanag": 62, "backend": [62, 68, 69, 77, 78, 81, 92, 96, 97, 98, 101, 108, 110, 113], "offer": [62, 64], "registr": [62, 65], "conveni": [62, 91, 104, 111, 115, 117], "control": [62, 65, 88, 100, 109, 115], "_aten_lowering_pass": 62, "my_custom_pass": 62, "front": [62, 71], "passlist": 62, "arbitrari": [62, 75], "remov": [62, 63, 71, 80, 96, 97, 108], "dump_lowering_pass": 62, "apply_lowering_pass": 62, "graph_modul": [62, 71], "_remove_lowering_pass": 62, "evolv": 62, "introduc": [63, 65, 111], "exportedprogram": [63, 68, 71, 76, 100, 106, 107, 109, 113, 117], "dynamo": [63, 64, 66, 68, 74, 75, 76, 78, 89, 93, 94, 95, 96, 97, 98, 100, 101, 102, 104, 105, 108, 109, 110, 113, 114], "frontend": [63, 71, 74, 90, 95, 98, 102, 104, 108, 110, 113], "simpl": [63, 64, 65, 82, 83, 88, 111, 112, 113], "usag": [63, 65, 69, 74, 78, 82, 89, 98, 109, 110, 113, 116], "eval": [63, 68, 89, 90, 93, 94, 96, 97, 99, 100, 101, 102, 104, 105, 106, 107, 108, 109, 112, 113, 114, 116], "exp_program": [63, 97, 100, 108, 113], "trt_gm": [63, 68, 97, 100, 113, 114, 116], "interact": [63, 82, 99, 101, 102, 103, 104, 105], "ideal": 63, "discuss": [63, 64, 112], "section": [63, 65, 80, 82, 83, 84, 86, 89, 91, 112, 116], "frequent": 63, "builder": [63, 64, 65, 71], "respect": [63, 64, 66, 71, 76, 106, 107, 114], "releas": [63, 64, 67, 82], "insid": [63, 82, 95, 112], "decomposit": [63, 64, 71, 76, 95], "downstream": [63, 111], "constraint": [63, 109], "guid": [64, 81], "present": [64, 100], "learn": [64, 66, 69, 89, 91, 95, 112, 118], "acceler": [64, 72, 76, 115, 117, 118], "workflow": [64, 65, 68, 69, 71, 72, 76, 89, 92, 97, 98, 99, 102, 103, 104, 106, 107, 110, 111], "wide": [64, 76, 86], "varieti": [64, 112], "primari": [64, 93, 97, 116], "simplic": 64, "optimized_model": [64, 68, 96, 101, 102, 104], "depth": [64, 80, 111], "challeng": [64, 99, 112], "addition": [64, 95], "fit": [64, 82], "compilationset": [64, 71, 75, 93, 95, 101], "_enum": [64, 71], "callabl": [64, 71, 76], "pass_through_build_failur": [64, 71, 75, 76, 95], "max_aux_stream": [64, 71, 75, 76, 95], "version_compat": [64, 71, 75, 76, 95], "optimization_level": [64, 71, 75, 76, 95, 101], "use_python_runtim": [64, 71, 75, 76, 95, 96, 97, 99, 100, 101], "truncate_doubl": [64, 71, 75, 76, 95, 96, 106, 107, 109], "use_fast_partition": [64, 71, 75, 76, 95], "enable_experimental_decomposit": [64, 71, 75, 76, 95], "_devic": [64, 71], "assume_dynamic_shape_support": [64, 71, 75, 76], "make_refitt": [64, 71, 75, 76, 96, 97, 99, 100], "engine_cap": [64, 71, 75, 76, 95], "dryrun": [64, 71, 75, 76, 95], "hardware_compat": [64, 71, 75, 76, 95], "timing_cache_path": [64, 71, 75, 76, 97], "tmp": [64, 71, 75, 76, 89, 96], "torch_tensorrt_engine_cach": [64, 71, 75, 76], "timing_cach": [64, 65, 71, 75, 76], "bin": [64, 66, 67, 71, 75, 76], "lazy_engine_init": [64, 71, 75, 76], "cache_built_engin": [64, 71, 75, 96, 97], "reuse_cached_engin": [64, 71, 75, 96, 97, 100], "use_explicit_typ": [64, 71, 75, 106, 107, 109, 114], "use_fp32_acc": [64, 71, 75, 106, 107], "enable_weight_stream": [64, 71, 75, 109], "enable_cross_compile_for_window": [64, 71, 75], "dpython": [64, 71, 76, 77], "per": [64, 71, 95, 115], "regardless": [64, 71, 83, 102, 104], "fail": [64, 71, 76, 89, 99, 100, 108, 118], "auxiliari": [64, 71], "stream": [64, 69, 71, 76, 95, 98, 110], "impli": [64, 71], "longer": [64, 66, 71, 76, 80, 94, 115], "search": [64, 69, 71, 76, 80], "strictli": [64, 71], "runtim": [64, 66, 68, 69, 71, 76, 89, 93, 98, 99, 101, 104, 105, 109, 110, 117], "presenc": [64, 71], "preferenti": [64, 71], "choos": [64, 65, 71, 88], "float64": [64, 71, 76, 77], "refitt": [64, 71, 97], "toggl": [64, 71, 76], "mode": [64, 65, 71, 75, 76, 90, 91, 93, 105, 108], "detail": [64, 65, 67, 71, 88, 89, 95, 97, 112, 115], "natur": [64, 71, 82], "architectur": [64, 66, 69, 71, 76, 94, 97, 111], "amper": [64, 71, 76], "newer": [64, 66, 71, 76], "storag": [64, 71, 91], "use_strong_typ": [64, 71], "strong": [64, 71, 82], "mix": [64, 69, 71], "happen": [64, 65, 71, 88, 99, 102, 113], "were": [64, 71, 95, 100, 115], "cross": [64, 71, 82, 98, 110], "window": [64, 71, 82], "sub": [64, 70, 82, 88, 101], "slate": 64, "futur": [64, 65, 71, 76, 77, 115], "occur": [64, 109], "first_output": 64, "subsequ": [64, 97], "second_output": 64, "session": [64, 68, 82, 97, 105], "point": [64, 66, 76, 80, 81, 82, 89, 108, 112], "cover": [64, 93], "benchmark": [64, 70], "automat": [64, 67, 76, 82, 89, 100, 113, 117], "vari": [64, 72, 109, 113], "distribut": [64, 67, 89, 91, 109, 115], "inf": 64, "dynamo_convers": 64, "contribut": 64, "demonstr": [64, 82, 83, 84, 91, 93, 95, 97, 99, 108, 111, 112], "break": [64, 65, 71, 75, 76, 82, 95, 107], "successfulli": [64, 94, 99, 100], "_dynamo": [64, 96, 97, 101, 102, 104, 113], "explain": [64, 65, 69], "veri": [64, 65, 83, 84, 91, 92, 106, 107, 112], "explan": [64, 65], "graph_break_count": 64, "furthermor": 64, "durat": [64, 82], "latter": [64, 75], "logic": [64, 65, 93], "guard": 64, "compos": [65, 88, 91, 93, 108, 112], "variou": [65, 118], "etc": [65, 80, 82, 95, 118], "environ": [65, 68, 112], "research": 65, "few": [65, 66, 76, 93], "nightli": 65, "lower_exampl": 65, "welcom": [65, 89], "finish": 65, "converison": 65, "pleas": [65, 67, 76, 82, 89, 98, 108, 110, 112, 113], "max_batch_s": [65, 72, 112], "2048": [65, 72], "max_workspace_s": [65, 72], "33554432": [65, 72], "explicit_batch_dimens": [65, 72], "lower_precis": [65, 72], "lowerprecis": [65, 72], "verbose_log": [65, 72], "timing_cache_prefix": [65, 72], "save_timing_cach": [65, 72], "cuda_graph_batch_s": [65, 72], "dynamic_batch": [65, 72], "turn": [65, 72, 105], "trtmodul": [65, 72], "otherwis": [65, 66, 72, 97, 115], "implicit": [65, 70, 72, 82], "config": [65, 66, 72, 112], "updat": [65, 66, 67, 71, 72, 76, 95, 100], "dim": [65, 70, 72, 95, 97, 108, 109, 112, 113], "fx2trt_exampl": 65, "acc_trac": 65, "come": [65, 66, 81, 95, 99, 112], "my_pytorch_model": 65, "build_model": 65, "prepar": [65, 112], "acc_mod": 65, "earli": [65, 100], "deprec": [65, 70], "continu": [65, 82, 115], "backward": [65, 75, 95, 117], "vision": [65, 98, 110, 112], "activ": [65, 75, 77, 82, 89, 91, 93, 111, 115, 118], "except": [65, 71, 76], "permut": [65, 70], "transpos": [65, 70, 113], "ll": [65, 97], "inputtensorspec": [65, 72, 76], "experiment": [65, 76, 77], "dataclass": [65, 101], "re": [65, 76, 82, 97, 99, 105, 115], "manual": [65, 76, 81, 82, 100, 109], "sampl": [65, 71, 82, 90, 91, 99, 100, 101, 102, 103, 104, 105, 106, 107, 112], "rand": [65, 89, 94, 97, 99, 100, 101], "from_tensor": [65, 76], "slightli": [65, 66, 95], "promis": 65, "optimize_target_shap": 65, "input_tensor_spec": 65, "shape_rang": [65, 72], "100": [65, 72, 95, 97, 108, 109], "accordingli": [65, 80, 113, 115], "trtinterpreterresult": [65, 72], "namedtupl": 65, "input_nam": [65, 72], "output_nam": [65, 72], "serialized_cach": [65, 72], "bytearrai": [65, 75, 77], "afford": 65, "temporari": [65, 97], "best": [65, 71, 76, 82, 99, 109, 114], "perforamnc": 65, "examin": 65, "suitabl": [65, 93], "force_fp32_output": 65, "strict_type_constraint": 65, "usual": [65, 66, 80], "unless": 65, "certain": [65, 66, 101, 106, 107, 109, 115], "algorithm_selector": 65, "profiling_verbos": 65, "trt_interpreter_result": 65, "64": [65, 76, 90, 107, 108, 113], "25": [65, 72, 89], "runtimeerror": [65, 108], "xxx": 65, "One": [65, 82, 83, 89, 111, 115], "reload_trt_mod": 65, "reload_model_output": 65, "far": [65, 82], "give": [65, 80, 82], "convtert": 65, "scheme": [65, 71, 76], "action": [65, 82], "tensort": [65, 117], "thing": [65, 66, 82], "compar": [65, 71, 76, 90, 100], "vanilla": 65, "mainli": 65, "builtin": 65, "purpos": [65, 111, 112], "acc_op": 65, "leverag": [65, 91], "power": [65, 82, 89, 109, 111], "goe": [65, 82], "whole": 65, "sigmoid": [65, 70], "tensorrt_convert": 65, "acc_ops_sigmoid": 65, "rest": [65, 82, 83], "input_v": [65, 93], "receiv": 65, "region": 65, "add_activ": 65, "get_output": [65, 95], "wherev": 65, "rememb": [65, 66], "mapper": 65, "todo": [65, 67, 80], "logist": 65, "down": [65, 66, 80, 107], "acc_norm": 65, "foo": [65, 82, 83], "register_acc_op": 65, "register_acc_op_map": 65, "this_arg_is_opt": 65, "op_and_target": 65, "arg_replacement_tupl": 65, "rule": [65, 66, 77], "third": [65, 83], "boolean": [65, 76, 93], "matter": [65, 95], "register_custom_acc_mapper_fn": 65, "design": [65, 74, 93, 99, 106, 109, 111, 118], "redund": 65, "throught": 65, "custom_mapp": 65, "_": [65, 82, 95, 108, 109, 114], "foo_kwarg": 65, "inserting_befor": 65, "foo_nod": 65, "meta": [65, 67, 86, 107, 109], "children": 65, "unit": [65, 76], "test_acc_trac": 65, "acc_op_convert": 65, "essenti": 65, "plugin": [65, 95], "yet": [65, 111], "folder": 65, "center": 66, "pypi": 66, "m": [66, 67, 83, 108], "pip": [66, 67, 98, 110, 112], "upload": [66, 112], "x86": [66, 115], "extra": [66, 75, 89, 95, 99], "url": [66, 80, 112], "download": [66, 67, 86, 91, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 112], "whl": [66, 67], "cu118": 66, "cu124": 66, "tarbal": [66, 89, 91], "easiest": [66, 95], "bazelisk": [66, 67], "bazelbuild": [66, 67], "export": [66, 67, 69, 71, 76, 97, 98, 100, 102, 106, 107, 108, 109, 110, 114, 116, 117], "bazel_vers": 66, "path_to_torchtrt_root": 66, "bazelvers": 66, "mkdir": 66, "cd": [66, 112], "curl": [66, 82], "fssl": 66, "o": [66, 82, 112], "dist": 66, "unzip": 66, "bash": 66, "sh": 66, "cp": [66, 67, 95], "usr": [66, 67], "driver": 66, "branch": [66, 67], "4e5b0f6e860910eb510fa70a76ee3eb9825e7a4d": 66, "l46": 66, "pull": [66, 97, 112], "latest": [66, 67, 80], "l53c1": 66, "fact": 66, "reproduc": 66, "l71": 66, "http_archiv": 66, "build_fil": 66, "archiv": [66, 67], "sha256": 66, "strip_prefix": 66, "OR": 66, "TO": [66, 89], "gnu": 66, "tar": [66, 67, 82, 91], "gz": [66, 82, 83, 91], "ld_library_path": 66, "comment": [66, 82], "uncom": 66, "l114c1": 66, "l124c3": 66, "uv": 66, "astral": 66, "project": [66, 81, 86], "simpler": [66, 91], "wheel": [66, 67], "dep": 66, "lighter": 66, "executor": 66, "avoid": [66, 93, 95, 100, 113], "implic": 66, "python_onli": 66, "legaci": [66, 74], "mainten": 66, "torchdynamo": [66, 113, 117], "technolog": [66, 117], "exclud": [66, 95], "speed": [66, 97, 100], "no_torchscript": 66, "dbg": 66, "pre_cxx11_abi": 66, "complic": 66, "incompat": 66, "popular": [66, 81, 98, 106, 107, 110, 111], "ngc": [66, 67, 112], "tabl": [66, 86], "bdist_wheel": 66, "preinstal": 66, "forum": 66, "correctli": [66, 95], "declar": 66, "intend": [66, 101, 102, 103, 104, 105], "microsoft": 66, "2022": [66, 69], "open": [66, 111, 112], "app": 66, "x64": 66, "prompt": [66, 99, 103, 106, 107], "admin": 66, "privileg": 66, "launcher": 66, "chocolatei": 66, "navig": [66, 80], "ninja": 66, "setuptool": 66, "r": [66, 67, 82, 98, 110], "txt": [66, 67, 98, 110], "distutils_use_sdk": 66, "cuda_win": 66, "libtorch_win": 66, "tensorrt_win": 66, "non": [66, 76, 83, 85, 115], "similarli": [66, 97, 105, 115], "ci_workspac": 66, "win": 66, "tmpl": [66, 67], "torchtrtc": [66, 69, 118], "websit": 66, "finder": 66, "dcmake_module_path": 66, "doesn": [66, 82, 88, 89], "dtorch_dir": 66, "dtensorrt_root": 66, "choic": [66, 74], "b": [66, 70, 76, 83, 109, 112], "dcmake_build_typ": 66, "72048": 66, "jp_workspac": [66, 67], "new_local_repositori": 66, "sudo": [66, 67], "home": 66, "unlik": [66, 92], "libtorch_pre_cxx11_abi": 66, "shift": [66, 70, 82], "jetpack": 66, "jetpack_x": 66, "jetpack_5": 66, "drop": [66, 80, 108], "nvida": 67, "ofjetpack": 67, "With": [67, 80, 82, 89, 91, 97, 112], "incorpor": [67, 83], "cudnn": 67, "9": [67, 86, 89, 94, 95, 112], "dlfw": 67, "09": 67, "jetson": [67, 111], "framework": 67, "instal": [67, 69, 86, 89, 98, 110, 112, 115], "kit": 67, "flash": 67, "board": 67, "apt": 67, "show": [67, 80, 82, 97, 103, 109, 111], "dev": 67, "everth": 67, "nvcc": 67, "cmd": 67, "toolkit": [67, 74], "libcusparselt": 67, "lib64": 67, "wget": [67, 112], "cusparselt": 67, "redist": 67, "libcusparse_lt": 67, "sbsa": 67, "xz": 67, "xf": 67, "v1": [67, 99, 103], "arm64": 67, "mv": 67, "chmod": 67, "pypa": 67, "en": [67, 80], "bootstrap": 67, "jp": 67, "v61": 67, "0a0": 67, "872d972e41": 67, "nv24": 67, "08": 67, "17622132": 67, "cp310": 67, "linux_aarch64": 67, "test_requir": 67, "jetpack6": 67, "lanl": 67, "cuda_vers": 67, "grep": 67, "cut": [67, 82, 100], "sed": [67, 83, 85], "torch_install_path": 67, "dirnam": 67, "__file__": 67, "site_package_path": 67, "cuda_hom": 67, "envsubst": 67, "cxx11": [67, 115], "abi": [67, 115], "anywher": 68, "ahead": [68, 69, 89, 99], "ep": [68, 70, 94, 100, 114, 116], "output_format": [68, 76, 116], "input_tensor": [68, 95, 108, 109], "fill": 68, "aot": [69, 89, 98, 99, 100, 110, 117], "integr": [69, 99, 101], "seamlessli": [69, 76], "ecosystem": [69, 117], "hybrid": [69, 71, 76, 77, 117], "advanc": [69, 78, 83, 91, 98, 110], "bert": [69, 78, 98, 110], "triton": [69, 95], "cudagraph": [69, 98, 110], "overload": [69, 98, 110], "mutabl": [69, 98, 110], "diffus": [69, 78, 98, 110], "gpt2": [69, 98, 110], "llama2": [69, 98, 110], "page": [69, 84, 86, 112], "introductori": 69, "blog": [69, 115], "gtc": 69, "2020": [69, 89], "talk": 69, "fall": [69, 76, 95], "2021": 69, "dai": 69, "confer": 69, "_convolut": [70, 89], "stride": [70, 76, 95, 108], "pad": [70, 76, 95, 108], "dilat": 70, "output_pad": 70, "group": [70, 82, 83], "determinist": 70, "cudnn_en": 70, "allow_tf32": 70, "ab": 70, "aco": 70, "acosh": 70, "adaptive_avg_pool1d": 70, "output_s": 70, "adaptive_avg_pool2d": 70, "adaptive_avg_pool3d": 70, "adaptive_max_pool1d": 70, "adaptive_max_pool2d": 70, "adaptive_max_pool3d": 70, "argmax": [70, 109], "keepdim": 70, "argmin": 70, "asin": 70, "asinh": 70, "atan": 70, "atanh": 70, "avg_pool1d": 70, "kernel_s": [70, 95, 108], "ceil_mod": 70, "count_include_pad": 70, "avg_pool2d": 70, "divisor_overrid": 70, "avg_pool3d": 70, "gamma": 70, "var": 70, "momentum": 70, "bitwise_not": 70, "bmm": 70, "ceil": 70, "clamp": 70, "clamp_max": 70, "clamp_min": 70, "constant_pad_nd": 70, "co": [70, 83, 111], "cosh": 70, "cumsum": 70, "tensor_mod": 70, "rounding_mod": 70, "div_": 70, "elu": 70, "scale": [70, 91, 111], "input_scal": 70, "indic": [70, 71, 80, 82, 93, 100, 102, 113, 114], "padding_idx": 70, "eq": [70, 82], "erf": [70, 93], "exp": 70, "expand_a": 70, "fake_quantize_per_channel_affin": 70, "zero_point": 70, "axi": [70, 76], "quant_min": 70, "quant_max": 70, "fake_quantize_per_tensor_affin": 70, "using_int": [70, 89], "start_dim": [70, 89], "end_dim": [70, 89], "floor": 70, "floor_divid": 70, "ge": 70, "gru_cel": 70, "hx": 70, "w_ih": 70, "w_hh": 70, "b_ih": 70, "b_hh": 70, "gt": 70, "hardtanh_": 70, "instance_norm": 70, "running_mean": 70, "running_var": 70, "use_input_stat": 70, "layer_norm": 70, "normalized_shap": 70, "le": 70, "negative_slop": 70, "01": [70, 83, 89, 108], "leaky_relu_": 70, "lstm_cell": 70, "lt": 70, "masked_fil": 70, "mask": [70, 95], "max_pool1d": 70, "max_pool2d": [70, 88, 89], "max_pool3d": 70, "mul_": [70, 93], "narrow": 70, "neg": [70, 99], "norm": 70, "scalaropt_dim": 70, "pixel_shuffl": 70, "upscale_factor": 70, "pow": 70, "tensor_scalar": 70, "expon": 70, "tensor_tensor": 70, "prelu": 70, "prod": [70, 95], "dim_int": 70, "reciproc": 70, "reflection_pad1d": 70, "reflection_pad2d": 70, "relu_": 70, "repeat_interleav": 70, "self_int": 70, "replication_pad1d": 70, "replication_pad2d": 70, "replication_pad3d": 70, "reshap": [70, 95, 112], "roll": 70, "rsub": 70, "scatter": 70, "sigmoid_": 70, "sin": [70, 82], "sinh": 70, "slice": 70, "split_siz": 70, "split_with_s": 70, "sqrt": 70, "squar": 70, "squeez": [70, 111], "sub_": 70, "dim_intlist": 70, "tan": 70, "tanh": [70, 93], "tanh_": [70, 93], "non_block": [70, 108], "memory_format": [70, 76], "prim_devic": 70, "topk": 70, "k": [70, 91, 108], "largest": 70, "dim0": [70, 97], "dim1": 70, "unbind": 70, "unsqueez": 70, "upsample_bilinear2d": 70, "align_corn": 70, "scales_h": 70, "scales_w": 70, "vec": 70, "scale_factor": 70, "upsample_linear1d": 70, "upsample_nearest1d": 70, "upsample_nearest2d": 70, "upsample_nearest3d": 70, "scales_d": 70, "upsample_trilinear3d": 70, "view": [70, 80], "__and__": 70, "__derive_index": 70, "idx": 70, "__getitem__": 70, "__is__": 70, "t1": 70, "t2": 70, "obj": 70, "__isnot__": 70, "__not__": 70, "__or__": 70, "__range_length": 70, "lo": 70, "hi": [70, 82, 83], "__round_to_zero_floordiv": 70, "__xor__": 70, "append": [70, 93, 96, 97, 108, 109], "el": 70, "arang": [70, 95], "pin_memori": 70, "start_step": 70, "copy_": 70, "float_int": 70, "int_float": 70, "floordiv": 70, "is_floating_point": 70, "numel": 70, "l": [70, 108], "9223372036854775807": 70, "requires_grad": 70, "tupleindex": 70, "tup": 70, "exported_program": [71, 76, 116], "arg_input": [71, 76, 93, 100], "kwarg_input": [71, 76, 100], "engine_cache_dir": [71, 96, 97], "engine_cache_s": [71, 96, 97], "custom_engine_cach": [71, 97], "baseenginecach": [71, 97], "int32": [71, 76, 77, 95, 96, 104, 111], "channel_last": [71, 76, 77, 111], "244": [71, 76, 77], "alia": [71, 76], "better": [71, 76, 88, 111, 117], "understand": [71, 76, 113], "convolut": [71, 76, 77, 91, 95, 118], "_c": [71, 76, 77, 92], "oppos": [71, 76, 77], "lean": [71, 76], "spend": [71, 76], "integ": [71, 76, 85], "faster": [71, 76, 96, 97, 111], "parition": [71, 76], "increas": [71, 76, 97, 109], "amount": [71, 76, 109], "defer": [71, 76, 117], "lead": [71, 76, 82, 109, 115], "oversubscript": [71, 76], "hard": [71, 100], "disk": [71, 76, 97], "space": [71, 82, 83, 91], "byte": [71, 75, 76, 77, 95, 97, 109, 111], "1gb": [71, 96, 97], "exce": 71, "oldest": 71, "gear": [71, 91], "toward": [71, 91], "cross_compile_flag": 71, "cross_compil": 71, "refit_module_weight": [71, 100], "compiled_modul": [71, 100], "new_weight_modul": [71, 100], "verify_output": [71, 100], "use_weight_map_cach": [71, 100], "in_plac": [71, 100], "compmil": 71, "coverag": [71, 95], "min_acc_module_s": 72, "is_aten": 72, "use_experimental_fx_rt": 72, "correctness_atol": 72, "correctness_rtol": 72, "minim": [72, 91, 95], "submodul": [72, 88, 95], "fx2trt": 72, "cpu": [72, 106, 107, 109], "has_batch_dim": 72, "dtyep": 72, "prop": 72, "min_input_shap": 72, "optimized_input_shap": 72, "max_input_shap": 72, "popul": 72, "225": [72, 112], "explicit_precis": 72, "logger_level": 72, "model_trt": 73, "model_torchtrt": 73, "internal_error": 73, "dataloadercalibr": [74, 91], "preprocess": [74, 91, 112], "algo_typ": [74, 91], "calibrationalgo": [74, 91], "cachecalibr": [74, 91], "qualnam": [74, 76], "entropy_calibr": 74, "entropy_calibration_2": [74, 91], "legacy_calibr": 74, "minmax_calibr": 74, "set_multi_device_safe_mod": [75, 115], "_multidevicesafemodecontextmanag": 75, "impact": 75, "suppress": 75, "unsaf": 75, "trt_compiled_modul": 75, "torchtensorrtmodul": [75, 95], "encompass": [75, 77], "simpili": 75, "de": 75, "initi": [75, 76, 82, 100, 101, 102, 104, 