Merge branch 'main' into dependabot/pip/requirements/lintrunner/pylin…

…t-3.1.0

.github/workflows/main.yaml

@@ -75,6 +75,8 @@ jobs:
       - name: Upload coverage to Codecov
         if: always()
         uses: codecov/codecov-action@v4
+        with:
+          token: ${{ secrets.CODECOV_TOKEN }}
       - name: Upload Test Results
         if: always()
         uses: actions/upload-artifact@v3

docs/examples/01_plot_selu.py

@@ -21,7 +21,7 @@ def Selu(X, alpha: float, gamma: float):
     neg = gammaX * (alphaX * op.Exp(X) - alphaX)
     pos = gammaX * X
     zero = op.CastLike(0, X)
-    return op.Where(X <= zero, neg, pos)
+    return op.Where(zero >= X, neg, pos)
 
 
 # %%

noxfile.py

@@ -13,7 +13,7 @@
     "jinja2",
     "numpy==1.24.4",
     "typing_extensions",
-    "beartype!=0.16.0",
+    "beartype==0.17.2",
     "types-PyYAML",
     "expecttest==0.1.6",
     "hypothesis",
@@ -26,7 +26,7 @@
     "pytest!=7.1.0",
     "pyyaml",
 )
-ONNX = "onnx==1.14.1"
+ONNX = "onnx==1.15"
 ONNX_RUNTIME = "onnxruntime==1.16.1"
 PYTORCH = "torch==2.1.0"
 TORCHVISON = "torchvision==0.16"

onnxscript/__init__.py

@@ -61,7 +61,7 @@
 from ._internal.utils import external_tensor
 from .values import OnnxFunction, TracedOnnxFunction
 
-try:
+try:  # noqa: SIM105
     __version__ = importlib.metadata.version("onnxscript")
 except importlib.metadata.PackageNotFoundError:
     # package is not installed

onnxscript/_internal/analysis.py

@@ -13,7 +13,7 @@
 
 def _get_loop_var(for_stmt: ast.For, formatter: sourceinfo.Formatter) -> str:
     if not isinstance(for_stmt.target, ast.Name):
-        raise ValueError(formatter(for_stmt, "For loop target must be a single variable."))
+        raise TypeError(formatter(for_stmt, "For loop target must be a single variable."))
     return for_stmt.target.id
 
 

onnxscript/_internal/utils.py

@@ -5,17 +5,14 @@
 from __future__ import annotations
 
 import numbers
-from typing import Any, Iterable, Optional, Sequence
+from typing import Optional, Sequence
 
 import numpy as np
 import onnx
 import onnx.helper
-from onnx import FunctionProto, ModelProto, TensorProto, ValueInfoProto
 
 from onnxscript import tensor
 
-# pylint: enable=unused-import, ungrouped-imports
-
 
 def external_tensor(
     name: str,
@@ -26,7 +23,7 @@ def external_tensor(
     length: Optional[int] = None,
     checksum: Optional[str] = None,
     basepath: Optional[str] = None,
-) -> TensorProto:
+) -> onnx.TensorProto:
     """Create a TensorProto referencing externally stored tensor-data.
 
     Args:
@@ -44,11 +41,11 @@ def external_tensor(
 
     See https://github.com/onnx/onnx/blob/main/docs/ExternalData.md for more details.
     """
-    tensor_proto = TensorProto()
+    tensor_proto = onnx.TensorProto()
     tensor_proto.name = name
     tensor_proto.data_type = data_type
     tensor_proto.dims.extend(dims)
-    tensor_proto.data_location = TensorProto.EXTERNAL
+    tensor_proto.data_location = onnx.TensorProto.EXTERNAL
 
     def add(k, v):
         entry = tensor_proto.external_data.add()
@@ -74,17 +71,17 @@ def value_to_type_proto(val):
         shape = val.shape
         return onnx.helper.make_tensor_type_proto(elem_type, shape)
     if isinstance(val, int):
-        return onnx.helper.make_tensor_type_proto(TensorProto.INT32, [])
+        return onnx.helper.make_tensor_type_proto(onnx.TensorProto.INT32, [])
     if isinstance(val, (float, np.float32)):
-        return onnx.helper.make_tensor_type_proto(TensorProto.FLOAT, [])
+        return onnx.helper.make_tensor_type_proto(onnx.TensorProto.FLOAT, [])
     if isinstance(val, list):
         if len(val) > 0:
             return onnx.helper.make_sequence_type_proto(value_to_type_proto(val[0]))
         # Edge-case. Cannot determine a suitable ONNX type for an empty list.
         # Should be using a typed