105, 106, 107], "scriptmodul": [75, 76, 77, 89, 90, 116, 117], "overridden": [75, 76], "subclass": 75, "although": [75, 82], "recip": [75, 91], "afterward": 75, "former": 75, "care": 75, "hook": 75, "silent": 75, "get_extra_st": 75, "state_dict": [75, 76, 99], "set_extra_st": 75, "picklabl": 75, "pickl": [75, 95, 97], "load_state_dict": [75, 99, 108], "pythontorchtensorrtmodul": 75, "serialized_engin": [75, 77], "_set": [75, 101], "weight_name_map": 75, "trt_modul": 75, "engine_str": 75, "my_modul": 75, "current_devic": 75, "cudagraphs_validate_shap": 75, "versu": 75, "disable_profil": 75, "enable_profil": 75, "iprofil": 75, "spent": 75, "get_layer_info": 75, "request": [76, 89, 112], "decid": 76, "deseri": [76, 77, 89, 95], "retrac": 76, "strict": [76, 115], "valueerror": [76, 94], "mutabletorchtensorrtmodul": [76, 99], "pytorch_model": 76, "regular": 76, "whenev": 76, "refit_gm": 76, "shape_mod": 76, "_shapemod": 76, "interv": 76, "notat": 76, "bound": 76, "torch_tensor": 76, "tracer": 76, "example_tensor": 76, "optimization_profile_field": 76, "classmethod": 76, "disable_memory_format_check": 76, "core_id": 76, "schedul": [76, 112], "use_default": 76, "try_to": 76, "anoth": [76, 82, 83, 88, 90, 100], "typeerror": 76, "unknown": 76, "succe": 76, "float_dtyp": 76, "failur": 76, "bf16": 76, "try_from": [76, 95], "complex128": 76, "16": [76, 86, 88, 89, 90, 102, 105], "brain": 76, "bfloat16": 76, "f64": 76, "f8": 76, "fp8": 76, "float8": 76, "i32": 76, "sign": [76, 112], "i64": 76, "u8": 76, "unsign": 76, "uint8": 76, "trt_dla": 76, "torchtrt_dla": 76, "_from": 76, "torchtrt_dla_ec": 76, "torchtrt_safety_ec": 76, "saefti": 76, "trt_dla_ec": 76, "standalon": [76, 82], "certifi": 76, "tf": 76, "torchtrt_linear": 76, "cdhw32": 76, "thirti": 76, "row": [76, 83], "spatial": 76, "31": [76, 89], "subscript": [76, 82], "chw16": 76, "sixteen": 76, "15": [76, 82, 86], "chw2": 76, "chw32": 76, "chw4": 76, "four": [76, 82, 83], "dhwc": 76, "equivi": 76, "channels_last_3d": 76, "dhwc8": 76, "eight": 76, "dla_hwc4": 76, "imag": [76, 91, 95, 99, 103, 108, 112], "roundup": 76, "elements": 76, "dla_linear": 76, "planar": 76, "hwc": 76, "channels_last": 76, "hwc16": 76, "hwc8": 76, "least": [76, 82, 83], "ishapelay": 77, "check_method_op_support": 77, "seriali": 77, "put_binding_nam": 77, "tensorrtcompilespec": [77, 92], "scriptclass": 77, "0x7fdbd3710cf0": 77, "_jit_to_tensorrt": 77, "00": 78, "000": [78, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "total": [78, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "galleri": [78, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110], "mem": 78, "torch_compile_advanced_usag": [78, 101], "torch_compile_resnet_exampl": [78, 102], "torch_compile_stable_diffus": [78, 103], "torch_compile_transformers_exampl": [78, 104], "v0": [79, 112], "pytorch_sphinx_them": [80, 87], "conf": [80, 87], "html_theme_opt": 80, "canonical_url": 80, "analytics_id": 80, "logo_onli": 80, "display_vers": 80, "prev_next_buttons_loc": 80, "bottom": 80, "style_external_link": 80, "vcs_pageview_mod": 80, "collapse_navig": 80, "sticky_navig": [80, 84], "navigation_depth": 80, "includehidden": 80, "titles_onli": 80, "canon": 80, "rank": 80, "trail": 80, "slash": 80, "googl": 80, "analyt": 80, "isn": [80, 82, 95], "shown": [80, 82, 89, 114], "sidebar": [80, 86], "button": [80, 82], "icon": [80, 82], "extern": [80, 82, 98, 110], "display_github": 80, "display_gitlab": 80, "gitlab": 80, "bitbucket": 80, "bar": [80, 82], "www": [80, 82, 89, 91, 112], "sphinx": [80, 81, 82, 83, 87, 93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110], "toctre": 80, "lose": 80, "scroll": [80, 84], "unlimit": 80, "header": [80, 82, 83, 89, 112], "render": 80, "github_url": 80, "bitbucket_url": 80, "gitlab_url": 80, "left": [80, 82], "upon": [80, 101, 104], "rst": [80, 82], "visitor": 80, "revert": 80, "misbuild": 80, "properti": [80, 95], "stick": 80, "screen": 80, "vertic": [80, 82], "too": [80, 82, 83], "sticki": [80, 86], "nav": [80, 86], "At": [81, 93, 100], "django": 81, "payment": 81, "dotpai": 81, "dotpayprovid": 81, "seller_id": 81, "pin": 81, "lock": 81, "lang": 81, "pl": 81, "polish": 81, "gatewai": 81, "transfer": 81, "purchas": 81, "item": [81, 83, 108], "param": 81, "seller": 81, "consult": 81, "ui": 81, "languag": [81, 82, 83, 88, 95, 98, 106, 110, 112], "data_item_1": 81, "emphasi": 82, "hyperlink": 82, "uri": 82, "web": 82, "anonym": 82, "label": [82, 91, 108, 111, 112], "substitut": 82, "charact": 82, "exceedingli": 82, "ugli": 82, "problem": [82, 107], "problemat": 82, "ext": [82, 83], "autodoc": [82, 83], "demo": [82, 91], "test_py_modul": [82, 86], "my": [82, 106], "role": 82, "pep": 82, "287": 82, "rfc": 82, "2822": 82, "superscript": 82, "gui": 82, "taken": 82, "height": 82, "interfer": 82, "press": 82, "keyboard": 82, "mous": 82, "mmb": 82, "menuselect": 82, "seen": [82, 83], "whitespac": 82, "signific": [82, 95], "strang": 82, "hyphen": 82, "word": [82, 111], "adjust": 82, "width": [82, 111], "browser": 82, "sentenc": [82, 109, 111], "suppli": [82, 100], "258": 82, "equat": 82, "x_": 82, "x_0": 82, "x_1": 82, "x_2": 82, "x_3": 82, "x_4": 82, "nabla": 82, "frac": 82, "theta": 82, "phi": 82, "restructuredtext": [82, 83], "parser": [82, 94, 108], "colon": 82, "indent": 82, "literal_block": 82, "spaces_and_linebreak": 82, "preserv": [82, 88, 91], "markup_process": 82, "Or": 82, "great": [82, 89, 95, 97, 117], "why": [82, 115], "didn": 82, "blank": 82, "align": 82, "permit": 82, "awai": 82, "eric": 82, "orchestra": 82, "leader": 82, "bee": 82, "philosoph": 82, "ipso": 82, "facto": 82, "But": [82, 89, 100, 109], "got": [82, 89], "vi": 82, "entiti": 82, "said": 82, "entir": [82, 117], "ancient": 82, "injuri": 82, "sing": 82, "elk": 82, "bracket": 82, "miss": [82, 89], "brontosaurus": 82, "thin": 82, "thicker": 82, "middl": 82, "That": [82, 89], "mine": 82, "belong": 82, "me": [82, 83], "ann": 82, "begun": 82, "past": 82, "pars": [82, 89], "someurl": 82, "dev0": 82, "bc95015": 82, "caption": [82, 85], "pane": 82, "shell_command": 82, "echo": 82, "did": 82, "window_nam": 82, "session_nam": 82, "shorthand": 82, "some_funct": 82, "highlight": 82, "THE": 82, "heaven": 82, "hexagram": 82, "six": 82, "unbroken": 82, "primal": 82, "light": [82, 116], "spirit": 82, "weak": 82, "essenc": 82, "energi": 82, "unrestrict": 82, "conceiv": 82, "motion": 82, "regard": [82, 117], "basi": 82, "thu": 82, "persist": 82, "dual": 82, "sens": [82, 89], "univers": 82, "world": 82, "men": 82, "express": 82, "deiti": 82, "human": 82, "denot": [82, 95], "holi": 82, "man": [82, 83], "sage": 82, "ruler": 82, "who": 82, "awaken": 82, "utf": [82, 83], "sphinx_rtd_them": [82, 83], "docstr": [82, 83, 90], "dl": 82, "dt": 82, "tag": [82, 112], "tt": 82, "descnam": 82, "descclassnam": 82, "wrote": 82, "anyth": [82, 83, 115], "programm": 82, "myclass": 82, "dothismethod": 82, "flush": 82, "meth": 82, "capit": 82, "flox": 82, "unreferenc": 82, "nonexist": 82, "extrem": 82, "stuff": 82, "mayb": 82, "bold": 82, "ital": 82, "heck": 82, "put": [82, 111], "13": [82, 86], "backlink": 82, "knowledg": 82, "mind": 82, "ey": 82, "thought": 82, "medium": 82, "peopl": 82, "subsect": 82, "interpol": 82, "indirect": 82, "phrase": 82, "docutil": [82, 83], "sourceforg": [82, 83], "ref": 82, "clickabl": 82, "legend": 82, "revis": [82, 83, 99, 103], "revisit": 82, "enhanc": 82, "structuredtext": 82, "wooden": 82, "nickel": 82, "mad": 82, "scientist": 82, "bigger": 82, "bread": 82, "box": [82, 113, 117], "wash": 82, "behind": 82, "ear": 82, "room": 82, "closet": 82, "bathroom": 82, "trash": 82, "sink": 82, "mother": 82, "g_": 82, "mu": 82, "nu": 82, "pi": 82, "t_": 82, "rho_": 82, "servic": 82, "thing1": 82, "thing2": 82, "thing3": 82, "prose": 82, "provok": 82, "mental": 82, "exert": 82, "reader": 82, "discret": 82, "strongli": [82, 109], "advis": 82, "subtitl": 82, "outsid": 82, "often": 82, "besid": 82, "border": 82, "background": [82, 88], "ok": [82, 89], "transmit": 82, "disconnect": 82, "nonetheless": 82, "semant": 82, "blue": [82, 95], "white": 82, "arab": 83, "roman": 83, "upper": 83, "iii": 83, "iv": 83, "classifi": [83, 88, 89, 108, 111], "paragraph": [83, 86], "z": 83, "commonli": 83, "vm": 83, "david": 83, "goodger": 83, "address": [83, 95, 99], "123": 83, "street": 83, "canada": 83, "a1b": 83, "2c3": 83, "contact": 83, "myself": 83, "organ": 83, "humankind": 83, "2012": 83, "03": 83, "19": [83, 86], "53": 83, "0000": 83, "tue": 83, "jan": 83, "progress": 83, "7302": 83, "wish": 83, "redistribut": 83, "reattribut": 83, "sell": 83, "bui": 83, "rent": 83, "leas": 83, "improv": [83, 115], "quot": 83, "excerpt": 83, "collat": 83, "fold": 83, "stapl": 83, "mutil": 83, "anyon": 83, "heart": 83, "bibliograph": 83, "markup": [83, 86], "literal": 83, "yahoo": 83, "oh": 83, "liter": 83, "heh": 83, "child": 83, "beat": 83, "text": [83, 85, 106, 107, 111], "hehe": 83, "kept": 83, "sai": [83, 111], "cackl": 83, "night": 83, "lone": 83, "guangzhou": 83, "destini": 83, "hope": 83, "dream": 83, "forth": 83, "fifth": 83, "sixth": 83, "lorem": [83, 85], "ipsum": [83, 85], "dolor": [83, 85], "sit": [83, 85], "amet": [83, 85], "consectetur": [83, 85], "adipisc": [83, 85], "elit": [83, 85], "donec": [83, 85], "porttitor": [83, 85], "odio": [83, 85], "posuer": [83, 85], "vita": [83, 85], "ornar": [83, 85], "libero": [83, 85], "matti": 83, "loborti": [83, 85], "justo": [83, 85], "vestibulum": [83, 85], "nibh": [83, 85], "aliquet": [83, 85], "feugiat": [83, 85], "sagitti": [83, 85], "nequ": [83, 85], "qui": [83, 85], "eleifend": 83, "dui": [83, 85], "rutrum": [83, 85], "lectu": [83, 85], "suscipit": [83, 85], "letter": [83, 111], "column": 83, "cell": 83, "span": 83, "nam": [83, 85], "mauri": [83, 85], "arcu": [83, 85], "stub": 83, "behav": 84, "area": 84, "interdum": 85, "nec": 85, "finibu": 85, "dictum": 85, "velit": 85, "ut": 85, "eu": 85, "efficitur": 85, "aliquam": 85, "erat": 85, "diam": 85, "gravida": 85, "imperdiet": 85, "tellu": 85, "nisl": 85, "praesent": 85, "eget": 85, "elementum": 85, "rhoncu": 85, "tincidunt": 85, "suspendiss": 85, "volutpat": 85, "scelerisqu": 85, "tristiqu": 85, "aenean": 85, "condimentum": 85, "risu": 85, "accumsan": 85, "laoreet": 85, "maximu": 85, "sapien": 85, "ligula": 85, "fringilla": 85, "commodo": 85, "proin": 85, "et": 85, "pharetra": 85, "etiam": 85, "turpi": 85, "ant": 85, "luctu": 85, "vel": 85, "malesuada": 85, "dignissim": 85, "mi": 85, "nunc": 85, "augu": 85, "sem": 85, "cursu": 85, "nulla": 85, "pellentesqu": 85, "habit": 85, "morbi": 85, "senectu": 85, "netu": 85, "fame": 85, "ac": 85, "egesta": 85, "placerat": 85, "tortor": 85, "iaculi": 85, "venenati": 85, "cra": 85, "puru": 85, "ero": 85, "vehicula": 85, "fusc": 85, "auctor": 85, "phasellu": 85, "est": 85, "viverra": 85, "conval": 85, "faucibu": 85, "vulput": 85, "feli": 85, "sodal": 85, "maecena": 85, "congu": 85, "semper": 85, "enim": 85, "blandit": 85, "sollicitudin": 85, "urna": 85, "orci": 85, "lacu": 85, "quisqu": 85, "facilisi": 85, "hendrerit": 85, "curabitur": 85, "variu": 85, "bibendum": 85, "massa": 85, "magna": 85, "tempu": 85, "metu": 85, "nisi": 85, "pretium": 85, "leo": 85, "euismod": 85, "ultric": 85, "dapibu": 85, "lacinia": 85, "vivamu": 85, "molesti": 85, "hac": 85, "habitass": 85, "platea": 85, "dictumst": 85, "git": 86, "content": [86, 91, 112], "changelog": 86, "math": 86, "14": [86, 96, 104, 112], "17": 86, "18": [86, 89, 99], "submenu": 86, "symlink": 87, "subtre": 87, "_theme": 87, "html_theme": 87, "html_theme_path": 87, "optimiz": 88, "tutori": [88, 91, 93, 95, 97, 99, 100], "beginn": 88, "intro_to_torchscript_tutori": 88, "briefli": 88, "lenet": [88, 89], "lenetfeatextractor": 88, "conv1": [88, 89], "conv2d": [88, 95, 108], "conv2": [88, 89], "lenetclassifi": 88, "fc1": [88, 89], "120": [88, 89], "fc2": [88, 89], "84": [88, 89], "fc3": [88, 89], "feat": [88, 89], "obvious": 88, "pathwai": 88, "input_data": [88, 90], "traced_model": 88, "pick": [88, 114], "script_model": [88, 92], "perspect": 88, "___torch_mangle_10": 88, "129": 88, "___torch_mangle_9": 88, "119": 88, "___torch_mangle_5": 88, "137": 88, "callmethod": 88, "138": 88, "38": 88, "39": 88, "torch_script_modul": [88, 89], "in_tensor": 88, "fly": 88, "lenet_script": [88, 89], "haven": 89, "acquir": 89, "dyanmo": 89, "almost": [89, 117], "trt_lenet_script": 89, "apr": 89, "56": 89, "04": [89, 112], "credit": 89, "stop": 89, "argc": 89, "argv": 89, "cerr": 89, "cout": 89, "even": [89, 99], "cppdoc": 89, "pretti": 89, "fashion": [89, 111], "enable_precis": 89, "And": 89, "convertgraphtotrtengin": 89, "engine_converted_from_jit": 89, "close": [89, 93], "saw": 89, "576": 89, "346": 89, "539": 89, "0464": 89, "0383": 89, "0678": 89, "0932": 89, "1045": 89, "0805": 89, "0435": 89, "0818": 89, "0208": 89, "0358": 89, "cudafloattyp": 89, "0530": 89, "1691": 89, "2802": 89, "1502": 89, "1056": 89, "1549": 89, "input0": [89, 90], "1063": 89, "input1": [89, 90], "input2": 89, "28": 89, "29": 89, "33": 89, "35": 89, "36": 89, "37": 89, "compilegraph": [89, 91], "transform": [89, 91, 96, 98, 100, 104, 106, 107, 108, 109, 110, 112, 116], "laid": 89, "translat": [89, 100], "aren": 89, "techniqu": [89, 91, 107, 115], "checkmethodoperatorsupport": 89, "modular": 89, "ship": [89, 115], "exhaust": 89, "109": 89, "addlay": 89, "yourself": 89, "question": [89, 93], "outself": 89, "flatten_convert": 89, "unwraptoint": 89, "in_shap": 89, "tovec": 89, "out_shap": 89, "shuffl": [89, 91, 108], "addshuffl": 89, "setreshapedimens": 89, "todim": 89, "extens": [89, 117], "ctype": 89, "cdll": 89, "contributor": 89, "upstream": 89, "pr": 89, "usecas": 90, "sole": [90, 91, 117], "individu": 90, "accuraci": [91, 111], "loss": [91, 111], "infrastructur": [91, 112], "streamlin": 91, "expos": [91, 95], "cpp_frontend": 91, "loading_data_recip": 91, "cifar10": [91, 108], "cstddef": 91, "ktrain": 91, "ktest": 91, "un": 91, "cs": 91, "toronto": 91, "edu": 91, "kriz": 91, "cifar": 91, "is_train": 91, "trim": 91, "use_subset": 91, "new_siz": 91, "mode_": 91, "images_": 91, "targets_": 91, "calibration_dataset": 91, "data_dir": 91, "320": 91, "4914": [91, 108], "4822": [91, 108], "4465": [91, 108], "2023": [91, 108], "1994": [91, 108], "2010": [91, 108], "dataloaderopt": 91, "worker": 91, "virtual": 91, "input_shap": [91, 118], "compile_spec": [91, 94, 102, 118], "kf16": [91, 118], "ki8": 91, "vgg16": [91, 108], "testing_dataset": [91, 108], "totensor": [91, 108, 112], "testing_dataload": [91, 108], "num_work": [91, 108], "vgg": [91, 108], "test_ptq_dataloader_calibr": 91, "test_ptq_trt_calibr": 91, "krizhevski": 91, "hinton": 91, "2009": 91, "tini": 91, "simonyan": 91, "zisserman": 91, "2014": 91, "recognit": [91, 111], "arxiv": 91, "preprint": 91, "1409": 91, "1556": 91, "_jit_to_backend": 92, "mobilenet_v2": 92, "pretrain": [92, 97, 99, 102, 105, 111, 112], "gelu": 93, "sy": 93, "approxim": 93, "suppos": 93, "my_mod": 93, "ex_input": [93, 95], "baselin": 93, "my_standard_gelu": 93, "supports_dynamic_shap": 93, "supersed": 93, "converterprior": 93, "vers": 93, "prior": [93, 97, 113, 115], "distinct": 93, "prepend": 93, "candid": 93, "primit": 93, "compiler_ir": 93, "boilerpl": 93, "focu": [93, 99], "interoper": 93, "aten_ops_gelu": 93, "sourceir": 93, "cheap": 93, "unqiu": 93, "op_count": 93, "get_op_count": 93, "nonloc": 93, "elementwis": 93, "source_ir": 93, "lhs_val": 93, "rhs_val": 93, "x_7": 93, "x_8": 93, "79788456080000003": 93, "x_9": 93, "044714999999999998": 93, "x_10": 93, "x_11": 93, "x_12": 93, "x_13": 93, "x_14": 93, "x_15": 93, "my_custom_gelu": 93, "allclos": [93, 99, 100], "my_mod_erf": 93, "my_gelu_erf": 93, "notic": 93, "minut": [93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "converter_overload": 93, "jupyt": [93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110], "ipynb": [93, 94, 95, 96, 97, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "geforcertx": 94, "4080": 94, "3080": 94, "cross_runtime_compilation_for_window": 94, "trt_resnet": 94, "argpars": [94, 108], "argumentpars": [94, 108], "comil": 94, "add_argu": [94, 108], "parse_arg": [94, 108], "manual_se": [94, 96, 97, 99, 100], "resnet18": [94, 97, 99, 100, 102, 105], "amd64": 94, "loaded_model": 94, "load_cross_compiled_exported_program": 94, "trt_output": 94, "cross_compile_for_window": 94, "cost": [95, 97, 100, 115], "perhap": 95, "overhead": [95, 109, 115], "sake": 95, "circular": 95, "red": 95, "green": 95, "twice": 95, "written": 95, "openai": 95, "formal": 95, "tl": 95, "custom_op": 95, "circ_pad_kernel": 95, "all_pads_0": 95, "all_pads_2": 95, "all_pads_4": 95, "all_pads_6": 95, "orig_dims_0": 95, "orig_dims_1": 95, "orig_dims_2": 95, "orig_dims_3": 95, "y_shape_1": 95, "y_shape_2": 95, "y_shape_3": 95, "x_len": 95, "y_len": 95, "block_siz": 95, "pid": 95, "program_id": 95, "mask_i": 95, "i3": 95, "i2": 95, "i1": 95, "i0": 95, "j0": 95, "j1": 95, "j2": 95, "j3": 95, "load_idx": 95, "mask_x": 95, "launch": [95, 112], "torchtrt_ex": 95, "triton_circular_pad": 95, "mutates_arg": 95, "out_dim": 95, "tolist": 95, "all_pad": 95, "zero": 95, "orig_dim": 95, "blocksiz": 95, "256": [95, 108, 109, 112], "numblock": 95, "tracabl": 95, "prerequisit": 95, "fake": 95, "real": 95, "faketensor": 95, "register_fak": 95, "autograd": 95, "beyond": 95, "register_autograd": 95, "padded_x": 95, "my_model": 95, "2604": 95, "4232": 95, "3041": 95, "0833": 95, "2461": 95, "1270": 95, "2450": 95, "4079": 95, "2887": 95, "2828": 95, "0373": 95, "0332": 95, "3143": 95, "6344": 95, "5638": 95, "1867": 95, "5068": 95, "4363": 95, "7937": 95, "3488": 95, "1350": 95, "7966": 95, "3517": 95, "1379": 95, "5537": 95, "1088": 95, "8950": 95, "0550": 95, "6163": 95, "0109": 95, "5245": 95, "9632": 95, "5686": 95, "3775": 95, "8162": 95, "4216": 95, "4311": 95, "1649": 95, "2091": 95, "3668": 95, "1006": 95, "1447": 95, "0352": 95, "7689": 95, "8131": 95, "_run_on_gpu_0": 95, "_run_on_acc_1": 95, "dry": 95, "50": [95, 111], "count": 95, "__": 95, "aggreg": 95, "stat": 95, "latenc": [95, 109, 115], "abstractli": 95, "pkl": [95, 99], "cupi": 95, "gap": 95, "prealloc": 95, "circularpaddingplugin": 95, "ipluginv2dynamicext": 95, "field_collect": 95, "pluginfieldcollect": 95, "x_shape": 95, "num_output": 95, "plugin_namespac": 95, "plugin_typ": 95, "plugin_vers": 95, "assert": [95, 99, 100], "get_output_datatyp": 95, "input_typ": 95, "get_output_dimens": 95, "output_index": 95, "dimsexpr": 95, "exprbuild": 95, "iexprbuild": 95, "output_dim": 95, "dimensionoper": 95, "configure_plugin": 95, "inp": 95, "dynamicplugintensordesc": 95, "x_dim": 95, "desc": 95, "supports_format_combin": 95, "po": 95, "in_out": 95, "plugintensordesc": 95, "num_input": 95, "enqueu": 95, "input_desc": 95, "output_desc": 95, "in_dtyp": 95, "a_mem": 95, "unownedmemori": 95, "items": 95, "c_mem": 95, "a_ptr": 95, "memorypoint": 95, "c_ptr": 95, "a_d": 95, "memptr": 95, "c_d": 95, "a_t": 95, "as_tensor": 95, "c_t": 95, "cloned_plugin": 95, "__dict__": 95, "circularpaddingplugincr": 95, "iplugincr": 95, "field_nam": 95, "pluginfield": 95, "pluginfieldtyp": 95, "create_plugin": 95, "pluginfieldcollection_": 95, "deserialize_plugin": 95, "pads_dict": 95, "creator": 95, "trt_plugin_registri": 95, "get_plugin_registri": 95, "register_cr": 95, "untyp": 95, "get_trt_tensor": 95, "set_layer_nam": 95, "recal": 95, "intlist": 95, "circular_padding_convert": 95, "retriev": 95, "elsewher": 95, "plugin_registri": 95, "plugin_cr": 95, "get_plugin_cr": 95, "field_config": 95, "eventu": 95, "freez": 95, "_input": 95, "add_plugin_v2": 95, "circular_padding_plugin": 95, "_run_on_acc_0": 95, "grad_fn": 95, "subbackward0": 95, "custom_kernel_plugin": 95, "engine_caching_exampl": [96, 97], "remove_timing_cach": [96, 97], "bertmodel": [96, 104], "random": [96, 97, 99, 100, 109], "seed": [96, 97, 99, 100], "from_pretrain": [96, 99, 103, 104, 106, 107, 109], "uncas": [96, 104, 111], "return_dict": 96, "randint": [96, 104, 109], "compile_bert": 96, "enable_tim": [96, 97], "1st": [96, 97], "measur": [96, 97, 109], "2nd": [96, 97], "3rd": [96, 97], "slower": [96, 97], "messur": [96, 97], "compilation_kwarg": [96, 104], "torch_trt_bert_engine_cach": 96, "30": [96, 97, 99, 100, 102, 104, 114], "synchron": [96, 97, 109], "elapsed_tim": [96, 97], "millisecond": 96, "__name__": [96, 101, 104], "__main__": [96, 101, 104], "engine_caching_bert_exampl": 96, "paid": 97, "upfront": 97, "invalid": 97, "repeatedli": 97, "mitig": 97, "explor": 97, "torch_trt": [97, 99, 100], "_default": 97, "_engine_cach": 97, "flexibl": [97, 117], "histor": 97, "barrier": 97, "reconstruct": 97, "ti": 97, "hash": 97, "magnitud": 97, "torch_compil": [97, 101, 102, 104, 105, 113, 117], "compiled_model": 97, "ms": [97, 109], "dynamo_compil": 97, "example_input": 97, "200": 97, "dynamic_shap": [97, 113], "remot": 97, "systen": 97, "agnost": 97, "implent": 97, "ramenginecach": 97, "held": 97, "engine_cach": 97, "torch_compile_my_cach": 97, "_torch_export_gpt2": [98, 110], "_torch_export_llama2": [98, 110], "sphx_glr_tutorials__rendered_examples_dynamo_cross_runtime_compilation_for_window": [98, 110], "straightforward": 99, "especi": 99, "hug": [99, 106, 107], "face": [99, 106, 107], "difficult": 99, "ever": 99, "walk": [99, 100, 106], "lora": [99, 100], "use_python": 99, "mutable_modul": 99, "model2": [99, 100], "expected_output": [99, 100], "refitted_output": [99, 100], "reload": [99, 117], "checkpoint": [99, 108], "civitai": 99, "12597": 99, "moxin": 99, "diffusionpipelin": [99, 103], "no_grad": [99, 106, 107, 108, 109], "model_id": [99, 103], "runwayml": 99, "hous": 99, "forest": 99, "shuimobysim": 99, "wuchangshuo": 99, "qualiti": 99, "worst": 99, "lowr": 99, "cloudi": 99, "watermark": 99, "pipe": [99, 103], "torch_dtyp": [99, 103], "unet": [99, 103], "negative_prompt": 99, "num_inference_step": 99, "without_lora_mut": 99, "jpg": [99, 112], "procedur": 99, "load_lora_weight": 99, "stablediffusionapi": 99, "load_lora_embed": 99, "weight_nam": 99, "safetensor": 99, "adapter_nam": 99, "lora1": 99, "set_adapt": 99, "adapter_weight": 99, "fuse_lora": 99, "unload_lora_weight": 99, "with_lora_mut": 99, "mutable_torchtrt_module_exampl": 99, "expens": 100, "involv": 100, "occasion": [100, 101, 104], "adapt": 100, "infeas": 100, "focus": 100, "mostli": 100, "recogn": 100, "behalf": 100, "init": [100, 108], "sett": 100, "randomli": 100, "exp_program2": 100, "compiled_trt_ep": 100, "new_trt_gm": 100, "accomplish": 100, "gaurente": 100, "attempt": [100, 108, 113], "rebuild": 100, "heurist": 100, "refit_engine_exampl": 100, "x_out": 101, "y_out": 101, "x_y_out": 101, "invoc": 101, "sample_inputs_half": 101, "model_half": 101, "backend_kwarg": 101, "optimized_model_custom": 101, "exit": [101, 104, 112], "2052": [101, 104], "compile_engine_and_inf": [101, 104], "new_input": [102, 104], "new_output": [102, 104], "new_batch_size_input": 102, "new_batch_size_output": 102, "inputs_bs8": 102, "mark_dynam": [102, 113], "outputs_bs8": 102, "No": [102, 113], "inputs_bs12": 102, "outputs_bs12": 102, "compvi": 103, "majest": 103, "castl": 103, "cloud": 103, "majestic_castl": 103, "png": 103, "enable_cudagraph": [105, 115], "out_trt": 105, "set_cudagraphs_mod": [105, 115], "inputs_2": 105, "inputs_3": 105, "out_trt_2": 105, "out_trt_3": 105, "torch_export_cudagraph": 105, "automodelforcausallm": [106, 107, 109], "autotoken": [106, 107], "export_llm": [106, 107, 109], "max_token": [106, 107, 109], "kv_cach": [106, 107], "token": [106, 107, 111], "pad_token_id": 106, "eos_token_id": [106, 107], "attn_implement": [106, 107, 109], "eager": [106, 107, 109], "enjoi": 106, "cute": 106, "dog": 106, "model_input": [106, 107], "return_tensor": [106, 107], "input_id": [106, 107], "regress": [106, 107], "huggingfac": [106, 107, 111], "pyt_gen_token": [106, 107], "gpt2_ep": 106, "max_seq_len": [106, 107, 109], "trt_gen_token": [106, 107], "skip_special_token": [106, 107], "parallel": 106, "paradigm": 106, "torch_export_gpt2": 106, "llama_path": [107, 109], "llama": [107, 109], "7b": [107, 109], "chat": [107, 109], "hf": [107, 109], "llama2_ep": [107, 109], "batch_decod": 107, "clean_up_tokenization_spac": 107, "solv": [107, 112], "smaller": [107, 111], "subproblem": 107, "torch_export_llama2": 107, "modelopt": 108, "mtq": 108, "export_torch_mod": 108, "layer_spec": 108, "num_class": 108, "1000": [108, 109, 112], "init_weight": 108, "in_channel": 108, "pool": [108, 118], "maxpool2d": 108, "batchnorm2d": 108, "sequenti": 108, "avgpool": 108, "adaptiveavgpool2d": 108, "4096": 108, "dropout": 108, "_initialize_weight": 108, "kaiming_normal_": 108, "fan_out": 108, "nonlinear": 108, "constant_": 108, "elif": 108, "normal_": 108, "vgg16_cfg": 108, "128": [108, 109], "ckpt": 108, "model_state_dict": 108, "device_count": 108, "ordereddict": 108, "new_state_dict": 108, "forget": 108, "training_dataset": 108, "randomcrop": 108, "randomhorizontalflip": 108, "training_dataload": 108, "drop_last": 108, "crit": 108, "crossentropyloss": 108, "calibrate_loop": 108, "pred": 108, "5f": 108, "acc": 108, "2f": 108, "quantize_typ": 108, "quant_cfg": 108, "int8_default_cfg": 108, "fp8_default_cfg": 108, "forward_loop": 108, "qdq": 108, "incomplet": 108, "functionaltensor": 108, "functionaltensormod": 108, "_trace": 108, "_export": 108, "float8_e4m3fn": 108, "class_prob": 108, "class_pr": 108, "test_prob": 108, "test_pr": 108, "test_loss": 108, "test_acc": 108, "vgg16_ptq": 108, "overcom": 109, "throughput": 109, "sometim": [109, 113], "outweigh": 109, "slowdown": 109, "hardwar": [109, 118], "experi": 109, "balanc": 109, "timeit": 109, "time_gener": 109, "output_seq_length": 109, "seq_len": [109, 113], "llm": 109, "input_seq": 109, "start_tim": 109, "default_tim": 109, "inputs_copi": 109, "decod": 109, "logit": 109, "next_token_logit": 109, "next_token": 109, "end_tim": 109, "time_mean_m": 109, "isl": 109, "osl": 109, "warm": 109, "solut": 109, "insight": 109, "weight_streaming_ctx": 109, "weight_stream": 109, "mean_lat": 109, "percentag": 109, "weight_budget_pct": 109, "device_budget": 109, "total_device_budget": 109, "permiss": 109, "equal": 109, "proportion": 109, "streamabl": 109, "streamable_budget": 109, "requested_budget": 109, "get_automatic_weight_streaming_budget": 109, "weight_streaming_exampl": 109, "_rendered_examples_python": 110, "_rendered_examples_jupyt": 110, "acoust": 111, "speech": 111, "quartznet": 111, "contextnet": 111, "subword": 111, "piec": 111, "excit": 111, "se": 111, "audio": 111, "transcrib": 111, "speedup": 111, "obtain": [111, 116], "feedforward": 111, "cnn": 111, "uniformli": 111, "resolut": 111, "highli": [111, 112], "compound": 111, "coeffici": 111, "b0": 111, "corpu": 111, "english": 111, "supervis": 111, "walkthrough": 111, "overal": 111, "adopt": 111, "mobilenetv2": 111, "classif": 111, "imagenet": 111, "imagenett": 111, "qat": 111, "simul": 111, "hand": 112, "consider": 112, "concurr": 112, "grpc": 112, "aforement": 112, "familiar": 112, "resnet50": 112, "torchhub": 112, "docker": 112, "login": 112, "xx": 112, "yy": 112, "mm": 112, "publish": 112, "22": 112, "pwd": 112, "scratch_spac": 112, "nvcr": 112, "py3": 112, "proce": 112, "hub": 112, "_validate_not_a_forked_repo": 112, "suggest": 112, "simplest": 112, "model_repositori": 112, "pbtxt": 112, "pytorch_libtorch": 112, "input__0": 112, "data_typ": 112, "type_fp32": 112, "output__0": 112, "exact": 112, "encourag": 112, "rm": 112, "8000": 112, "8001": 112, "8002": 112, "the_model_repositori": 112, "tritonserv": 112, "spin": 112, "proceed": 112, "flesh": 112, "img1": 112, "hakaimagazin": 112, "wp": 112, "gulf": 112, "bird": 112, "attrdict": 112, "pyindex": 112, "tritoncli": 112, "jump": 112, "firstli": 112, "resiz": 112, "pil": 112, "httpclient": 112, "triton_to_np_dtyp": 112, "rn50_preprocess": 112, "img_path": 112, "img": 112, "centercrop": 112, "485": 112, "456": 112, "406": 112, "229": 112, "transformed_img": 112, "inferenceservercli": 112, "localhost": 112, "secondli": 112, "inferinput": 112, "set_data_from_numpi": 112, "binary_data": 112, "inferrequestedoutput": 112, "class_count": 112, "lastli": 112, "send": 112, "model_nam": 112, "inference_output": 112, "as_numpi": 112, "468750": 112, "90": 112, "523438": 112, "92": 112, "664062": 112, "429688": 112, "136": 112, "234375": 112, "confidence_scor": 112, "classification_index": 112, "eagerli": 113, "swap": 113, "exactli": 113, "_tracer": 113, "queri": 113, "attn_weight": 113, "compiler_dynamic_shap": 113, "inputs_bs2": 113, "mymodul": 114, "linear1": 114, "linear2": 114, "linear3": 114, "40": 114, "__myl_mulsum_myl0_0": 114, "layertyp": 114, "kgen": 114, "__mye116_dconst": 114, "__myln_k_arg__bb1_2": 114, "tacticnam": 114, "__myl_mulsum_0xfa6c1858aea1b13b03f90165d7149ec6": 114, "streamid": 114, "__myl_addresmulsum_myl0_1": 114, "__mye131_dconst": 114, "addmm_constant_0": 114, "addmm_add_broadcast_to_same_shape_lhs_broadcast_constantfloat": 114, "__myln_k_arg__bb1_3": 114, "__myl_addresmulsum_0xb3915d7ebfe48be45b6d49083479e12f": 114, "__myl_addresmulsumadd_myl0_2": 114, "__mye146_dconst": 114, "addmm_2_constant_0": 114, "addmm_2_add_broadcast_to_same_shape_lhs_broadcast_constantfloat": 114, "addmm_1_constant_0": 114, "addmm_1_add_broadcast_to_same_shape_lhs_broadcast_constantfloat": 114, "__myl_addresmulsumadd_0xcdd0085ad25f5f45ac5fafb72acbffd6": 114, "__myl_mulsumaddcas_myl0_0": 114, "__mye112_dconst": 114, "__myl_mulsumaddcas_0xacf8f5dd9be2f3e7bb09cdddeac6c936": 114, "__myl_resmulsumaddcas_myl0_1": 114, "__mye127_dconst": 114, "addmm_1_add_broadcast_to_same_shape_lhs_broadcast_constanthalf": 114, "__myl_resmulsumaddcas_0x5a3b318b5a1c97b7d5110c0291481337": 114, "__myl_resmulsumadd_myl0_2": 114, "__mye142_dconst": 114, "__myl_resmulsumadd_0x3fad91127c640fd6db771aa9cde67db0": 114, "libtorchtrt_runtim": 115, "dl_open": 115, "ld_preload": 115, "load_librari": 115, "wl": 115, "ltorchtrt": 115, "torchtrt_runtime_exampl": 115, "libtorchtrt_plugin": 115, "neglig": 115, "thread": 115, "alert": 115, "switch": 115, "mismatch": 115, "crash": 115, "sacrif": 115, "incur": 115, "intens": 115, "trt_ep": 116, "stai": 116, "trt_t": 116, "ergonom": 117, "deleg": 117, "believ": 117, "amen": 117, "artifact": 117, "pack": 117, "year": 117, "superset": 117, "codebas": 117, "immedi": 117, "traceabl": 117, "scriptabl": 117, "neural": 118, "deconvolut": 118, "scripted_model": 118}, "objects": {"": [[5, 0, 1, "c.STR", "STR"], [9, 0, 1, "c.TORCHTRT_API", "TORCHTRT_API"], [11, 0, 1, "c.TORCHTRT_HIDDEN", "TORCHTRT_HIDDEN"], [7, 0, 1, "c.TORCH_TENSORRT_MAJOR_VERSION", "TORCH_TENSORRT_MAJOR_VERSION"], [8, 0, 1, "c.TORCH_TENSORRT_MINOR_VERSION", "TORCH_TENSORRT_MINOR_VERSION"], [6, 0, 1, "c.TORCH_TENSORRT_PATCH_VERSION", "TORCH_TENSORRT_PATCH_VERSION"], [12, 0, 1, "c.TORCH_TENSORRT_VERSION", "TORCH_TENSORRT_VERSION"], [10, 0, 1, "c.XSTR", "XSTR"], [0, 1, 1, "_CPPv4N14torch_tensorrt8DataTypeE", "torch_tensorrt::DataType"], [0, 2, 1, "_CPPv4N14torch_tensorrt8DataType8DataTypeE5Value", "torch_tensorrt::DataType::DataType"], [0, 2, 1, "_CPPv4N14torch_tensorrt8DataType8DataTypeEN3c1010ScalarTypeE", "torch_tensorrt::DataType::DataType"], [0, 2, 1, "_CPPv4N14torch_tensorrt8DataType8DataTypeEv", "torch_tensorrt::DataType::DataType"], [0, 3, 1, "_CPPv4N14torch_tensorrt8DataType8DataTypeE5Value", "torch_tensorrt::DataType::DataType::t"], [0, 3, 1, "_CPPv4N14torch_tensorrt8DataType8DataTypeEN3c1010ScalarTypeE", "torch_tensorrt::DataType::DataType::t"], [0, 4, 1, "_CPPv4N14torch_tensorrt8DataType5ValueE", "torch_tensorrt::DataType::Value"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kBoolE", "torch_tensorrt::DataType::Value::kBool"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kCharE", "torch_tensorrt::DataType::Value::kChar"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value7kDoubleE", "torch_tensorrt::DataType::Value::kDouble"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value6kFloatE", "torch_tensorrt::DataType::Value::kFloat"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kHalfE", "torch_tensorrt::DataType::Value::kHalf"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value4kIntE", "torch_tensorrt::DataType::Value::kInt"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kLongE", "torch_tensorrt::DataType::Value::kLong"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value8kUnknownE", "torch_tensorrt::DataType::Value::kUnknown"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kBoolE", "torch_tensorrt::DataType::kBool"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kCharE", "torch_tensorrt::DataType::kChar"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value7kDoubleE", "torch_tensorrt::DataType::kDouble"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value6kFloatE", "torch_tensorrt::DataType::kFloat"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kHalfE", "torch_tensorrt::DataType::kHalf"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value4kIntE", "torch_tensorrt::DataType::kInt"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value5kLongE", "torch_tensorrt::DataType::kLong"], [0, 5, 1, "_CPPv4N14torch_tensorrt8DataType5Value8kUnknownE", "torch_tensorrt::DataType::kUnknown"], [0, 2, 1, "_CPPv4NK14torch_tensorrt8DataTypecv5ValueEv", "torch_tensorrt::DataType::operator Value"], [0, 2, 1, "_CPPv4N14torch_tensorrt8DataTypecvbEv", "torch_tensorrt::DataType::operator bool"], [0, 2, 1, "_CPPv4NK14torch_tensorrt8DataTypeneE8DataType", "torch_tensorrt::DataType::operator!="], [0, 2, 1, "_CPPv4NK14torch_tensorrt8DataTypeneEN8DataType5ValueE", "torch_tensorrt::DataType::operator!="], [0, 3, 1, "_CPPv4NK14torch_tensorrt8DataTypeneE8DataType", "torch_tensorrt::DataType::operator!=::other"], [0, 3, 1, "_CPPv4NK14torch_tensorrt8DataTypeneEN8DataType5ValueE", "torch_tensorrt::DataType::operator!=::other"], [0, 2, 1, "_CPPv4NK14torch_tensorrt8DataTypeeqE8DataType", "torch_tensorrt::DataType::operator=="], [0, 2, 1, "_CPPv4NK14torch_tensorrt8DataTypeeqEN8DataType5ValueE", "torch_tensorrt::DataType::operator=="], [0, 3, 1, "_CPPv4NK14torch_tensorrt8DataTypeeqE8DataType", "torch_tensorrt::DataType::operator==::other"], [0, 3, 1, "_CPPv4NK14torch_tensorrt8DataTypeeqEN8DataType5ValueE", "torch_tensorrt::DataType::operator==::other"], [46, 1, 1, "_CPPv4N14torch_tensorrt6DeviceE", "torch_tensorrt::Device"], [46, 2, 1, "_CPPv4N14torch_tensorrt6Device6DeviceEv", "torch_tensorrt::Device::Device"], [1, 1, 1, "_CPPv4N14torch_tensorrt6Device10DeviceTypeE", "torch_tensorrt::Device::DeviceType"], [46, 1, 1, "_CPPv4N14torch_tensorrt6Device10DeviceTypeE", "torch_tensorrt::Device::DeviceType"], [1, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeE5Value", "torch_tensorrt::Device::DeviceType::DeviceType"], [1, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeEN3c1010DeviceTypeE", "torch_tensorrt::Device::DeviceType::DeviceType"], [1, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeEv", "torch_tensorrt::Device::DeviceType::DeviceType"], [46, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeE5Value", "torch_tensorrt::Device::DeviceType::DeviceType"], [46, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeEN3c1010DeviceTypeE", "torch_tensorrt::Device::DeviceType::DeviceType"], [46, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeEv", "torch_tensorrt::Device::DeviceType::DeviceType"], [1, 3, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeE5Value", "torch_tensorrt::Device::DeviceType::DeviceType::t"], [1, 3, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeEN3c1010DeviceTypeE", "torch_tensorrt::Device::DeviceType::DeviceType::t"], [46, 3, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeE5Value", "torch_tensorrt::Device::DeviceType::DeviceType::t"], [46, 3, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType10DeviceTypeEN3c1010DeviceTypeE", "torch_tensorrt::Device::DeviceType::DeviceType::t"], [1, 4, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5ValueE", "torch_tensorrt::Device::DeviceType::Value"], [46, 4, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5ValueE", "torch_tensorrt::Device::DeviceType::Value"], [1, 5, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5Value4kDLAE", "torch_tensorrt::Device::DeviceType::Value::kDLA"], [46, 5, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5Value4kDLAE", "torch_tensorrt::Device::DeviceType::Value::kDLA"], [1, 5, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5Value4kGPUE", "torch_tensorrt::Device::DeviceType::Value::kGPU"], [46, 5, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5Value4kGPUE", "torch_tensorrt::Device::DeviceType::Value::kGPU"], [1, 5, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5Value4kDLAE", "torch_tensorrt::Device::DeviceType::kDLA"], [1, 5, 1, "_CPPv4N14torch_tensorrt6Device10DeviceType5Value4kGPUE", "torch_tensorrt::Device::DeviceType::kGPU"], [1, 2, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypecv5ValueEv", "torch_tensorrt::Device::DeviceType::operator Value"], [46, 2, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypecv5ValueEv", "torch_tensorrt::Device::DeviceType::operator Value"], [1, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceTypecvbEv", "torch_tensorrt::Device::DeviceType::operator bool"], [46, 2, 1, "_CPPv4N14torch_tensorrt6Device10DeviceTypecvbEv", "torch_tensorrt::Device::DeviceType::operator bool"], [1, 2, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeneE10DeviceType", "torch_tensorrt::Device::DeviceType::operator!="], [46, 2, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeneE10DeviceType", "torch_tensorrt::Device::DeviceType::operator!="], [1, 3, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeneE10DeviceType", "torch_tensorrt::Device::DeviceType::operator!=::other"], [46, 3, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeneE10DeviceType", "torch_tensorrt::Device::DeviceType::operator!=::other"], [1, 2, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeeqE10DeviceType", "torch_tensorrt::Device::DeviceType::operator=="], [46, 2, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeeqE10DeviceType", "torch_tensorrt::Device::DeviceType::operator=="], [1, 3, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeeqE10DeviceType", "torch_tensorrt::Device::DeviceType::operator==::other"], [46, 3, 1, "_CPPv4NK14torch_tensorrt6Device10DeviceTypeeqE10DeviceType", "torch_tensorrt::Device::DeviceType::operator==::other"], [46, 6, 1, "_CPPv4N14torch_tensorrt6Device18allow_gpu_fallbackE", "torch_tensorrt::Device::allow_gpu_fallback"], [46, 6, 1, "_CPPv4N14torch_tensorrt6Device11device_typeE", "torch_tensorrt::Device::device_type"], [46, 6, 1, "_CPPv4N14torch_tensorrt6Device8dla_coreE", "torch_tensorrt::Device::dla_core"], [46, 6, 1, "_CPPv4N14torch_tensorrt6Device6gpu_idE", "torch_tensorrt::Device::gpu_id"], [17, 4, 1, "_CPPv4N14torch_tensorrt16EngineCapabilityE", "torch_tensorrt::EngineCapability"], [17, 5, 1, "_CPPv4N14torch_tensorrt16EngineCapability15kDLA_STANDALONEE", "torch_tensorrt::EngineCapability::kDLA_STANDALONE"], [17, 5, 1, "_CPPv4N14torch_tensorrt16EngineCapability7kSAFETYE", "torch_tensorrt::EngineCapability::kSAFETY"], [17, 5, 1, "_CPPv4N14torch_tensorrt16EngineCapability9kSTANDARDE", "torch_tensorrt::EngineCapability::kSTANDARD"], [47, 1, 1, "_CPPv4N14torch_tensorrt11GraphInputsE", "torch_tensorrt::GraphInputs"], [47, 6, 1, "_CPPv4N14torch_tensorrt11GraphInputs15input_signatureE", "torch_tensorrt::GraphInputs::input_signature"], [47, 6, 1, "_CPPv4N14torch_tensorrt11GraphInputs6inputsE", "torch_tensorrt::GraphInputs::inputs"], [48, 1, 1, "_CPPv4N14torch_tensorrt5InputE", "torch_tensorrt::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN2at6TensorE", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input"], [48, 2, 1, "_CPPv4N14torch_tensorrt5Input5InputEv", "torch_tensorrt::Input::Input"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::dtype"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::format"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::max_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::min_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::opt_shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataType12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::shape"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN2at6TensorE", "torch_tensorrt::Input::Input::tensor"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputEN3c108ArrayRefI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEE8DataTypeNSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 3, 1, "_CPPv4N14torch_tensorrt5Input5InputENSt6vectorI7int64_tEENSt6vectorIdEE12TensorFormat", "torch_tensorrt::Input::Input::tensor_domain"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input5dtypeE", "torch_tensorrt::Input::dtype"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input6formatE", "torch_tensorrt::Input::format"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input9max_shapeE", "torch_tensorrt::Input::max_shape"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input9min_shapeE", "torch_tensorrt::Input::min_shape"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input9opt_shapeE", "torch_tensorrt::Input::opt_shape"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input5shapeE", "torch_tensorrt::Input::shape"], [48, 6, 1, "_CPPv4N14torch_tensorrt5Input13tensor_domainE", "torch_tensorrt::Input::tensor_domain"], [2, 1, 1, "_CPPv4N14torch_tensorrt12TensorFormatE", "torch_tensorrt::TensorFormat"], [2, 2, 1, "_CPPv4N14torch_tensorrt12TensorFormat12TensorFormatE5Value", "torch_tensorrt::TensorFormat::TensorFormat"], [2, 2, 1, "_CPPv4N14torch_tensorrt12TensorFormat12TensorFormatEN2at12MemoryFormatE", "torch_tensorrt::TensorFormat::TensorFormat"], [2, 2, 1, "_CPPv4N14torch_tensorrt12TensorFormat12TensorFormatEv", "torch_tensorrt::TensorFormat::TensorFormat"], [2, 3, 1, "_CPPv4N14torch_tensorrt12TensorFormat12TensorFormatE5Value", "torch_tensorrt::TensorFormat::TensorFormat::t"], [2, 3, 1, "_CPPv4N14torch_tensorrt12TensorFormat12TensorFormatEN2at12MemoryFormatE", "torch_tensorrt::TensorFormat::TensorFormat::t"], [2, 4, 1, "_CPPv4N14torch_tensorrt12TensorFormat5ValueE", "torch_tensorrt::TensorFormat::Value"], [2, 5, 1, "_CPPv4N14torch_tensorrt12TensorFormat5Value13kChannelsLastE", "torch_tensorrt::TensorFormat::Value::kChannelsLast"], [2, 5, 1, "_CPPv4N14torch_tensorrt12TensorFormat5Value11kContiguousE", "torch_tensorrt::TensorFormat::Value::kContiguous"], [2, 5, 1, "_CPPv4N14torch_tensorrt12TensorFormat5Value8kUnknownE", "torch_tensorrt::TensorFormat::Value::kUnknown"], [2, 5, 1, "_CPPv4N14torch_tensorrt12TensorFormat5Value13kChannelsLastE", "torch_tensorrt::TensorFormat::kChannelsLast"], [2, 5, 1, "_CPPv4N14torch_tensorrt12TensorFormat5Value11kContiguousE", "torch_tensorrt::TensorFormat::kContiguous"], [2, 5, 1, "_CPPv4N14torch_tensorrt12TensorFormat5Value8kUnknownE", "torch_tensorrt::TensorFormat::kUnknown"], [2, 2, 1, "_CPPv4NK14torch_tensorrt12TensorFormatcv5ValueEv", "torch_tensorrt::TensorFormat::operator Value"], [2, 2, 1, "_CPPv4N14torch_tensorrt12TensorFormatcvbEv", "torch_tensorrt::TensorFormat::operator bool"], [2, 2, 1, "_CPPv4NK14torch_tensorrt12TensorFormatneE12TensorFormat", "torch_tensorrt::TensorFormat::operator!="], [2, 2, 1, "_CPPv4NK14torch_tensorrt12TensorFormatneEN12TensorFormat5ValueE", "torch_tensorrt::TensorFormat::operator!="], [2, 3, 1, "_CPPv4NK14torch_tensorrt12TensorFormatneE12TensorFormat", "torch_tensorrt::TensorFormat::operator!=::other"], [2, 3, 1, "_CPPv4NK14torch_tensorrt12TensorFormatneEN12TensorFormat5ValueE", "torch_tensorrt::TensorFormat::operator!=::other"], [2, 2, 1, "_CPPv4NK14torch_tensorrt12TensorFormateqE12TensorFormat", "torch_tensorrt::TensorFormat::operator=="], [2, 2, 1, "_CPPv4NK14torch_tensorrt12TensorFormateqEN12TensorFormat5ValueE", "torch_tensorrt::TensorFormat::operator=="], [2, 3, 1, "_CPPv4NK14torch_tensorrt12TensorFormateqE12TensorFormat", "torch_tensorrt::TensorFormat::operator==::other"], [2, 3, 1, "_CPPv4NK14torch_tensorrt12TensorFormateqEN12TensorFormat5ValueE", "torch_tensorrt::TensorFormat::operator==::other"], [37, 2, 1, "_CPPv4N14torch_tensorrt15dump_build_infoEv", "torch_tensorrt::dump_build_info"], [35, 2, 1, "_CPPv4N14torch_tensorrt14get_build_infoEv", "torch_tensorrt::get_build_info"], [17, 5, 1, "_CPPv4N14torch_tensorrt16EngineCapability15kDLA_STANDALONEE", "torch_tensorrt::kDLA_STANDALONE"], [17, 5, 1, "_CPPv4N14torch_tensorrt16EngineCapability7kSAFETYE", "torch_tensorrt::kSAFETY"], [17, 5, 1, "_CPPv4N14torch_tensorrt16EngineCapability9kSTANDARDE", "torch_tensorrt::kSTANDARD"], [16, 4, 1, "_CPPv4N14torch_tensorrt7logging5LevelE", "torch_tensorrt::logging::Level"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level6kDEBUGE", "torch_tensorrt::logging::Level::kDEBUG"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level6kERRORE", "torch_tensorrt::logging::Level::kERROR"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level6kGRAPHE", "torch_tensorrt::logging::Level::kGRAPH"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level5kINFOE", "torch_tensorrt::logging::Level::kINFO"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level15kINTERNAL_ERRORE", "torch_tensorrt::logging::Level::kINTERNAL_ERROR"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level8kWARNINGE", "torch_tensorrt::logging::Level::kWARNING"], [24, 2, 1, "_CPPv4N14torch_tensorrt7logging24get_is_colored_output_onEv", "torch_tensorrt::logging::get_is_colored_output_on"], [22, 2, 1, "_CPPv4N14torch_tensorrt7logging18get_logging_prefixEv", "torch_tensorrt::logging::get_logging_prefix"], [23, 2, 1, "_CPPv4N14torch_tensorrt7logging24get_reportable_log_levelEv", "torch_tensorrt::logging::get_reportable_log_level"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level6kDEBUGE", "torch_tensorrt::logging::kDEBUG"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level6kERRORE", "torch_tensorrt::logging::kERROR"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level6kGRAPHE", "torch_tensorrt::logging::kGRAPH"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level5kINFOE", "torch_tensorrt::logging::kINFO"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level15kINTERNAL_ERRORE", "torch_tensorrt::logging::kINTERNAL_ERROR"], [16, 5, 1, "_CPPv4N14torch_tensorrt7logging5Level8kWARNINGE", "torch_tensorrt::logging::kWARNING"], [26, 2, 1, "_CPPv4N14torch_tensorrt7logging3logE5LevelNSt6stringE", "torch_tensorrt::logging::log"], [26, 3, 1, "_CPPv4N14torch_tensorrt7logging3logE5LevelNSt6stringE", "torch_tensorrt::logging::log::lvl"], [26, 3, 1, "_CPPv4N14torch_tensorrt7logging3logE5LevelNSt6stringE", "torch_tensorrt::logging::log::msg"], [27, 2, 1, "_CPPv4N14torch_tensorrt7logging24set_is_colored_output_onEb", "torch_tensorrt::logging::set_is_colored_output_on"], [27, 3, 1, "_CPPv4N14torch_tensorrt7logging24set_is_colored_output_onEb", "torch_tensorrt::logging::set_is_colored_output_on::colored_output_on"], [28, 2, 1, "_CPPv4N14torch_tensorrt7logging18set_logging_prefixENSt6stringE", "torch_tensorrt::logging::set_logging_prefix"], [28, 3, 1, "_CPPv4N14torch_tensorrt7logging18set_logging_prefixENSt6stringE", "torch_tensorrt::logging::set_logging_prefix::prefix"], [25, 2, 1, "_CPPv4N14torch_tensorrt7logging24set_reportable_log_levelE5Level", "torch_tensorrt::logging::set_reportable_log_level"], [25, 3, 1, "_CPPv4N14torch_tensorrt7logging24set_reportable_log_levelE5Level", "torch_tensorrt::logging::set_reportable_log_level::lvl"], [3, 1, 1, "_CPPv4I0EN14torch_tensorrt3ptq19Int8CacheCalibratorE", "torch_tensorrt::ptq::Int8CacheCalibrator"], [3, 7, 1, "_CPPv4I0EN14torch_tensorrt3ptq19Int8CacheCalibratorE", "torch_tensorrt::ptq::Int8CacheCalibrator::Algorithm"], [3, 2, 1, "_CPPv4N14torch_tensorrt3ptq19Int8CacheCalibrator19Int8CacheCalibratorERKNSt6stringE", "torch_tensorrt::ptq::Int8CacheCalibrator::Int8CacheCalibrator"], [3, 3, 1, "_CPPv4N14torch_tensorrt3ptq19Int8CacheCalibrator19Int8CacheCalibratorERKNSt6stringE", "torch_tensorrt::ptq::Int8CacheCalibrator::Int8CacheCalibrator::cache_file_path"], [3, 2, 1, "_CPPv4N14torch_tensorrt3ptq19Int8CacheCalibratorcvPN8nvinfer115IInt8CalibratorEEv", "torch_tensorrt::ptq::Int8CacheCalibrator::operator nvinfer1::IInt8Calibrator*"], [4, 1, 1, "_CPPv4I00EN14torch_tensorrt3ptq14Int8CalibratorE", "torch_tensorrt::ptq::Int8Calibrator"], [4, 7, 1, "_CPPv4I00EN14torch_tensorrt3ptq14Int8CalibratorE", "torch_tensorrt::ptq::Int8Calibrator::Algorithm"], [4, 7, 1, "_CPPv4I00EN14torch_tensorrt3ptq14Int8CalibratorE", "torch_tensorrt::ptq::Int8Calibrator::DataLoaderUniquePtr"], [4, 2, 1, "_CPPv4N14torch_tensorrt3ptq14Int8Calibrator14Int8CalibratorE19DataLoaderUniquePtrRKNSt6stringEb", "torch_tensorrt::ptq::Int8Calibrator::Int8Calibrator"], [4, 3, 1, "_CPPv4N14torch_tensorrt3ptq14Int8Calibrator14Int8CalibratorE19DataLoaderUniquePtrRKNSt6stringEb", "torch_tensorrt::ptq::Int8Calibrator::Int8Calibrator::cache_file_path"], [4, 3, 1, "_CPPv4N14torch_tensorrt3ptq14Int8Calibrator14Int8CalibratorE19DataLoaderUniquePtrRKNSt6stringEb", "torch_tensorrt::ptq::Int8Calibrator::Int8Calibrator::dataloader"], [4, 3, 1, "_CPPv4N14torch_tensorrt3ptq14Int8Calibrator14Int8CalibratorE19DataLoaderUniquePtrRKNSt6stringEb", "torch_tensorrt::ptq::Int8Calibrator::Int8Calibrator::use_cache"], [4, 2, 1, "_CPPv4N14torch_tensorrt3ptq14Int8CalibratorcvPN8nvinfer115IInt8CalibratorEEv", "torch_tensorrt::ptq::Int8Calibrator::operator nvinfer1::IInt8Calibrator*"], [29, 2, 1, "_CPPv4I0EN14torch_tensorrt3ptq26make_int8_cache_calibratorE19Int8CacheCalibratorI9AlgorithmERKNSt6stringE", "torch_tensorrt::ptq::make_int8_cache_calibrator"], [29, 7, 1, "_CPPv4I0EN14torch_tensorrt3ptq26make_int8_cache_calibratorE19Int8CacheCalibratorI9AlgorithmERKNSt6stringE", "torch_tensorrt::ptq::make_int8_cache_calibrator::Algorithm"], [29, 3, 1, "_CPPv4I0EN14torch_tensorrt3ptq26make_int8_cache_calibratorE19Int8CacheCalibratorI9AlgorithmERKNSt6stringE", "torch_tensorrt::ptq::make_int8_cache_calibrator::cache_file_path"], [30, 2, 1, "_CPPv4I00EN14torch_tensorrt3ptq20make_int8_calibratorE14Int8CalibratorI9Algorithm10DataLoaderE10DataLoaderRKNSt6stringEb", "torch_tensorrt::ptq::make_int8_calibrator"], [30, 7, 1, "_CPPv4I00EN14torch_tensorrt3ptq20make_int8_calibratorE14Int8CalibratorI9Algorithm10DataLoaderE10DataLoaderRKNSt6stringEb", "torch_tensorrt::ptq::make_int8_calibrator::Algorithm"], [30, 7, 1, "_CPPv4I00EN14torch_tensorrt3ptq20make_int8_calibratorE14Int8CalibratorI9Algorithm10DataLoaderE10DataLoaderRKNSt6stringEb", "torch_tensorrt::ptq::make_int8_calibrator::DataLoader"], [30, 3, 1, "_CPPv4I00EN14torch_tensorrt3ptq20make_int8_calibratorE14Int8CalibratorI9Algorithm10DataLoaderE10DataLoaderRKNSt6stringEb", "torch_tensorrt::ptq::make_int8_calibrator::cache_file_path"], [30, 3, 1, "_CPPv4I00EN14torch_tensorrt3ptq20make_int8_calibratorE14Int8CalibratorI9Algorithm10DataLoaderE10DataLoaderRKNSt6stringEb", "torch_tensorrt::ptq::make_int8_calibrator::dataloader"], [30, 3, 1, "_CPPv4I00EN14torch_tensorrt3ptq20make_int8_calibratorE14Int8CalibratorI9Algorithm10DataLoaderE10DataLoaderRKNSt6stringEb", "torch_tensorrt::ptq::make_int8_calibrator::use_cache"], [36, 2, 1, "_CPPv4N14torch_tensorrt10set_deviceEKi", "torch_tensorrt::set_device"], [36, 3, 1, "_CPPv4N14torch_tensorrt10set_deviceEKi", "torch_tensorrt::set_device::gpu_id"], [49, 1, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpecE", "torch_tensorrt::torchscript::CompileSpec"], [49, 2, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecEN5torch3jit6IValueE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec"], [49, 2, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecENSt6vectorI5InputEE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec"], [49, 2, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecENSt6vectorIN3c108ArrayRefI7int64_tEEEE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec"], [49, 2, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecENSt6vectorINSt6vectorI7int64_tEEEE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec"], [49, 3, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecENSt6vectorIN3c108ArrayRefI7int64_tEEEE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec::fixed_sizes"], [49, 3, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecENSt6vectorINSt6vectorI7int64_tEEEE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec::fixed_sizes"], [49, 3, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecEN5torch3jit6IValueE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec::input_signature"], [49, 3, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec11CompileSpecENSt6vectorI5InputEE", "torch_tensorrt::torchscript::CompileSpec::CompileSpec::inputs"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec19allow_shape_tensorsE", "torch_tensorrt::torchscript::CompileSpec::allow_shape_tensors"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec10capabilityE", "torch_tensorrt::torchscript::CompileSpec::capability"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec5debugE", "torch_tensorrt::torchscript::CompileSpec::debug"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec6deviceE", "torch_tensorrt::torchscript::CompileSpec::device"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec12disable_tf32E", "torch_tensorrt::torchscript::CompileSpec::disable_tf32"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec20dla_global_dram_sizeE", "torch_tensorrt::torchscript::CompileSpec::dla_global_dram_size"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec19dla_local_dram_sizeE", "torch_tensorrt::torchscript::CompileSpec::dla_local_dram_size"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec13dla_sram_sizeE", "torch_tensorrt::torchscript::CompileSpec::dla_sram_size"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec18enabled_precisionsE", "torch_tensorrt::torchscript::CompileSpec::enabled_precisions"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec12graph_inputsE", "torch_tensorrt::torchscript::CompileSpec::graph_inputs"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec14min_block_sizeE", "torch_tensorrt::torchscript::CompileSpec::min_block_size"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec20num_avg_timing_itersE", "torch_tensorrt::torchscript::CompileSpec::num_avg_timing_iters"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec14ptq_calibratorE", "torch_tensorrt::torchscript::CompileSpec::ptq_calibrator"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec5refitE", "torch_tensorrt::torchscript::CompileSpec::refit"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec24require_full_compilationE", "torch_tensorrt::torchscript::CompileSpec::require_full_compilation"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec14sparse_weightsE", "torch_tensorrt::torchscript::CompileSpec::sparse_weights"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec22torch_executed_modulesE", "torch_tensorrt::torchscript::CompileSpec::torch_executed_modules"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec18torch_executed_opsE", "torch_tensorrt::torchscript::CompileSpec::torch_executed_ops"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec24truncate_long_and_doubleE", "torch_tensorrt::torchscript::CompileSpec::truncate_long_and_double"], [49, 6, 1, "_CPPv4N14torch_tensorrt11torchscript11CompileSpec14workspace_sizeE", "torch_tensorrt::torchscript::CompileSpec::workspace_size"], [31, 2, 1, "_CPPv4N14torch_tensorrt11torchscript29check_method_operator_supportERKN5torch3jit6ModuleENSt6stringE", "torch_tensorrt::torchscript::check_method_operator_support"], [31, 3, 1, "_CPPv4N14torch_tensorrt11torchscript29check_method_operator_supportERKN5torch3jit6ModuleENSt6stringE", "torch_tensorrt::torchscript::check_method_operator_support::method_name"], [31, 3, 1, "_CPPv4N14torch_tensorrt11torchscript29check_method_operator_supportERKN5torch3jit6ModuleENSt6stringE", "torch_tensorrt::torchscript::check_method_operator_support::module"], [32, 2, 1, "_CPPv4N14torch_tensorrt11torchscript7compileERKN5torch3jit6ModuleE11CompileSpec", "torch_tensorrt::torchscript::compile"], [32, 3, 1, "_CPPv4N14torch_tensorrt11torchscript7compileERKN5torch3jit6ModuleE11CompileSpec", "torch_tensorrt::torchscript::compile::info"], [32, 3, 1, "_CPPv4N14torch_tensorrt11torchscript7compileERKN5torch3jit6ModuleE11CompileSpec", "torch_tensorrt::torchscript::compile::module"], [34, 2, 1, "_CPPv4N14torch_tensorrt11torchscript28convert_method_to_trt_engineERKN5torch3jit6ModuleENSt6stringE11CompileSpec", "torch_tensorrt::torchscript::convert_method_to_trt_engine"], [34, 3, 1, "_CPPv4N14torch_tensorrt11torchscript28convert_method_to_trt_engineERKN5torch3jit6ModuleENSt6stringE11CompileSpec", "torch_tensorrt::torchscript::convert_method_to_trt_engine::info"], [34, 3, 1, "_CPPv4N14torch_tensorrt11torchscript28convert_method_to_trt_engineERKN5torch3jit6ModuleENSt6stringE11CompileSpec", "torch_tensorrt::torchscript::convert_method_to_trt_engine::method_name"], [34, 3, 1, "_CPPv4N14torch_tensorrt11torchscript28convert_method_to_trt_engineERKN5torch3jit6ModuleENSt6stringE11CompileSpec", "torch_tensorrt::torchscript::convert_method_to_trt_engine::module"], [33, 2, 1, "_CPPv4N14torch_tensorrt11torchscript26embed_engine_in_new_moduleERKNSt6stringE6DeviceRKNSt6vectorINSt6stringEEERKNSt6vectorINSt6stringEEE", "torch_tensorrt::torchscript::embed_engine_in_new_module"], [33, 3, 1, "_CPPv4N14torch_tensorrt11torchscript26embed_engine_in_new_moduleERKNSt6stringE6DeviceRKNSt6vectorINSt6stringEEERKNSt6vectorINSt6stringEEE", "torch_tensorrt::torchscript::embed_engine_in_new_module::device"], [33, 3, 1, "_CPPv4N14torch_tensorrt11torchscript26embed_engine_in_new_moduleERKNSt6stringE6DeviceRKNSt6vectorINSt6stringEEERKNSt6vectorINSt6stringEEE", "torch_tensorrt::torchscript::embed_engine_in_new_module::engine"], [33, 3, 1, "_CPPv4N14torch_tensorrt11torchscript26embed_engine_in_new_moduleERKNSt6stringE6DeviceRKNSt6vectorINSt6stringEEERKNSt6vectorINSt6stringEEE", "torch_tensorrt::torchscript::embed_engine_in_new_module::input_binding_names"], [33, 3, 1, "_CPPv4N14torch_tensorrt11torchscript26embed_engine_in_new_moduleERKNSt6stringE6DeviceRKNSt6vectorINSt6stringEEERKNSt6vectorINSt6stringEEE", "torch_tensorrt::torchscript::embed_engine_in_new_module::output_binding_names"], [76, 8, 0, "-", "torch_tensorrt"]], "torch_tensorrt": [[76, 9, 1, "", "Device"], [76, 9, 1, "", "DeviceType"], [76, 9, 1, "", "EngineCapability"], [76, 9, 1, "", "Input"], [76, 9, 1, "", "MutableTorchTensorRTModule"], [76, 12, 1, "", "compile"], [76, 12, 1, "", "convert_method_to_trt_engine"], [76, 9, 1, "", "dtype"], [116, 8, 0, "-", "dynamo"], [72, 8, 0, "-", "fx"], [76, 12, 1, "", "load"], [73, 8, 0, "-", "logging"], [76, 9, 1, "", "memory_format"], [75, 8, 0, "-", "runtime"], [76, 12, 1, "", "save"], [77, 8, 0, "-", "ts"]], "torch_tensorrt.Device": [[76, 10, 1, "", "__init__"], [76, 11, 1, "", "device_type"], [76, 11, 1, "", "dla_core"], [76, 11, 1, "", "gpu_id"]], "torch_tensorrt.DeviceType": [[76, 11, 1, "", "DLA"], [76, 11, 1, "", "GPU"], [76, 11, 1, "", "UNKNOWN"], [76, 10, 1, "", "to"], [76, 10, 1, "", "try_from"], [76, 10, 1, "", "try_to"]], "torch_tensorrt.EngineCapability": [[76, 11, 1, "", "DLA_STANDALONE"], [76, 11, 1, "", "SAFETY"], [76, 11, 1, "", "STANDARD"], [76, 10, 1, "", "to"], [76, 10, 1, "", "try_from"], [76, 10, 1, "", "try_to"]], "torch_tensorrt.Input": [[76, 10, 1, "", "__init__"], [76, 11, 1, "", "dtype"], [76, 10, 1, "", "example_tensor"], [76, 11, 1, "", "format"], [76, 10, 1, "", "from_tensor"], [76, 10, 1, "", "from_tensors"]], "torch_tensorrt.MutableTorchTensorRTModule": [[76, 10, 1, "", "__init__"], [76, 10, 1, "", "compile"], [76, 10, 1, "", "refit_gm"]], "torch_tensorrt.dtype": [[76, 11, 1, "", "b"], [76, 11, 1, "", "bf16"], [76, 11, 1, "", "f16"], [76, 11, 1, "", "f32"], [76, 11, 1, "", "f64"], [76, 11, 1, "", "f8"], [76, 11, 1, "", "i32"], [76, 11, 1, "", "i64"], [76, 11, 1, "", "i8"], [76, 10, 1, "", "to"], [76, 10, 1, "", "try_from"], [76, 10, 1, "", "try_to"], [76, 11, 1, "", "u8"], [76, 11, 1, "", "unknown"]], "torch_tensorrt.dynamo": [[71, 9, 1, "", "CompilationSettings"], [71, 12, 1, "", "compile"], [71, 12, 1, "", "export"], [71, 12, 1, "", "refit_module_weights"], [71, 12, 1, "", "trace"]], "torch_tensorrt.fx": [[72, 9, 1, "", "InputTensorSpec"], [72, 9, 1, "", "TRTInterpreter"], [72, 9, 1, "", "TRTInterpreterResult"], [72, 9, 1, "", "TRTModule"], [72, 12, 1, "", "compile"]], "torch_tensorrt.logging": [[73, 9, 1, "", "debug"], [73, 9, 1, "", "errors"], [73, 9, 1, "", "graphs"], [73, 9, 1, "", "info"], [73, 9, 1, "", "internal_errors"], [73, 9, 1, "", "warnings"]], "torch_tensorrt.memory_format": [[76, 11, 1, "", "cdhw32"], [76, 11, 1, "", "chw16"], [76, 11, 1, "", "chw2"], [76, 11, 1, "", "chw32"], [76, 11, 1, "", "chw4"], [76, 11, 1, "", "dhwc"], [76, 11, 1, "", "dhwc8"], [76, 11, 1, "", "dla_hwc4"], [76, 11, 1, "", "dla_linear"], [76, 11, 1, "", "hwc"], [76, 11, 1, "", "hwc16"], [76, 11, 1, "", "hwc8"], [76, 11, 1, "", "linear"], [76, 10, 1, "", "to"], [76, 10, 1, "", "try_from"], [76, 10, 1, "", "try_to"]], "torch_tensorrt.runtime": [[75, 9, 1, "", "PythonTorchTensorRTModule"], [75, 9, 1, "", "TorchTensorRTModule"], [75, 12, 1, "", "set_multi_device_safe_mode"]], "torch_tensorrt.runtime.PythonTorchTensorRTModule": [[75, 10, 1, "", "__init__"], [75, 10, 1, "", "cudagraphs_validate_shapes"], [75, 10, 1, "", "disable_profiling"], [75, 10, 1, "", "enable_profiling"], [75, 10, 1, "", "forward"], [75, 10, 1, "", "get_layer_info"]], "torch_tensorrt.runtime.TorchTensorRTModule": [[75, 10, 1, "", "__init__"], [75, 10, 1, "", "forward"], [75, 10, 1, "", "get_extra_state"], [75, 10, 1, "", "set_extra_state"]], "torch_tensorrt.ts": [[77, 12, 1, "", "TensorRTCompileSpec"], [77, 12, 1, "", "check_method_op_support"], [77, 12, 1, "", "compile"], [77, 12, 1, "", "convert_method_to_trt_engine"], [77, 12, 1, "", "embed_engine_in_new_module"], [74, 8, 0, "-", "ptq"]], "torch_tensorrt.ts.ptq": [[74, 9, 1, "", "CacheCalibrator"], [74, 9, 1, "", "CalibrationAlgo"], [74, 9, 1, "", "DataLoaderCalibrator"]], "torch_tensorrt.ts.ptq.CalibrationAlgo": [[74, 11, 1, "", "ENTROPY_CALIBRATION"], [74, 11, 1, "", "ENTROPY_CALIBRATION_2"], [74, 11, 1, "", "LEGACY_CALIBRATION"], [74, 11, 1, "", "MINMAX_CALIBRATION"]]}, "objtypes": {"0": "c:macro", "1": "cpp:class", "2": "cpp:function", "3": "cpp:functionParam", "4": "cpp:enum", "5": "cpp:enumerator", "6": "cpp:member", "7": "cpp:templateParam", "8": "py:module", "9": "py:class", "10": "py:method", "11": "py:attribute", "12": "py:function"}, "objnames": {"0": ["c", "macro", "C macro"], "1": ["cpp", "class", "C++ class"], "2": ["cpp", "function", "C++ function"], "3": ["cpp", "functionParam", "C++ function parameter"], "4": ["cpp", "enum", "C++ enum"], "5": ["cpp", "enumerator", "C++ enumerator"], "6": ["cpp", "member", "C++ member"], "7": ["cpp", "templateParam", "C++ template parameter"], "8": ["py", "module", "Python module"], "9": ["py", "class", "Python class"], "10": ["py", "method", "Python method"], "11": ["py", "attribute", "Python attribute"], "12": ["py", "function", "Python function"]}, "titleterms": {"class": [0, 1, 2, 3, 4, 20, 21, 38, 40, 41, 50, 71, 72, 74, 75, 76], "datatyp": 0, "document": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 16, 17, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 46, 47, 48, 49, 61, 69, 85, 86], "devic": [1, 46, 115], "devicetyp": 1, "nest": [1, 46], "relationship": [1, 3, 4, 46, 48], "tensorformat": 2, "templat": [3, 4, 29, 30], "int8cachecalibr": 3, "inherit": [3, 4, 48], "base": [3, 4, 48, 80], "type": [3, 4, 46, 48, 54], "int8calibr": 4, "defin": [5, 6, 7, 8, 9, 10, 11, 12, 19, 50, 108], "str": 5, "torch_tensorrt_patch_vers": 6, "torch_tensorrt_major_vers": 7, "torch_tensorrt_minor_vers": 8, "torchtrt_api": 9, "xstr": 10, "torchtrt_hidden": 11, "torch_tensorrt_vers": 12, "directori": [13, 14, 15, 51], "cpp": [13, 18, 19, 20, 21, 56], "subdirectori": [13, 14], "includ": [14, 18, 19, 20, 21], "torch_tensorrt": [15, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 45, 67, 71, 72, 73, 74, 75, 76, 77, 102, 104, 105, 117], "file": [15, 18, 19, 20, 21, 42, 43, 44, 45, 50, 51], "enum": [16, 17, 38, 39, 50, 74, 76], "level": [16, 80, 82, 83], "enginecap": 17, "log": [18, 22, 23, 24, 25, 26, 27, 28, 39, 42, 73], "h": [18, 19, 20, 21, 42, 43, 44, 45, 56], "content": [18, 19, 20, 21, 38, 39, 40, 41, 80, 81, 82, 83, 84, 85], "definit": [18, 19, 20, 21, 83, 94, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "By": [18, 19], "namespac": [18, 19, 20, 21, 38, 39, 40, 41, 50], "macro": [19, 43], "ptq": [20, 29, 30, 40, 44, 74, 91, 108], "function": [22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 50, 61, 71, 72, 75, 76, 77, 108], "get_logging_prefix": 22, "get_reportable_log_level": 23, "get_is_colored_output_on": 24, "set_reportable_log_level": 25, "set_is_colored_output_on": 27, "set_logging_prefix": 28, "make_int8_cache_calibr": 29, "make_int8_calibr": 30, "torchscript": [31, 32, 33, 34, 41, 60, 66, 69, 88, 89, 92, 116, 117], "check_method_operator_support": 31, "compil": [32, 57, 59, 63, 64, 66, 68, 69, 89, 94, 97, 100, 101, 102, 103, 104, 105, 106, 107, 109, 111, 113, 114, 116, 117], "embed_engine_in_new_modul": 33, "convert_method_to_trt_engin": 34, "get_build_info": 35, "set_devic": 36, "dump_build_info": 37, "program": [42, 43, 44, 45, 63, 100, 115], "list": [42, 43, 44, 45, 83], "struct": [46, 47, 48, 49, 50], "graphinput": 47, "input": [48, 102, 104], "compilespec": 49, "torch": [50, 61, 63, 64, 65, 66, 68, 69, 89, 90, 92, 93, 95, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 110, 111, 112, 113, 114, 115, 116, 117], "tensorrt": [50, 58, 61, 63, 64, 65, 66, 69, 89, 90, 92, 93, 95, 99, 100, 106, 107, 108, 110, 111, 112, 113, 114, 115, 116, 117], "c": [50, 61, 66, 68, 69, 89, 91, 111], "api": [50, 51, 61, 66, 69], "hierarchi": 50, "full": [50, 51], "torchtrtc": [52, 89], "convers": [53, 57, 59, 60], "phase": [53, 55, 56, 57, 58, 59], "node": 53, "evalu": [53, 54, 70], "convert": [53, 54, 60, 65, 70, 89, 93], "write": [54, 60, 62, 93, 95], "dynamo": [54, 62, 69, 71, 106, 107, 116, 117], "implement": [54, 93], "registr": 54, "capabl": 54, "valid": 54, "contract": [54, 60], "exampl": [54, 62, 82, 84, 94], "convolut": 54, "oper": [54, 64, 70, 89, 95], "decomposit": 54, "addmm": [54, 55], "lower": [55, 57, 59, 62], "pass": [55, 62], "us": [55, 61, 89, 90, 92, 93, 95, 101, 102, 103, 104, 105, 106, 107, 108, 111, 113], "eliminatecommonsubexpress": 55, "elimin": 55, "dead": 55, "code": [55, 69, 82], "except": 55, "Or": 55, "pattern": 55, "redund": 55, "guard": 55, "freez": 55, "modul": [55, 88, 89, 99, 117], "fuse": 55, "branch": 55, "linear": 55, "flatten": 55, "graph": [55, 58, 117], "tupl": 55, "fallback": [55, 56], "peephol": 55, "optim": [55, 68, 112], "remov": 55, "contigu": 55, "dropout": 55, "To": 55, "unpack": 55, "logsoftmax": 55, "unrol": 55, "loop": [55, 108], "replac": [55, 82], "tile": 55, "repeat": 55, "partit": [56, 57, 59], "partitoninfo": 56, "segmentedblock": 56, "shape_analysi": 56, "automat": [56, 109], "depend": [56, 66, 98, 110], "awar": [56, 111], "runtim": [57, 58, 59, 75, 94, 115], "background": [58, 60], "engin": [58, 65, 95, 96, 97], "executor": 58, "op": [58, 65, 95], "construct": 58, "result": 58, "serial": [58, 64, 68], "deseri": 58, "abi": [58, 66], "version": [58, 66], "format": [58, 117], "system": [59, 66], "overview": [59, 67], "what": 60, "guarante": 60, "respons": 60, "context": [60, 80, 109], "arg": [60, 81], "weight": [60, 100, 108, 109], "other": 60, "advic": 60, "link": [61, 82], "develop": 61, "avail": 61, "layer": 61, "expect": 61, "dimens": 61, "python": [61, 66, 68, 69, 88, 90, 91], "sometim": 61, "easier": 61, "read": 61, "pytorch": [61, 65, 69, 92, 95, 106, 107, 111], "native_op": 61, "ir": [61, 116, 117], "aten": 62, "basic": 62, "requir": 62, "regist": [62, 89], "export": [63, 68, 105, 113], "customiz": [63, 64], "set": [63, 64, 99, 101, 105, 112], "under": [63, 89, 113], "hood": [63, 89, 113], "trace": 63, "backend": [64, 102, 103, 104, 106, 107], "kei": 64, "featur": 64, "custom": [64, 89, 93, 95, 97, 101, 113], "usag": [64, 100, 101], "after": 64, "model": [64, 65, 69, 94, 95, 98, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 114, 116], "perform": 64, "coverag": 64, "feasibl": 64, "dynam": [64, 102, 111, 113], "shape": [64, 102, 111, 113], "support": [64, 70], "recompil": [64, 102], "condit": 64, "fx": [65, 69, 72, 111, 117], "frontend": [65, 66, 69, 92, 111, 117], "user": [65, 69], "guid": [65, 69], "acc": 65, "tracer": 65, "fx2trt": 65, "how": [65, 80, 91], "add": 65, "miss": 65, "instal": [66, 87], "precompil": 66, "binari": 66, "specif": 66, "cuda": [66, 101, 104], "nightli": 66, "build": [66, 67, 80, 112], "onli": 66, "from": [66, 92], "sourc": 66, "linux": 66, "packag": [66, 115], "addit": 66, "option": [66, 68, 80, 81, 83, 102, 104, 109, 117], "distribut": 66, "No": 66, "librari": [66, 115], "standalon": 66, "releas": 66, "debug": 66, "pre": [66, 108], "cxx11": 66, "choos": 66, "right": 66, "window": [66, 94], "step": [66, 68, 112], "advanc": [66, 100, 101], "setup": 66, "troubleshoot": 66, "altern": 66, "cmake": 66, "nativ": 66, "aarch64": 66, "jetson": 66, "prerequisit": [66, 67], "environ": 66, "cli": [66, 69], "jetpack": 67, "6": [67, 84], "1": [67, 68, 84, 112], "quick": 68, "start": [68, 69], "2": [68, 84, 85, 112], "deploi": [68, 108, 111, 115], "deploy": 68, "In": [69, 100], "framework": 69, "infer": [69, 102, 103, 104, 105, 108, 112], "nvidia": 69, "gpu": 69, "get": 69, "tutori": [69, 110], "zoo": [69, 98, 110], "contributor": 69, "indic": 69, "legaci": [69, 111, 117], "further": 69, "inform": 69, "current": 70, "through": 70, "ts": [74, 77, 117], "submodul": 76, "comput": 78, "time": [78, 117], "changelog": 79, "configur": 80, "project": 80, "wide": 80, "html": 80, "theme": [80, 86], "toc": 80, "page": 80, "tabl": [80, 81, 82, 83, 84, 85], "mod": 81, "test_py_modul": 81, "gener": [81, 106, 107], "index": 81, "paramet": 81, "data": 81, "paragraph": [82, 85], "markup": 82, "inlin": 82, "math": 82, "meta": 82, "block": 82, "liter": 82, "line": 82, "quot": 82, "doctest": 82, "emphas": 82, "number": [82, 83], "sidebar": 82, "ch": 82, "ien": 82, "The": [82, 89], "creativ": 82, "A": 82, "refer": 82, "footnot": 82, "citat": [82, 91], "glossari": 82, "target": 82, "direct": 82, "center": 82, "text": 82, "imag": [82, 83], "figur": 82, "admonit": 82, "And": 82, "wai": 82, "topic": 82, "rubric": 82, "titl": 82, "compound": 82, "download": [82, 87], "enumer": 83, "field": 83, "bullet": 83, "second": 83, "But": 83, "deeper": 83, "down": 83, "rabbit": 83, "hole": 83, "hlist": 83, "grid": 83, "giant": 83, "can": 83, "have": 83, "caption": [83, 86], "like": 83, "thi": [83, 86], "one": 83, "long": [84, 86], "sticki": 84, "nav": 84, "menu": [84, 86], "3": [84, 112], "4": 84, "5": 84, "7": 84, "8": 84, "9": 84, "10": 84, "11": 84, "12": 84, "13": 84, "14": 84, "15": 84, "16": 84, "17": 84, "18": 84, "19": 84, "20": 84, "submenu": 84, "subsubmenu": 84, "structur": 85, "element": 85, "section": 85, "subsect": 85, "subsubsect": 85, "demo": 86, "an": 86, "incred": 86, "via": 87, "git": 87, "creat": [88, 91], "work": [88, 89], "save": [88, 99, 116], "disk": 88, "quickstart": 89, "unsupport": 89, "post": 91, "train": [91, 108, 111], "quantiz": [91, 108, 111], "your": [91, 112], "own": 91, "applic": 91, "directli": 92, "overload": 93, "metadata": 93, "our": [93, 95], "cross": 94, "import": [94, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109], "kernel": 95, "within": 95, "test": 95, "wrap": 95, "insert": 95, "cach": [96, 97, 100], "bert": [96, 104, 111], "jit": [97, 113], "aot": [97, 113], "mutabl": 99, "initi": 99, "make": [99, 100], "modif": 99, "stabl": [99, 103], "diffus": [99, 103], "huggingfac": 99, "refit": 100, "new": 100, "standard": 100, "workflow": 100, "refitt": 100, "pretrain": 100, "map": 100, "place": 100, "default": [101, 105], "cleanup": [101, 104], "driver": [101, 104], "error": [101, 104], "note": [101, 104], "resnet": 102, "argument": [102, 104], "avoid": 102, "specifi": 102, "befor": 102, "trt": 102, "cudagraph": [105, 115], "integr": 105, "gpt2": 106, "output": [106, 107], "decod": [106, 107], "sentenc": [106, 107], "llama2": 107, "load": [108, 116], "dataset": 108, "loss": 108, "calibr": 108, "tune": 108, "fp8": 108, "stream": 109, "run": 109, "budget": 109, "size": 109, "manag": 109, "notebook": 111, "citrinet": 111, "efficientnet": 111, "mask": 111, "languag": 111, "mlm": 111, "hug": 111, "face": 111, "transform": 111, "acceler": 111, "serv": [111, 112], "resnet50": 111, "lenet": 111, "deep": 111, "learn": 111, "object": 111, "detect": 111, "ssd": 111, "int8": 111, "triton": 112, "up": 112, "server": 112, "client": 112, "queri": 112, "constraint": 113, "mix": 114, "precis": 114, "libtorchtrt": 115, "so": 115, "plugin": 115, "multi": 115, "safe": 115, "mode": 115, "exportedprogram": 116, "b": 116, "explain": 117, "just": 117, "accept": 117, "return": 117, "ahead": 117, "dla": 118}, "envversion": {"sphinx.domains.c": 2, "sphinx.domains.changeset": 1, "sphinx.domains.citation": 1, "sphinx.domains.cpp": 6, "sphinx.domains.index": 1, "sphinx.domains.javascript": 2, "sphinx.domains.math": 2, "sphinx.domains.python": 3, "sphinx.domains.rst": 2, "sphinx.domains.std": 2, "nbsphinx": 4, "sphinx.ext.intersphinx": 1, "sphinx.ext.todo": 2, "sphinx.ext.viewcode": 1, "sphinx": 56}}) \ No newline at end of file diff --git a/docs/sg_execution_times.html b/docs/sg_execution_times.html index e86b2f97aa..3d8c138084 100644 --- a/docs/sg_execution_times.html +++ b/docs/sg_execution_times.html @@ -10,7 +10,7 @@ - Computation times — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Computation times — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/src/pytorch-sphinx-theme/docs/changelog.html b/docs/src/pytorch-sphinx-theme/docs/changelog.html index 83687253b7..722f4859cc 100644 --- a/docs/src/pytorch-sphinx-theme/docs/changelog.html +++ b/docs/src/pytorch-sphinx-theme/docs/changelog.html @@ -10,7 +10,7 @@ - Changelog — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Changelog — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/src/pytorch-sphinx-theme/docs/configuring.html b/docs/src/pytorch-sphinx-theme/docs/configuring.html index 9f5413c795..d004da318c 100644 --- a/docs/src/pytorch-sphinx-theme/docs/configuring.html +++ b/docs/src/pytorch-sphinx-theme/docs/configuring.html @@ -10,7 +10,7 @@ - Configuration — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Configuration — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/src/pytorch-sphinx-theme/docs/demo/api.html b/docs/src/pytorch-sphinx-theme/docs/demo/api.html index 33e44dd2f8..92895ef7c0 100644 --- a/docs/src/pytorch-sphinx-theme/docs/demo/api.html +++ b/docs/src/pytorch-sphinx-theme/docs/demo/api.html @@ -10,7 +10,7 @@ - 5. :mod:`test_py_module` — Torch-TensorRT v2.6.0.dev0+e43833d documentation + 5. :mod:`test_py_module` — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/src/pytorch-sphinx-theme/docs/demo/demo.html b/docs/src/pytorch-sphinx-theme/docs/demo/demo.html index e5e3a2818b..2b9910e404 100644 --- a/docs/src/pytorch-sphinx-theme/docs/demo/demo.html +++ b/docs/src/pytorch-sphinx-theme/docs/demo/demo.html @@ -12,7 +12,7 @@ - 3. Paragraph Level Markup — Torch-TensorRT v2.6.0.dev0+e43833d documentation + 3. Paragraph Level Markup — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    @@ -664,7 +664,7 @@

    3.4.4.

    3.4.5. Code Blocks¶

    # parsed-literal test
    -curl -O http://someurl/release-v2.6.0.dev0+e43833d.tar-gz
    +curl -O http://someurl/release-v2.6.0.dev0+bc95015.tar-gz

    Code Blocks can have captions.¶
    {
    diff --git a/docs/src/pytorch-sphinx-theme/docs/demo/lists_tables.html b/docs/src/pytorch-sphinx-theme/docs/demo/lists_tables.html
    index 0f495c8d92..0aee59635b 100644
    --- a/docs/src/pytorch-sphinx-theme/docs/demo/lists_tables.html
    +++ b/docs/src/pytorch-sphinx-theme/docs/demo/lists_tables.html
    @@ -10,7 +10,7 @@
     
       
       
    -  4. Lists & Tables — Torch-TensorRT v2.6.0.dev0+e43833d documentation
    +  4. Lists & Tables — Torch-TensorRT v2.6.0.dev0+bc95015 documentation
       
     
       
    @@ -273,7 +273,7 @@
                   
                   
                     
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/src/pytorch-sphinx-theme/docs/demo/long.html b/docs/src/pytorch-sphinx-theme/docs/demo/long.html index 5ff749ccab..ccc32a6079 100644 --- a/docs/src/pytorch-sphinx-theme/docs/demo/long.html +++ b/docs/src/pytorch-sphinx-theme/docs/demo/long.html @@ -10,7 +10,7 @@ - 1. Long Sticky Nav — Torch-TensorRT v2.6.0.dev0+e43833d documentation + 1. Long Sticky Nav — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/src/pytorch-sphinx-theme/docs/demo/structure.html b/docs/src/pytorch-sphinx-theme/docs/demo/structure.html index 97bca07882..705fab8300 100644 --- a/docs/src/pytorch-sphinx-theme/docs/demo/structure.html +++ b/docs/src/pytorch-sphinx-theme/docs/demo/structure.html @@ -10,7 +10,7 @@ - 1. Structural Elements — Torch-TensorRT v2.6.0.dev0+e43833d documentation + 1. Structural Elements — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/src/pytorch-sphinx-theme/docs/index.html b/docs/src/pytorch-sphinx-theme/docs/index.html index 34d6711acb..f3f727a041 100644 --- a/docs/src/pytorch-sphinx-theme/docs/index.html +++ b/docs/src/pytorch-sphinx-theme/docs/index.html @@ -10,7 +10,7 @@ - <no title> — Torch-TensorRT v2.6.0.dev0+e43833d documentation + <no title> — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/src/pytorch-sphinx-theme/docs/installing.html b/docs/src/pytorch-sphinx-theme/docs/installing.html index f1b24af379..7aa2f8855f 100644 --- a/docs/src/pytorch-sphinx-theme/docs/installing.html +++ b/docs/src/pytorch-sphinx-theme/docs/installing.html @@ -10,7 +10,7 @@ - Installation — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Installation — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/ts/creating_torchscript_module_in_python.html b/docs/ts/creating_torchscript_module_in_python.html index 57df80d557..f2cd314d4e 100644 --- a/docs/ts/creating_torchscript_module_in_python.html +++ b/docs/ts/creating_torchscript_module_in_python.html @@ -10,7 +10,7 @@ - Creating a TorchScript Module — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Creating a TorchScript Module — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/ts/getting_started_with_cpp_api.html b/docs/ts/getting_started_with_cpp_api.html index 4b67101919..c76744ed07 100644 --- a/docs/ts/getting_started_with_cpp_api.html +++ b/docs/ts/getting_started_with_cpp_api.html @@ -10,7 +10,7 @@ - Using Torch-TensorRT in C++ — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Using Torch-TensorRT in C++ — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/ts/getting_started_with_python_api.html b/docs/ts/getting_started_with_python_api.html index b6c102ae0d..eb1b1a44d7 100644 --- a/docs/ts/getting_started_with_python_api.html +++ b/docs/ts/getting_started_with_python_api.html @@ -10,7 +10,7 @@ - Using Torch-TensorRT in Python — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Using Torch-TensorRT in Python — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/ts/ptq.html b/docs/ts/ptq.html index d315b04b55..180589f98f 100644 --- a/docs/ts/ptq.html +++ b/docs/ts/ptq.html @@ -10,7 +10,7 @@ - Post Training Quantization (PTQ) — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Post Training Quantization (PTQ) — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/ts/torchscript_frontend_from_pytorch.html b/docs/ts/torchscript_frontend_from_pytorch.html index a5c8e40aab..4434857da4 100644 --- a/docs/ts/torchscript_frontend_from_pytorch.html +++ b/docs/ts/torchscript_frontend_from_pytorch.html @@ -10,7 +10,7 @@ - Using Torch-TensorRT TorchScript Frontend Directly From PyTorch — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Using Torch-TensorRT TorchScript Frontend Directly From PyTorch — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/converter_overloading.html b/docs/tutorials/_rendered_examples/dynamo/converter_overloading.html index 300479caa0..44a66d6ec6 100644 --- a/docs/tutorials/_rendered_examples/dynamo/converter_overloading.html +++ b/docs/tutorials/_rendered_examples/dynamo/converter_overloading.html @@ -10,7 +10,7 @@ - Overloading Torch-TensorRT Converters with Custom Converters — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Overloading Torch-TensorRT Converters with Custom Converters — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/cross_runtime_compilation_for_windows.html b/docs/tutorials/_rendered_examples/dynamo/cross_runtime_compilation_for_windows.html new file mode 100644 index 0000000000..ad5d896048 --- /dev/null +++ b/docs/tutorials/_rendered_examples/dynamo/cross_runtime_compilation_for_windows.html @@ -0,0 +1,883 @@ + + + + + + + + + + + + + Cross runtime compilation for windows example — Torch-TensorRT v2.6.0.dev0+bc95015 documentation + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
    +
    +
    + + + + + +
    +
    +
    + + + + + + + + + + + +
    +
    +
    + + + + + + + + + + + + + + + + +
    + +
      + +
    • + + + Docs + + > +
    • + + +
    • Cross runtime compilation for windows example
    • + + +
    • + + + + + +
    • + +
    + + +
    +
    + +
    + Shortcuts +
    +
    + +
    +
    + + + + + + +
    + +
    +
    + + +

    cross runtime compilation limitations: +The cross compile and saved model can only be loaded in Windows, it can no longer be loaded in Linux +The cross compile and saved model can only be loaded in the same Compute Capability as the Linux which it was cross compiled +(for example, if the model was cross compiled in Linux with GeForceRTX 4080 which has Compute Capability of 8.9, +It cannot be loaded in Windows with GeForceRTX 3080 which has Compute Capability of 8.6)

    +
    +

    Cross runtime compilation for windows example¶

    +

    Compile and save the Resnet Model using Torch-TensorRT in Linux:

    +

    python examples/dynamo/cross_runtime_compilation_for_windows.py –path trt_resnet.ep

    +

    Load the Resnet Model saved in Windows:

    +

    python examples/dynamo/cross_runtime_compilation_for_windows.py –path trt_resnet.ep –load True

    +
    +

    Imports and Model Definition¶

    +
    import argparse
    +import platform
    +
    +import torch
    +import torch_tensorrt as torchtrt
    +import torchvision.models as models
    +
    +PARSER = argparse.ArgumentParser(
    +    description="Cross runtime comilation for windows example: Resnet Model"
    +)
    +PARSER.add_argument(
    +    "--load", default=False, type=bool, required=False, help="Load the model in Windows"
    +)
    +PARSER.add_argument(
    +    "--path",
    +    type=str,
    +    required=True,
    +    help="Path to the saved model file",
    +)
    +
    +args = PARSER.parse_args()
    +torch.manual_seed(0)
    +model = models.resnet18().eval().cuda()
    +input = torch.rand((1, 3, 224, 224)).to("cuda")
    +inputs = [input]
    +
    +
    +

    According to the argument, it is either cross compile and save resnet model for windows in Linux +or load the saved resnet model in Windows +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

    +
    if args.load:
    +    # load the saved model in Windows
    +    if platform.system() != "Windows" or platform.machine() != "AMD64":
    +        raise ValueError(
    +            "cross runtime compiled model for windows can only be loaded in Windows system"
    +        )
    +    loaded_model = torchtrt.load_cross_compiled_exported_program(args.path).module()
    +    print(f"model has been successfully loaded from ${args.path}")
    +    # inference
    +    trt_output = loaded_model(input)
    +    print(f"inference result: {trt_output}")
    +else:
    +    if platform.system() != "Linux" or platform.architecture()[0] != "64bit":
    +        raise ValueError(
    +            "cross runtime compiled model for windows can only be compiled in Linux system"
    +        )
    +    compile_spec = {
    +        "debug": True,
    +        "min_block_size": 1,
    +    }
    +    torchtrt.cross_compile_for_windows(
    +        model, file_path=args.path, inputs=inputs, **compile_spec
    +    )
    +    print(
    +        f"model has been successfully cross compiled and saved in Linux to {args.path}"
    +    )
    +
    +
    +

    Total running time of the script: ( 0 minutes 0.000 seconds)

    + +

    Gallery generated by Sphinx-Gallery

    +
    +
    + + +
    + +
    +
    + + + + +
    + + + +
    +

    + © Copyright 2024, NVIDIA Corporation. + +

    +
    + +
    + Built with Sphinx using a theme provided by Read the Docs. +
    + + +
    + +
    +
    + + +
    +
    + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
    +
    +
    +
    +

    Docs

    +

    Access comprehensive developer documentation for PyTorch

    + View Docs +
    + +
    +

    Tutorials

    +

    Get in-depth tutorials for beginners and advanced developers

    + View Tutorials +
    + +
    +

    Resources

    +

    Find development resources and get your questions answered

    + View Resources +
    +
    +
    +
    + + + + + + + + + +
    +
    +
    +
    + + +
    +
    +
    + + +
    + + + + + + + + \ No newline at end of file diff --git a/docs/tutorials/_rendered_examples/dynamo/custom_kernel_plugins.html b/docs/tutorials/_rendered_examples/dynamo/custom_kernel_plugins.html index 6d1d3c5dda..f7fc6b84dc 100644 --- a/docs/tutorials/_rendered_examples/dynamo/custom_kernel_plugins.html +++ b/docs/tutorials/_rendered_examples/dynamo/custom_kernel_plugins.html @@ -10,7 +10,7 @@ - Using Custom Kernels within TensorRT Engines with Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Using Custom Kernels within TensorRT Engines with Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/engine_caching_bert_example.html b/docs/tutorials/_rendered_examples/dynamo/engine_caching_bert_example.html index 11e5fcd607..2281d30f06 100644 --- a/docs/tutorials/_rendered_examples/dynamo/engine_caching_bert_example.html +++ b/docs/tutorials/_rendered_examples/dynamo/engine_caching_bert_example.html @@ -10,7 +10,7 @@ - Engine Caching (BERT) — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Engine Caching (BERT) — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/engine_caching_example.html b/docs/tutorials/_rendered_examples/dynamo/engine_caching_example.html index caaf7f9350..2b3b2d7f93 100644 --- a/docs/tutorials/_rendered_examples/dynamo/engine_caching_example.html +++ b/docs/tutorials/_rendered_examples/dynamo/engine_caching_example.html @@ -10,7 +10,7 @@ - Engine Caching — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Engine Caching — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/index.html b/docs/tutorials/_rendered_examples/dynamo/index.html index 47e90ab73d..398eb2dbd1 100644 --- a/docs/tutorials/_rendered_examples/dynamo/index.html +++ b/docs/tutorials/_rendered_examples/dynamo/index.html @@ -10,7 +10,7 @@ - Dependencies — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Dependencies — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    @@ -488,6 +488,9 @@

    Model Zoo

    Torch Export with Cudagraphs

    Torch Export with Cudagraphs
    +

    +

    sphx_glr_tutorials__rendered_examples_dynamo_cross_runtime_compilation_for_windows.py

    +
    cross runtime compilation limitations:

    Refitting Torch-TensorRT Programs with New Weights

    Refitting Torch-TensorRT Programs with New Weights
    diff --git a/docs/tutorials/_rendered_examples/dynamo/mutable_torchtrt_module_example.html b/docs/tutorials/_rendered_examples/dynamo/mutable_torchtrt_module_example.html index c3b00c04ee..5f86cb8e27 100644 --- a/docs/tutorials/_rendered_examples/dynamo/mutable_torchtrt_module_example.html +++ b/docs/tutorials/_rendered_examples/dynamo/mutable_torchtrt_module_example.html @@ -10,7 +10,7 @@ - Mutable Torch TensorRT Module — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Mutable Torch TensorRT Module — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/refit_engine_example.html b/docs/tutorials/_rendered_examples/dynamo/refit_engine_example.html index 03ec4c07f7..6b9b593b4c 100644 --- a/docs/tutorials/_rendered_examples/dynamo/refit_engine_example.html +++ b/docs/tutorials/_rendered_examples/dynamo/refit_engine_example.html @@ -10,7 +10,7 @@ - Refitting Torch-TensorRT Programs with New Weights — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Refitting Torch-TensorRT Programs with New Weights — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/torch_compile_advanced_usage.html b/docs/tutorials/_rendered_examples/dynamo/torch_compile_advanced_usage.html index b8152240d5..040198395e 100644 --- a/docs/tutorials/_rendered_examples/dynamo/torch_compile_advanced_usage.html +++ b/docs/tutorials/_rendered_examples/dynamo/torch_compile_advanced_usage.html @@ -10,7 +10,7 @@ - Torch Compile Advanced Usage — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Torch Compile Advanced Usage — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/torch_compile_resnet_example.html b/docs/tutorials/_rendered_examples/dynamo/torch_compile_resnet_example.html index 3a2dc1a285..a8b5a14a11 100644 --- a/docs/tutorials/_rendered_examples/dynamo/torch_compile_resnet_example.html +++ b/docs/tutorials/_rendered_examples/dynamo/torch_compile_resnet_example.html @@ -10,7 +10,7 @@ - Compiling ResNet with dynamic shapes using the torch.compile backend — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Compiling ResNet with dynamic shapes using the torch.compile backend — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/torch_compile_stable_diffusion.html b/docs/tutorials/_rendered_examples/dynamo/torch_compile_stable_diffusion.html index de2b866292..441f75d7d4 100644 --- a/docs/tutorials/_rendered_examples/dynamo/torch_compile_stable_diffusion.html +++ b/docs/tutorials/_rendered_examples/dynamo/torch_compile_stable_diffusion.html @@ -10,7 +10,7 @@ - Compiling Stable Diffusion model using the torch.compile backend — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Compiling Stable Diffusion model using the torch.compile backend — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/torch_compile_transformers_example.html b/docs/tutorials/_rendered_examples/dynamo/torch_compile_transformers_example.html index 9cd1c16e65..157b723576 100644 --- a/docs/tutorials/_rendered_examples/dynamo/torch_compile_transformers_example.html +++ b/docs/tutorials/_rendered_examples/dynamo/torch_compile_transformers_example.html @@ -10,7 +10,7 @@ - Compiling BERT using the torch.compile backend — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Compiling BERT using the torch.compile backend — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/torch_export_cudagraphs.html b/docs/tutorials/_rendered_examples/dynamo/torch_export_cudagraphs.html index 609fc48c41..0a40531228 100644 --- a/docs/tutorials/_rendered_examples/dynamo/torch_export_cudagraphs.html +++ b/docs/tutorials/_rendered_examples/dynamo/torch_export_cudagraphs.html @@ -10,7 +10,7 @@ - Torch Export with Cudagraphs — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Torch Export with Cudagraphs — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/torch_export_gpt2.html b/docs/tutorials/_rendered_examples/dynamo/torch_export_gpt2.html index 6c766fbc21..e704950ef1 100644 --- a/docs/tutorials/_rendered_examples/dynamo/torch_export_gpt2.html +++ b/docs/tutorials/_rendered_examples/dynamo/torch_export_gpt2.html @@ -10,7 +10,7 @@ - Compiling GPT2 using the dynamo backend — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Compiling GPT2 using the dynamo backend — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/torch_export_llama2.html b/docs/tutorials/_rendered_examples/dynamo/torch_export_llama2.html index 46d3d02703..61259a4bf1 100644 --- a/docs/tutorials/_rendered_examples/dynamo/torch_export_llama2.html +++ b/docs/tutorials/_rendered_examples/dynamo/torch_export_llama2.html @@ -10,7 +10,7 @@ - Compiling Llama2 using the dynamo backend — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Compiling Llama2 using the dynamo backend — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/vgg16_ptq.html b/docs/tutorials/_rendered_examples/dynamo/vgg16_ptq.html index 649decfbf8..8060f3a221 100644 --- a/docs/tutorials/_rendered_examples/dynamo/vgg16_ptq.html +++ b/docs/tutorials/_rendered_examples/dynamo/vgg16_ptq.html @@ -10,7 +10,7 @@ - Deploy Quantized Models using Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Deploy Quantized Models using Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/dynamo/weight_streaming_example.html b/docs/tutorials/_rendered_examples/dynamo/weight_streaming_example.html index dc3e38dca4..c4e3e6b770 100644 --- a/docs/tutorials/_rendered_examples/dynamo/weight_streaming_example.html +++ b/docs/tutorials/_rendered_examples/dynamo/weight_streaming_example.html @@ -10,7 +10,7 @@ - Weight Streaming — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Weight Streaming — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/_rendered_examples/index.html b/docs/tutorials/_rendered_examples/index.html index 53c3dd6f0a..873e9770c0 100644 --- a/docs/tutorials/_rendered_examples/index.html +++ b/docs/tutorials/_rendered_examples/index.html @@ -10,7 +10,7 @@ - Torch-TensorRT Tutorials — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Torch-TensorRT Tutorials — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -273,7 +273,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    @@ -490,6 +490,9 @@

    Model Zoo

    Torch Export with Cudagraphs

    Torch Export with Cudagraphs
    +

    +

    sphx_glr_tutorials__rendered_examples_dynamo_cross_runtime_compilation_for_windows.py

    +
    cross runtime compilation limitations:

    Refitting Torch-TensorRT Programs with New Weights

    Refitting Torch-TensorRT Programs with New Weights
    diff --git a/docs/tutorials/notebooks.html b/docs/tutorials/notebooks.html index ae0b76f2d9..a930a2584f 100644 --- a/docs/tutorials/notebooks.html +++ b/docs/tutorials/notebooks.html @@ -10,7 +10,7 @@ - Legacy notebooks — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Legacy notebooks — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/tutorials/serving_torch_tensorrt_with_triton.html b/docs/tutorials/serving_torch_tensorrt_with_triton.html index fad19dd7c2..a974712825 100644 --- a/docs/tutorials/serving_torch_tensorrt_with_triton.html +++ b/docs/tutorials/serving_torch_tensorrt_with_triton.html @@ -10,7 +10,7 @@ - Serving a Torch-TensorRT model with Triton — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Serving a Torch-TensorRT model with Triton — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/user_guide/dynamic_shapes.html b/docs/user_guide/dynamic_shapes.html index bbbf462f79..4ba862426c 100644 --- a/docs/user_guide/dynamic_shapes.html +++ b/docs/user_guide/dynamic_shapes.html @@ -10,7 +10,7 @@ - Dynamic shapes with Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Dynamic shapes with Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/user_guide/mixed_precision.html b/docs/user_guide/mixed_precision.html index ad94ac87d4..96578d4101 100644 --- a/docs/user_guide/mixed_precision.html +++ b/docs/user_guide/mixed_precision.html @@ -10,7 +10,7 @@ - Compile Mixed Precision models with Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Compile Mixed Precision models with Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/user_guide/runtime.html b/docs/user_guide/runtime.html index eab011e947..a00365166c 100644 --- a/docs/user_guide/runtime.html +++ b/docs/user_guide/runtime.html @@ -10,7 +10,7 @@ - Deploying Torch-TensorRT Programs — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Deploying Torch-TensorRT Programs — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/user_guide/saving_models.html b/docs/user_guide/saving_models.html index d6199a9e15..a584c6aa16 100644 --- a/docs/user_guide/saving_models.html +++ b/docs/user_guide/saving_models.html @@ -10,7 +10,7 @@ - Saving models compiled with Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Saving models compiled with Torch-TensorRT — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/user_guide/torch_tensorrt_explained.html b/docs/user_guide/torch_tensorrt_explained.html index c4cd64970d..4514965c06 100644 --- a/docs/user_guide/torch_tensorrt_explained.html +++ b/docs/user_guide/torch_tensorrt_explained.html @@ -10,7 +10,7 @@ - Torch-TensorRT Explained — Torch-TensorRT v2.6.0.dev0+e43833d documentation + Torch-TensorRT Explained — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/docs/user_guide/using_dla.html b/docs/user_guide/using_dla.html index 03ce69f629..5f4aff74bb 100644 --- a/docs/user_guide/using_dla.html +++ b/docs/user_guide/using_dla.html @@ -10,7 +10,7 @@ - DLA — Torch-TensorRT v2.6.0.dev0+e43833d documentation + DLA — Torch-TensorRT v2.6.0.dev0+bc95015 documentation @@ -275,7 +275,7 @@
    - v2.6.0.dev0+e43833d + v2.6.0.dev0+bc95015
    diff --git a/examples/dynamo/cross_runtime_compilation_for_windows.py b/examples/dynamo/cross_runtime_compilation_for_windows.py new file mode 100644 index 0000000000..184470ffa0 --- /dev/null +++ b/examples/dynamo/cross_runtime_compilation_for_windows.py @@ -0,0 +1,82 @@ +""" +.. _resnet_cross_runtime_compilation_for_windows_example: + +cross runtime compilation limitations: +The cross compile and saved model can only be loaded in Windows, it can no longer be loaded in Linux +The cross compile and saved model can only be loaded in the same Compute Capability as the Linux which it was cross compiled +(for example, if the model was cross compiled in Linux with GeForceRTX 4080 which has Compute Capability of 8.9, +It cannot be loaded in Windows with GeForceRTX 3080 which has Compute Capability of 8.6) + +Cross runtime compilation for windows example +====================================================== + +Compile and save the Resnet Model using Torch-TensorRT in Linux: + +python examples/dynamo/cross_runtime_compilation_for_windows.py --path trt_resnet.ep + +Load the Resnet Model saved in Windows: + +python examples/dynamo/cross_runtime_compilation_for_windows.py --path trt_resnet.ep --load True + +""" + +# %% +# Imports and Model Definition +# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +import argparse +import platform + +import torch +import torch_tensorrt as torchtrt +import torchvision.models as models + +PARSER = argparse.ArgumentParser( + description="Cross runtime comilation for windows example: Resnet Model" +) +PARSER.add_argument( + "--load", default=False, type=bool, required=False, help="Load the model in Windows" +) +PARSER.add_argument( + "--path", + type=str, + required=True, + help="Path to the saved model file", +) + +args = PARSER.parse_args() +torch.manual_seed(0) +model = models.resnet18().eval().cuda() +input = torch.rand((1, 3, 224, 224)).to("cuda") +inputs = [input] + +# %% +# According to the argument, it is either cross compile and save resnet model for windows in Linux +# or load the saved resnet model in Windows +# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +if args.load: + # load the saved model in Windows + if platform.system() != "Windows" or platform.machine() != "AMD64": + raise ValueError( + "cross runtime compiled model for windows can only be loaded in Windows system" + ) + loaded_model = torchtrt.load_cross_compiled_exported_program(args.path).module() + print(f"model has been successfully loaded from ${args.path}") + # inference + trt_output = loaded_model(input) + print(f"inference result: {trt_output}") +else: + if platform.system() != "Linux" or platform.architecture()[0] != "64bit": + raise ValueError( + "cross runtime compiled model for windows can only be compiled in Linux system" + ) + compile_spec = { + "debug": True, + "min_block_size": 1, + } + torchtrt.cross_compile_for_windows( + model, file_path=args.path, inputs=inputs, **compile_spec + ) + print( + f"model has been successfully cross compiled and saved in Linux to {args.path}" + ) diff --git a/py/torch_tensorrt/_compile.py b/py/torch_tensorrt/_compile.py index 66cff1f9ea..eb2642755a 100644 --- a/py/torch_tensorrt/_compile.py +++ b/py/torch_tensorrt/_compile.py @@ -2,6 +2,7 @@ import collections.abc import logging +import platform from enum import Enum from typing import Any, Callable, List, Optional, Sequence, Set @@ -29,11 +30,27 @@ from torch_tensorrt.dynamo._compiler import ( convert_exported_program_to_serialized_trt_engine as dynamo_convert_exported_program_to_serialized_trt_engine, ) + from torch_tensorrt.dynamo._compiler import ( + cross_compile_for_windows as dynamo_cross_compile_for_windows, + ) + from torch_tensorrt.dynamo._compiler import ( + load_cross_compiled_exported_program as dynamo_load_cross_compiled_exported_program, + ) + from torch_tensorrt.dynamo._compiler import ( + save_cross_compiled_exported_program as dynamo_save_cross_compiled_exported_program, + ) from torch_tensorrt.dynamo._tracer import trace as dynamo_trace logger = logging.getLogger(__name__) -__all__ = ["compile", "convert_method_to_trt_engine", "save", "load"] +__all__ = [ + "compile", + "cross_compile_for_windows", + "load_cross_compiled_exported_program", + "convert_method_to_trt_engine", + "save", + "load", +] def _non_fx_input_interface( @@ -281,6 +298,105 @@ def compile( raise RuntimeError("Module is an unknown format or the ir requested is unknown") +def cross_compile_for_windows( + module: torch.nn.Module, + file_path: str, + inputs: Optional[Sequence[Input | torch.Tensor]] = None, + arg_inputs: Optional[Sequence[Sequence[Any]]] = None, + kwarg_inputs: Optional[dict[Any, Any]] = None, + enabled_precisions: Optional[Set[torch.dtype | dtype]] = None, + **kwargs: Any, +) -> None: + """Compile a PyTorch module using TensorRT in Linux for Inference in Windows + + Takes an existing PyTorch module and a set of settings to configure the compiler + and it will convert methods to AOT graphs which call equivalent TensorRT serialized + engine info into the disk in the specified file_path user provided. + It will then allow user to load the deserialized model from the disk in Windows. + Note: the model cross compiled for windows in Linux environmen can only be loaded + in Windows. + + Argument: + module (torch.nn.Module): Source module + file_path (str): the file path to store the serialized module into the disk + + Keyword Arguments: + inputs (List[Union(torch_tensorrt.Input, torch.Tensor)]): **Required** List of specifications of input shape, dtype and memory layout for inputs to the module. This argument is required. Input Sizes can be specified as torch sizes, tuples or lists. dtypes can be specified using + torch datatypes or torch_tensorrt datatypes and you can use either torch devices or the torch_tensorrt device type enum + to select device type. :: + + inputs=[ + torch_tensorrt.Input((1, 3, 224, 224)), # Static NCHW input shape for input #1 + torch_tensorrt.Input( + min_shape=(1, 224, 224, 3), + opt_shape=(1, 512, 512, 3), + max_shape=(1, 1024, 1024, 3), + dtype=torch.int32 + format=torch.channel_last + ), # Dynamic input shape for input #2 + torch.randn((1, 3, 224, 244)) # Use an example tensor and let torch_tensorrt infer settings + ] + arg_inputs (Tuple[Any, ...]): Same as inputs. Alias for better understanding with kwarg_inputs. + kwarg_inputs (dict[Any, ...]): Optional, kwarg inputs to the module forward function. + enabled_precision (Set(Union(torch.dtype, torch_tensorrt.dtype))): The set of datatypes that TensorRT can use when selecting kernels + **kwargs: Additional settings for the specific requested strategy (See submodules for more info) + + """ + + if platform.system() != "Linux" or platform.architecture()[0] != "64bit": + raise RuntimeError( + f"Cross compile for windows is only supported on x86-64 Linux architecture, current platform: {platform.system()=}, {platform.architecture()[0]=}" + ) + + if not file_path: + raise ValueError("File path cannot be empty. Please provide a valid file path") + + enabled_precisions_set: Set[dtype | torch.dtype] = ( + enabled_precisions + if enabled_precisions is not None + else _defaults.ENABLED_PRECISIONS + ) + + # Prepare torch and torchtrt inputs + if not arg_inputs and not inputs: + raise AssertionError("'arg_inputs' and 'inputs' should not both be None.") + + elif arg_inputs and inputs: + raise AssertionError( + "'arg_inputs' and 'inputs' should not be used at the same time." + ) + + arg_inputs = inputs or arg_inputs + + if kwarg_inputs is None: + kwarg_inputs = {} + + from torch_tensorrt.dynamo.utils import prepare_inputs + + if not isinstance(arg_inputs, collections.abc.Sequence): + arg_inputs = [arg_inputs] # type: ignore + + # Export the module + torchtrt_arg_inputs = prepare_inputs(arg_inputs) + torchtrt_kwarg_inputs = prepare_inputs(kwarg_inputs) + + exp_program = dynamo_trace( + module, torchtrt_arg_inputs, kwarg_inputs=torchtrt_kwarg_inputs, **kwargs + ) + logger.debug("successfully exported the module") + + # Compile and save the module + trt_gm = dynamo_cross_compile_for_windows( + exp_program, + arg_inputs=torchtrt_arg_inputs, + enabled_precisions=enabled_precisions_set, + **kwargs, + ) + + dynamo_save_cross_compiled_exported_program(trt_gm, file_path) + logger.debug("successfully compiled and saved the module for windows") + + def torch_compile(module: torch.nn.Module, **kwargs: Any) -> Any: """ Returns a boxed model which is the output of torch.compile. @@ -406,6 +522,19 @@ def convert_method_to_trt_engine( raise RuntimeError("Module is an unknown format or the ir requested is unknown") +def load_cross_compiled_exported_program(file_path: str = "") -> Any: + """ + Load an ExportedProgram file in Windows which was previously cross compiled in Linux + + Arguments: + file_path (str): Path to file on the disk + + Raises: + ValueError: If the api is not called in windows or there is no file or the file is not a valid ExportedProgram file + """ + return dynamo_load_cross_compiled_exported_program(file_path) + + def load(file_path: str = "") -> Any: """ Load either a Torchscript model or ExportedProgram. diff --git a/py/torch_tensorrt/dynamo/__init__.py b/py/torch_tensorrt/dynamo/__init__.py index 79bd113ab8..6fabdad633 100644 --- a/py/torch_tensorrt/dynamo/__init__.py +++ b/py/torch_tensorrt/dynamo/__init__.py @@ -7,7 +7,13 @@ logger = logging.getLogger(__name__) if version.parse(sanitized_torch_version()) >= version.parse("2.1.dev"): - from ._compiler import compile, convert_exported_program_to_serialized_trt_engine + from ._compiler import ( + compile, + convert_exported_program_to_serialized_trt_engine, + cross_compile_for_windows, + load_cross_compiled_exported_program, + save_cross_compiled_exported_program, + ) from ._exporter import export from ._refit import refit_module_weights from ._settings import CompilationSettings diff --git a/py/torch_tensorrt/dynamo/_compiler.py b/py/torch_tensorrt/dynamo/_compiler.py index 5e41abe473..730d47a254 100644 --- a/py/torch_tensorrt/dynamo/_compiler.py +++ b/py/torch_tensorrt/dynamo/_compiler.py @@ -2,6 +2,7 @@ import collections.abc import logging +import platform import warnings from typing import Any, Collection, List, Optional, Sequence, Set, Tuple, Union @@ -19,6 +20,7 @@ parse_non_trt_nodes, ) from torch_tensorrt.dynamo._engine_cache import BaseEngineCache, DiskEngineCache +from torch_tensorrt.dynamo._exporter import replace_execute_engine_no_op_node from torch_tensorrt.dynamo.conversion import ( CompilationSettings, UnsupportedOperatorException, @@ -47,6 +49,285 @@ logger = logging.getLogger(__name__) +def cross_compile_for_windows( + exported_program: ExportedProgram, + inputs: Optional[Sequence[Sequence[Any]]] = None, + *, + arg_inputs: Optional[Sequence[Sequence[Any]]] = None, + kwarg_inputs: Optional[dict[Any, Any]] = None, + device: Optional[Union[Device, torch.device, str]] = _defaults.DEVICE, + disable_tf32: bool = _defaults.DISABLE_TF32, + assume_dynamic_shape_support: bool = _defaults.ASSUME_DYNAMIC_SHAPE_SUPPORT, + sparse_weights: bool = _defaults.SPARSE_WEIGHTS, + enabled_precisions: Union[ + Set[Union[torch.dtype, dtype]], Tuple[Union[torch.dtype, dtype]] + ] = _defaults.ENABLED_PRECISIONS, + engine_capability: EngineCapability = _defaults.ENGINE_CAPABILITY, + make_refittable: bool = _defaults.MAKE_REFITTABLE, + debug: bool = _defaults.DEBUG, + num_avg_timing_iters: int = _defaults.NUM_AVG_TIMING_ITERS, + workspace_size: int = _defaults.WORKSPACE_SIZE, + dla_sram_size: int = _defaults.DLA_SRAM_SIZE, + dla_local_dram_size: int = _defaults.DLA_LOCAL_DRAM_SIZE, + dla_global_dram_size: int = _defaults.DLA_GLOBAL_DRAM_SIZE, + truncate_double: bool = _defaults.TRUNCATE_DOUBLE, + require_full_compilation: bool = _defaults.REQUIRE_FULL_COMPILATION, + min_block_size: int = _defaults.MIN_BLOCK_SIZE, + torch_executed_ops: Optional[Collection[Target]] = None, + torch_executed_modules: Optional[List[str]] = None, + pass_through_build_failures: bool = _defaults.PASS_THROUGH_BUILD_FAILURES, + max_aux_streams: Optional[int] = _defaults.MAX_AUX_STREAMS, + version_compatible: bool = _defaults.VERSION_COMPATIBLE, + optimization_level: Optional[int] = _defaults.OPTIMIZATION_LEVEL, + use_python_runtime: bool = _defaults.USE_PYTHON_RUNTIME, + use_fast_partitioner: bool = _defaults.USE_FAST_PARTITIONER, + enable_experimental_decompositions: bool = _defaults.ENABLE_EXPERIMENTAL_DECOMPOSITIONS, + dryrun: bool = _defaults.DRYRUN, + hardware_compatible: bool = _defaults.HARDWARE_COMPATIBLE, + timing_cache_path: str = _defaults.TIMING_CACHE_PATH, + lazy_engine_init: bool = _defaults.LAZY_ENGINE_INIT, + cache_built_engines: bool = _defaults.CACHE_BUILT_ENGINES, + reuse_cached_engines: bool = _defaults.REUSE_CACHED_ENGINES, + engine_cache_dir: str = _defaults.ENGINE_CACHE_DIR, + engine_cache_size: int = _defaults.ENGINE_CACHE_SIZE, + custom_engine_cache: Optional[BaseEngineCache] = _defaults.CUSTOM_ENGINE_CACHE, + use_explicit_typing: bool = _defaults.USE_EXPLICIT_TYPING, + use_fp32_acc: bool = _defaults.USE_FP32_ACC, + enable_weight_streaming: bool = _defaults.ENABLE_WEIGHT_STREAMING, + **kwargs: Any, +) -> torch.fx.GraphModule: + """Compile an ExportedProgram module using TensorRT in Linux for Inference in Windows + + Takes an exported program and a set of settings to configure the compiler + and it will convert methods to AOT graphs which call equivalent TensorRT engines + + Arguments: + exported_program (torch.export.ExportedProgram): Source module, running torch.export on a ``torch.nn.Module`` + inputs (Tuple[Any, ...]): List of specifications of input shape, dtype and memory layout for inputs to the module. This argument is required. Input Sizes can be specified as torch sizes, tuples or lists. dtypes can be specified using + torch datatypes or torch_tensorrt datatypes and you can use either torch devices or the torch_tensorrt device type enum + to select device type. + + .. code-block:: py + + inputs=[ + torch_tensorrt.Input((1, 3, 224, 224)), # Static NCHW input shape for input #1 + torch_tensorrt.Input( + min_shape=(1, 224, 224, 3), + opt_shape=(1, 512, 512, 3), + max_shape=(1, 1024, 1024, 3), + dtype=torch.int32 + format=torch.channel_last + ), # Dynamic input shape for input #2 + torch.randn((1, 3, 224, 244)) # Use an example tensor and let torch_tensorrt infer settings + ] + + Keyword Arguments: + arg_inputs (Tuple[Any, ...]): Same as inputs. Alias for better understanding with kwarg_inputs. + kwarg_inputs (dict[Any, ...]): Optional, kwarg inputs to the module forward function. + device (Union(torch_tensorrt.Device, torch.device, dict)): Target device for TensorRT engines to run on :: + + device=torch_tensorrt.Device("dla:1", allow_gpu_fallback=True) + + disable_tf32 (bool): Force FP32 layers to use traditional as FP32 format vs the default behavior of rounding the inputs to 10-bit mantissas before multiplying, but accumulates the sum using 23-bit mantissas + assume_dynamic_shape_support (bool): Setting this to true enables the converters work for both dynamic and static shapes. Default: False + sparse_weights (bool): Enable sparsity for convolution and fully connected layers. + enabled_precision (Set(Union(torch.dtype, torch_tensorrt.dtype))): The set of datatypes that TensorRT can use when selecting kernels + refit (bool): Enable refitting + debug (bool): Enable debuggable engine + capability (torch_tensorrt.EngineCapability): Restrict kernel selection to safe gpu kernels or safe dla kernels + num_avg_timing_iters (int): Number of averaging timing iterations used to select kernels + workspace_size (int): Maximum size of workspace given to TensorRT + dla_sram_size (int): Fast software managed RAM used by DLA to communicate within a layer. + dla_local_dram_size (int): Host RAM used by DLA to share intermediate tensor data across operations + dla_global_dram_size (int): Host RAM used by DLA to store weights and metadata for execution + truncate_double (bool): Truncate weights provided in double (float64) to float32 + calibrator (Union(torch_tensorrt._C.IInt8Calibrator, tensorrt.IInt8Calibrator)): Calibrator object which will provide data to the PTQ system for INT8 Calibration + require_full_compilation (bool): Require modules to be compiled end to end or return an error as opposed to returning a hybrid graph where operations that cannot be run in TensorRT are run in PyTorch + min_block_size (int): The minimum number of contiguous TensorRT convertible operations in order to run a set of operations in TensorRT + torch_executed_ops (Collection[Target]): Set of aten operators that must be run in PyTorch. An error will be thrown if this set is not empty but ``require_full_compilation`` is True + torch_executed_modules (List[str]): List of modules that must be run in PyTorch. An error will be thrown if this list is not empty but ``require_full_compilation`` is True + pass_through_build_failures (bool): Error out if there are issues during compilation (only applicable to torch.compile workflows) + max_aux_stream (Optional[int]): Maximum streams in the engine + version_compatible (bool): Build the TensorRT engines compatible with future versions of TensorRT (Restrict to lean runtime operators to provide version forward compatibility for the engines) + optimization_level: (Optional[int]): Setting a higher optimization level allows TensorRT to spend longer engine building time searching for more optimization options. The resulting engine may have better performance compared to an engine built with a lower optimization level. The default optimization level is 3. Valid values include integers from 0 to the maximum optimization level, which is currently 5. Setting it to be greater than the maximum level results in identical behavior to the maximum level. + use_python_runtime: (bool): Return a graph using a pure Python runtime, reduces options for serialization + use_fast_partitioner: (bool): Use the adjacency based partitioning scheme instead of the global partitioner. Adjacency partitioning is faster but may not be optimal. Use the global paritioner (``False``) if looking for best performance + enable_experimental_decompositions (bool): Use the full set of operator decompositions. These decompositions may not be tested but serve to make the graph easier to convert to TensorRT, potentially increasing the amount of graphs run in TensorRT. + dryrun (bool): Toggle for "Dryrun" mode, running everything except conversion to TRT and logging outputs + hardware_compatible (bool): Build the TensorRT engines compatible with GPU architectures other than that of the GPU on which the engine was built (currently works for NVIDIA Ampere and newer) + timing_cache_path (str): Path to the timing cache if it exists (or) where it will be saved after compilation + lazy_engine_init (bool): Defer setting up engines until the compilation of all engines is complete. Can allow larger models with multiple graph breaks to compile but can lead to oversubscription of GPU memory at runtime. + cache_built_engines (bool): Whether to save the compiled TRT engines to storage + reuse_cached_engines (bool): Whether to load the compiled TRT engines from storage + engine_cache_dir (Optional[str]): Directory to store the cached TRT engines + engine_cache_size (Optional[int]): Maximum hard-disk space (bytes) to use for the engine cache, default is 1GB. If the cache exceeds this size, the oldest engines will be removed by default + custom_engine_cache (Optional[BaseEngineCache]): Engine cache instance to use for saving and loading engines. Users can provide their own engine cache by inheriting from BaseEngineCache. If used, engine_cache_dir and engine_cache_size will be ignored. + use_explicit_typing (bool): This flag enables strong typing in TensorRT compilation which respects the precisions set in the Pytorch model. This is useful when users have mixed precision graphs. + use_fp32_acc (bool): This option inserts cast to FP32 nodes around matmul layers and TensorRT ensures the accumulation of matmul happens in FP32. Use this only when FP16 precision is configured in enabled_precisions. + enable_weight_streaming (bool): Enable weight streaming. + **kwargs: Any, + Returns: + torch.fx.GraphModule: Compiled FX Module, when run it will execute via TensorRT + + """ + if platform.system() != "Linux" or platform.architecture()[0] != "64bit": + raise RuntimeError( + f"Cross compile for windows is only supported on x86-64 Linux architecture, current platform: {platform.system()=}, {platform.architecture()[0]=}" + ) + + if debug: + set_log_level(logger.parent, logging.DEBUG) + + if "truncate_long_and_double" in kwargs.keys(): + if truncate_double is not _defaults.TRUNCATE_DOUBLE: + raise ValueError( + 'Provided configuration for "truncate_double" and deprecated API "truncate_long_and_double", please only use "truncate_double"' + ) + else: + truncate_double = kwargs["truncate_long_and_double"] + warnings.warn( + 'Compiler option "truncate_long_and_double" is deprecated in favor of "truncate_double" as int64 is now natively supported, this option will be removed in the next version', + DeprecationWarning, + stacklevel=2, + ) + + if "refit" in kwargs.keys(): + warnings.warn( + "Refit is deprecated. Please use make_refittable=True if you want to enable refitting of the engine.", + DeprecationWarning, + stacklevel=2, + ) + if make_refittable: + raise ValueError("Use flag make_refittable only. Flag refit is deprecated.") + else: + make_refittable = kwargs["refit"] + + engine_capability = EngineCapability._from(engine_capability) + + if torch_executed_modules is not None and torch_executed_modules: + logger.warning( + f"Detected torch_executed_modules was non-empty: {torch_executed_modules}" + "\nThis feature is unimplemented in Torch-TRT Dynamo currently." + ) + + if use_explicit_typing: + if len(enabled_precisions) != 1 or not any( + x in enabled_precisions for x in {torch.float32, dtype.f32} + ): + raise AssertionError( + f"When use_explicit_typing is enabled, only torch.float32 is allowed in the enabled_precisions but found {enabled_precisions}" + ) + + if use_fp32_acc: + logger.debug( + "FP32 accumulation for matmul layers is enabled. This option should only be enabled if the model already has FP16 weights and has no effect if it has FP32 weights. \ + This flag inserts casts around matmul layers and ensures TensorRT executes the matmul layers in FP16 with FP32 accumulation." + ) + + if enable_weight_streaming and not use_explicit_typing: + raise AssertionError( + "When enable_weight_streaming is enabled, it requires use_explicit_typing to be set to True" + ) + # Aliasing inputs to arg_inputs for better understanding + if not arg_inputs and not inputs: + raise AssertionError("'arg_inputs' and 'inputs' should not both be None.") + + elif arg_inputs and inputs: + raise AssertionError( + "'arg_inputs' and 'inputs' should not be used at the same time." + ) + + arg_inputs = inputs or arg_inputs + + if kwarg_inputs is None: + kwarg_inputs = {} + + if not isinstance(arg_inputs, collections.abc.Sequence): + arg_inputs = [arg_inputs] # type: ignore + + # Prepare torch_trt inputs + trt_arg_inputs: Sequence[Input] = prepare_inputs(arg_inputs) + trt_kwarg_inputs: Optional[dict[Any, Any]] = prepare_inputs(kwarg_inputs) + device = to_torch_tensorrt_device(device) + enabled_precisions = {dtype._from(p) for p in enabled_precisions} + + compilation_options = { + "enabled_precisions": ( + enabled_precisions if enabled_precisions else _defaults.ENABLED_PRECISIONS + ), + "debug": debug, + "device": device, + "assume_dynamic_shape_support": assume_dynamic_shape_support, + "workspace_size": workspace_size, + "min_block_size": min_block_size, + "torch_executed_ops": ( + torch_executed_ops if torch_executed_ops is not None else set() + ), + "pass_through_build_failures": pass_through_build_failures, + "max_aux_streams": max_aux_streams, + "version_compatible": version_compatible, + "optimization_level": optimization_level, + "use_python_runtime": False, + "truncate_double": truncate_double, + "use_fast_partitioner": use_fast_partitioner, + "num_avg_timing_iters": num_avg_timing_iters, + "enable_experimental_decompositions": enable_experimental_decompositions, + "require_full_compilation": require_full_compilation, + "disable_tf32": disable_tf32, + "sparse_weights": sparse_weights, + "make_refittable": make_refittable, + "engine_capability": engine_capability, + "dla_sram_size": dla_sram_size, + "dla_local_dram_size": dla_local_dram_size, + "dla_global_dram_size": dla_global_dram_size, + "dryrun": dryrun, + "hardware_compatible": hardware_compatible, + "timing_cache_path": timing_cache_path, + "lazy_engine_init": lazy_engine_init, + "cache_built_engines": cache_built_engines, + "reuse_cached_engines": reuse_cached_engines, + "enable_cross_compile_for_windows": True, + "enable_weight_streaming": enable_weight_streaming, + } + + # disable the following settings is not supported for cross compilation for windows feature + unsupported_settings = ( + "use_python_runtime", + "lazy_engine_init", + "cache_built_engines", + "reuse_cached_engines", + ) + # disable these settings if anything is turned on + for key, value in compilation_options.items(): + if key in unsupported_settings and value: + compilation_options[key] = False + logger.warning( + f"arg: {key} is not supported for cross compilation for windows feature, hence it is disabled." + ) + + settings = CompilationSettings(**compilation_options) + logger.info("Compilation Settings: %s\n", settings) + exported_program = pre_export_lowering(exported_program, settings) + exported_program = exported_program.run_decompositions( + get_decompositions(enable_experimental_decompositions) + ) + + gm = exported_program.module() + logger.debug("Input graph: " + str(gm.graph)) + + # Apply lowering on the graph module + gm = post_lowering(gm, settings) + logger.debug("Lowered Input graph: " + str(gm.graph)) + + trt_gm = compile_module( + gm, + trt_arg_inputs, + trt_kwarg_inputs, + settings, + ) + return trt_gm + + def compile( exported_program: ExportedProgram, inputs: Optional[Sequence[Sequence[Any]]] = None, @@ -203,6 +484,14 @@ def compile( stacklevel=2, ) + if ( + "enable_cross_compile_for_windows" in kwargs.keys() + and kwargs["enable_cross_compile_for_windows"] + ): + raise ValueError( + "Please use cross_compile_for_windows() api if you want to cross compile the module in Linux for inferencing in Windows." + ) + engine_capability = EngineCapability._from(engine_capability) if torch_executed_modules is not None and torch_executed_modules: @@ -304,6 +593,7 @@ def compile( "refit_identical_engine_weights": refit_identical_engine_weights, "strip_engine_weights": strip_engine_weights, "immutable_weights": immutable_weights, + "enable_cross_compile_for_windows": False, "enable_weight_streaming": enable_weight_streaming, } @@ -539,7 +829,7 @@ def contains_metadata(gm: torch.fx.GraphModule) -> bool: # Replace all FX Modules with TRT Modules for name, trt_module in trt_modules.items(): setattr(partitioned_module, name, trt_module) - if settings.lazy_engine_init: + if settings.lazy_engine_init and not settings.enable_cross_compile_for_windows: getattr(partitioned_module, name).setup_engine() # Reset settings object to user specification after fallback to global partitioning mode @@ -780,3 +1070,58 @@ def convert_exported_program_to_serialized_trt_engine( serialized_engine: bytes = interpreter_result.serialized_engine return serialized_engine + + +def save_cross_compiled_exported_program( + gm: torch.fx.GraphModule, + file_path: str, +) -> None: + """ + Save cross compiled exported program to disk. + + Arguments: + module (torch.fx.GraphModule): Cross compiled Torch-TensorRT module + file_path (str): the file path where the exported program will be saved to disk + """ + if not file_path: + raise ValueError("File path cannot be empty. Please provide a valid file path") + + from torch_tensorrt.dynamo._exporter import export + + exp_program = export(gm, cross_compile_flag=True) + torch.export.save(exp_program, file_path) + logger.debug(f"successfully saved the module for windows at {file_path}") + + +def load_cross_compiled_exported_program(file_path: str = "") -> Any: + """ + Load an ExportedProgram file in Windows which was previously cross compiled in Linux + + Arguments: + file_path (str): Path to file on the disk + + Raises: + ValueError: If the api is not called in windows or there is no file or the file is a valid ExportedProgram file + """ + if not file_path: + raise ValueError("File path cannot be empty. Please provide a valid file path") + + if platform.system() != "Windows" or platform.machine() != "AMD64": + raise ValueError( + "cross runtime compiled model for windows can only be loaded in Windows system" + ) + + try: + logger.debug(f"Loading the provided file {file_path} using torch.export.load()") + # TODO: think about how to handle the torch.jit.load route? + exp_program = torch.export.load(file_path) + except Exception as e: + logger.info( + f"Loading the provided file {file_path} via torch.export.load() failed with the following error: {e}", + exc_info=True, + ) + raise ValueError( + f"cross_load the file {file_path} doesn't correspond to a valid ExportedProgram. Please verify the file path." + ) + + return replace_execute_engine_no_op_node(exp_program) diff --git a/py/torch_tensorrt/dynamo/_defaults.py b/py/torch_tensorrt/dynamo/_defaults.py index ee29e95b72..1341ca739f 100644 --- a/py/torch_tensorrt/dynamo/_defaults.py +++ b/py/torch_tensorrt/dynamo/_defaults.py @@ -45,6 +45,7 @@ STRIP_ENGINE_WEIGHTS = False IMMUTABLE_WEIGHTS = False ENABLE_WEIGHT_STREAMING = False +ENABLE_CROSS_COMPILE_FOR_WINDOWS = False def default_device() -> Device: diff --git a/py/torch_tensorrt/dynamo/_exporter.py b/py/torch_tensorrt/dynamo/_exporter.py index ae7c09caf8..c7a063d675 100644 --- a/py/torch_tensorrt/dynamo/_exporter.py +++ b/py/torch_tensorrt/dynamo/_exporter.py @@ -1,6 +1,7 @@ +import base64 import copy import operator -from typing import Any, Dict, Sequence, Tuple, cast +from typing import Any, Dict, Optional, Sequence, Tuple, cast import torch from torch._guards import detect_fake_mode @@ -16,24 +17,28 @@ OutputSpec, TensorArgument, ) +from torch_tensorrt.dynamo.runtime._TorchTensorRTModule import ENGINE_IDX, NAME_IDX def export( gm: torch.fx.GraphModule, + cross_compile_flag: Optional[bool] = False, ) -> ExportedProgram: """Export the result of TensorRT compilation into the desired output format. Arguments: gm (torch.fx.GraphModule): Compiled Torch-TensorRT module, generated by ``torch_tensorrt.dynamo.compile`` inputs (torch.Tensor): Torch input tensors + cross_compile_flag (bool): Flag to indicated whether it is cross_compilation enabled or not """ - patched_module = transform(gm) + patched_module = transform(gm, cross_compile_flag) exp_program = create_trt_exp_program(patched_module) return exp_program def transform( gm: torch.fx.GraphModule, + cross_compile_flag: Optional[bool] = False, ) -> torch.fx.GraphModule: """ Transforms the graphmodule by inlining Pytorch and TensorRT submodules. @@ -43,6 +48,7 @@ def transform( Arguments: gm (torch.fx.GraphModule): Compiled Torch-TensorRT module, generated by ``torch_tensorrt.dynamo.compile`` inputs (torch.Tensor): Torch input tensors + cross_compile_flag (bool): Flag to indicated whether it is cross_compilation enabled or not Returns an inlined torch.fx.GraphModule """ @@ -51,7 +57,7 @@ def transform( gm = copy.deepcopy(gm) # Inline TensorRT submodules - inline_trt_modules(gm) + inline_trt_modules(gm, cross_compile_flag) # Inline pytorch submodules inline_torch_modules(gm) @@ -350,7 +356,9 @@ def create_trt_exp_program( return trt_exp_program -def inline_trt_modules(gm: torch.fx.GraphModule) -> torch.fx.GraphModule: +def inline_trt_modules( + gm: torch.fx.GraphModule, cross_compile_flag: Optional[bool] = False +) -> torch.fx.GraphModule: """ Replace TRT submodules with trt engine nodes. """ @@ -373,25 +381,36 @@ def inline_trt_modules(gm: torch.fx.GraphModule) -> torch.fx.GraphModule: num_outputs = len(trt_module_node.meta["val"]) # Insert a call_function node to perform inference on TRT engine with gm.graph.inserting_before(trt_module_node): - engine_name = f"{name}_engine" - setattr(gm, engine_name, trt_module.engine) - engine_node = gm.graph.get_attr(engine_name) - - trt_node = gm.graph.call_function( - torch.ops.tensorrt.execute_engine.default, - (trt_module_node.args, engine_node), - ) + if not cross_compile_flag: + # for the normal workflow: use the execute_engine node + engine_name = f"{name}_engine" + setattr(gm, engine_name, trt_module.engine) + engine_node = gm.graph.get_attr(engine_name) + + trt_node = gm.graph.call_function( + torch.ops.tensorrt.execute_engine.default, + (trt_module_node.args, engine_node), + ) + # meta["val"] should be a lighter version of a tensor. For eg: it should be a FakeTensor (with output shape and dtype properties) + # Lighter version of a custom_obj is not defined clearly. meta["val"] does not have any type expectations but + # for custom object nodes, it should be CustomObjArgument + engine_node.meta["val"] = CustomObjArgument( + name=engine_node.name, class_fqn="" + ) + else: + # for the cross compile for windows workflow: use the no_op_placeholder node + engine_info = trt_module._pack_engine_info() + engine_bytes = engine_info[ENGINE_IDX] + engine_info[ENGINE_IDX] = base64.b64encode(engine_bytes).decode("utf-8") + # insert the no_placeholder node in the graph which should be replaced to the actual execute_engine node while load in the windows + trt_node = gm.graph.call_function( + torch.ops.tensorrt.no_op_placeholder_for_execute_engine.default, + (trt_module_node.args, *engine_info), + ) # set trt_node.meta with trt_module_node.meta assert num_outputs > 0 trt_node.meta["val"] = trt_module_node.meta["val"] - # meta["val"] should be a lighter version of a tensor. For eg: it should be a FakeTensor (with output shape and dtype properties) - # Lighter version of a custom_obj is not defined clearly. meta["val"] does not have any type expectations but - # for custom object nodes, it should be CustomObjArgument - engine_node.meta["val"] = CustomObjArgument( - name=engine_node.name, class_fqn="" - ) - if num_outputs == 1: # Insert getitem nodes as outputs (for export serialization to work) with gm.graph.inserting_after(trt_node): @@ -411,3 +430,57 @@ def inline_trt_modules(gm: torch.fx.GraphModule) -> torch.fx.GraphModule: gm.graph.erase_node(trt_module_node) return gm + + +def replace_execute_engine_no_op_node( + exp_program: ExportedProgram, +) -> ExportedProgram: + gm = exp_program.graph_module + no_op_placeholder_nodes = [] + for node in gm.graph.nodes: + if "no_op_placeholder_for_execute_engine" in node.name: + no_op_placeholder_nodes.append(node) + assert len(no_op_placeholder_nodes) > 0 + for no_op_placeholder_node in no_op_placeholder_nodes: + if "val" not in no_op_placeholder_node.meta: + raise ValueError(f"metadata info is missing for the node: {node.name}") + with gm.graph.inserting_before(no_op_placeholder_node): + packed_engine_info = list(no_op_placeholder_node.args[1:]) + engine_bytes = packed_engine_info[ENGINE_IDX] + engine_name = packed_engine_info[NAME_IDX] + + packed_engine_info[ENGINE_IDX] = base64.b64decode( + engine_bytes.encode("utf-8") + ) + trt_engine = torch.classes.tensorrt.Engine(tuple(packed_engine_info)) + setattr(gm, engine_name, trt_engine) + engine_node = gm.graph.get_attr(engine_name) + + trt_node = gm.graph.call_function( + torch.ops.tensorrt.execute_engine.default, + (no_op_placeholder_node.args[0], engine_node), + ) + trt_node.meta["val"] = no_op_placeholder_node.meta["val"] + engine_node.meta["val"] = CustomObjArgument( + name=engine_node.name, class_fqn="" + ) + + if len(no_op_placeholder_node.meta["val"]) == 1: + with gm.graph.inserting_after(trt_node): + getitem_output = gm.graph.call_function(operator.getitem, (trt_node, 0)) + getitem_output.meta["val"] = trt_node.meta["val"] + no_op_placeholder_node.replace_all_uses_with(getitem_output) + else: + no_op_placeholder_node.replace_all_uses_with(trt_node) + getitem_nodes = trt_node.users + for idx, getitem_node in enumerate(getitem_nodes): + getitem_node.meta["val"] = trt_node.meta["val"][idx] + + gm.graph.erase_node(no_op_placeholder_node) + + gm.delete_all_unused_submodules() + gm.graph.eliminate_dead_code() + gm.graph.lint() + gm.recompile() + + return exp_program diff --git a/py/torch_tensorrt/dynamo/_settings.py b/py/torch_tensorrt/dynamo/_settings.py index 05f6f1c0e6..7a22663af3 100644 --- a/py/torch_tensorrt/dynamo/_settings.py +++ b/py/torch_tensorrt/dynamo/_settings.py @@ -13,6 +13,7 @@ DLA_LOCAL_DRAM_SIZE, DLA_SRAM_SIZE, DRYRUN, + ENABLE_CROSS_COMPILE_FOR_WINDOWS, ENABLE_EXPERIMENTAL_DECOMPOSITIONS, ENABLE_WEIGHT_STREAMING, ENABLED_PRECISIONS, @@ -88,6 +89,8 @@ class CompilationSettings: strip_engine_weights (bool): Whether to strip the engine weights immutable_weights (bool): Build non-refittable engines. This is useful for some layers that are not refittable. If this argument is set to true, `strip_engine_weights` and `refit_identical_engine_weights` will be ignored enable_weight_streaming (bool): Enable weight streaming. + enable_cross_compile_for_windows (bool): By default this is False means TensorRT engines can only be executed on the same platform where they were built. + True will enable cross-platform compatibility which allows the engine to be built on Linux and run on Windows """ enabled_precisions: Set[dtype] = field(default_factory=lambda: ENABLED_PRECISIONS) @@ -127,6 +130,7 @@ class CompilationSettings: strip_engine_weights: bool = STRIP_ENGINE_WEIGHTS immutable_weights: bool = IMMUTABLE_WEIGHTS enable_weight_streaming: bool = ENABLE_WEIGHT_STREAMING + enable_cross_compile_for_windows: bool = ENABLE_CROSS_COMPILE_FOR_WINDOWS _SETTINGS_TO_BE_ENGINE_INVARIANT = ( diff --git a/py/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py b/py/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py index 02c0407ee7..7ffc02ca3d 100644 --- a/py/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py +++ b/py/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py @@ -316,6 +316,12 @@ def _populate_trt_builder_config( if tactic_sources is not None: builder_config.set_tactic_sources(tactic_sources=tactic_sources) + if self.compilation_settings.enable_cross_compile_for_windows: + builder_config.runtime_platform = trt.RuntimePlatform.WINDOWS_AMD64 + _LOGGER.info( + "Setting runtime_platform as trt.RuntimePlatform.WINDOWS_AMD64" + ) + if self.compilation_settings.enable_weight_streaming: builder_config.set_flag(trt.BuilderFlag.WEIGHT_STREAMING) diff --git a/py/torch_tensorrt/dynamo/conversion/impl/slice/base.py b/py/torch_tensorrt/dynamo/conversion/impl/slice/base.py index 018ac63b8c..a2af840a1f 100644 --- a/py/torch_tensorrt/dynamo/conversion/impl/slice/base.py +++ b/py/torch_tensorrt/dynamo/conversion/impl/slice/base.py @@ -32,5 +32,5 @@ def slice( ) if dynamic_shape: layer.set_input(2, shape) - set_layer_name(layer, target, name) + set_layer_name(layer, target, name, source_ir) return layer.get_output(0) diff --git a/py/torch_tensorrt/dynamo/runtime/_MutableTorchTensorRTModule.py b/py/torch_tensorrt/dynamo/runtime/_MutableTorchTensorRTModule.py index ac2bf1512f..f51707768e 100644 --- a/py/torch_tensorrt/dynamo/runtime/_MutableTorchTensorRTModule.py +++ b/py/torch_tensorrt/dynamo/runtime/_MutableTorchTensorRTModule.py @@ -493,7 +493,7 @@ def save(module: Any, path: str) -> None: def load(path: str) -> Any: # When the model get saved, init_finished is set to False. # Class is restored to MutableTorchTensorRTModule, and some attribute is deleted - module = torch.load(path) + module = torch.load(path, weights_only=False) module.pytorch_model = _make_refit_change_trigger( module.original_model, module.refit_state ) diff --git a/py/torch_tensorrt/dynamo/runtime/_TorchTensorRTModule.py b/py/torch_tensorrt/dynamo/runtime/_TorchTensorRTModule.py index 5558244bc0..d7cfc6608b 100644 --- a/py/torch_tensorrt/dynamo/runtime/_TorchTensorRTModule.py +++ b/py/torch_tensorrt/dynamo/runtime/_TorchTensorRTModule.py @@ -133,7 +133,11 @@ def __init__( self.serialized_engine = serialized_engine self.engine = None - if serialized_engine and not self.settings.lazy_engine_init: + if ( + serialized_engine + and not self.settings.lazy_engine_init + and not self.settings.enable_cross_compile_for_windows + ): self.setup_engine() def _pack_engine_info(self) -> List[str | bytes]: @@ -145,16 +149,16 @@ def _pack_engine_info(self) -> List[str | bytes]: metadata = {"settings": self.settings, "weight_name_map": self.weight_name_map} target_platform = ( Platform.current_platform() + if not self.settings.enable_cross_compile_for_windows + else Platform.WIN_X86_64 ) # Change to match target for engine engine_info: List[str | bytes] = [""] * SERIALIZATION_LEN - engine_info[ABI_TARGET_IDX] = torch.ops.tensorrt.ABI_VERSION() engine_info[NAME_IDX] = ( self.name + "_engine" if self.name != "" else "tensorrt_engine" ) engine_info[DEVICE_IDX] = target_device._to_serialized_rt_device() - assert self.serialized_engine engine_info[ENGINE_IDX] = self.serialized_engine diff --git a/py/torch_tensorrt/runtime/_utils.py b/py/torch_tensorrt/runtime/_utils.py index ab427285e1..90da7f69ad 100644 --- a/py/torch_tensorrt/runtime/_utils.py +++ b/py/torch_tensorrt/runtime/_utils.py @@ -1,5 +1,5 @@ import logging -from typing import Optional, Tuple +from typing import List, Optional, Tuple import torch import torch_tensorrt @@ -128,3 +128,24 @@ def _get_most_compatible_device( best_match = candidate return best_match + + +@torch.library.custom_op( + "tensorrt::no_op_placeholder_for_execute_engine", mutates_args=() +) +def no_op_placeholder_for_execute_engine( + inputs: List[torch.Tensor], + abi_version: str, + name: str, + serialized_device_info: str, + serialized_engine: str, + serialized_in_binding_names: str, + serialized_out_binding_names: str, + serialized_hardware_compatible: str, + serialized_metadata: str, + serialized_target_platform: str, +) -> List[torch.Tensor]: + + raise RuntimeError( + "The saved model is cross compiled for windows in Linux, should only be loadded in Windows via torch_tensorrt.load_cross_compiled_exported_program() api." + ) diff --git a/py/torch_tensorrt/ts/_compile_spec.py b/py/torch_tensorrt/ts/_compile_spec.py index 4843ec0145..5d6d27e4ad 100644 --- a/py/torch_tensorrt/ts/_compile_spec.py +++ b/py/torch_tensorrt/ts/_compile_spec.py @@ -307,7 +307,7 @@ def _parse_compile_spec(compile_spec_: Dict[str, Any]) -> _ts_C.CompileSpec: def TensorRTCompileSpec( inputs: Optional[List[torch.Tensor | Input]] = None, input_signature: Optional[Any] = None, - device: torch.device | Device = Device._current_device(), + device: Optional[torch.device | Device] = None, disable_tf32: bool = False, sparse_weights: bool = False, enabled_precisions: Optional[Set[torch.dtype | dtype]] = None, @@ -365,7 +365,7 @@ def TensorRTCompileSpec( compile_spec = { "inputs": inputs if inputs is not None else [], # "input_signature": input_signature, - "device": device, + "device": Device._current_device() if device is None else device, "disable_tf32": disable_tf32, # Force FP32 layers to use traditional as FP32 format vs the default behavior of rounding the inputs to 10-bit mantissas before multiplying, but accumulates the sum using 23-bit mantissas "sparse_weights": sparse_weights, # Enable sparsity for convolution and fully connected layers. "enabled_precisions": ( diff --git a/tests/py/dynamo/runtime/test_003_cross_compile_for_windows.py b/tests/py/dynamo/runtime/test_003_cross_compile_for_windows.py new file mode 100644 index 0000000000..acf2aa006f --- /dev/null +++ b/tests/py/dynamo/runtime/test_003_cross_compile_for_windows.py @@ -0,0 +1,66 @@ +import os +import platform +import tempfile +import unittest + +import pytest +import torch +import torch_tensorrt +from torch.testing._internal.common_utils import TestCase + +from ..testing_utilities import DECIMALS_OF_AGREEMENT + + +class TestCrossCompileSaveForWindows(TestCase): + + @unittest.skipIf( + platform.system() != "Linux" or platform.architecture()[0] != "64bit", + "Cross compile for windows can only be enabled on linux x86-64 platform", + ) + @pytest.mark.unit + def test_cross_compile_for_windows(self): + class Add(torch.nn.Module): + def forward(self, a, b): + return torch.add(a, b) + + model = Add().eval().cuda() + inputs = [torch.randn(2, 3).cuda(), torch.randn(2, 3).cuda()] + trt_ep_path = os.path.join(tempfile.gettempdir(), "trt.ep") + compile_spec = { + "inputs": inputs, + "min_block_size": 1, + } + try: + torch_tensorrt.cross_compile_for_windows( + model, file_path=trt_ep_path, **compile_spec + ) + except Exception as e: + pytest.fail(f"unexpected exception raised: {e}") + + @unittest.skipIf( + platform.system() != "Linux" or platform.architecture()[0] != "64bit", + "Cross compile for windows can only be enabled on linux x86-64 platform", + ) + @pytest.mark.unit + def test_dynamo_cross_compile_for_windows(self): + class Add(torch.nn.Module): + def forward(self, a, b): + return torch.add(a, b) + + model = Add().eval().cuda() + inputs = (torch.randn(2, 3).cuda(), torch.randn(2, 3).cuda()) + trt_ep_path = os.path.join(tempfile.gettempdir(), "trt.ep") + exp_program = torch.export.export(model, inputs) + compile_spec = { + "inputs": inputs, + "min_block_size": 1, + } + try: + trt_gm = torch_tensorrt.dynamo.cross_compile_for_windows( + exp_program, **compile_spec + ) + torch_tensorrt.dynamo.save_cross_compiled_exported_program( + trt_gm, file_path=trt_ep_path + ) + except Exception as e: + pytest.fail(f"unexpected exception raised: {e}")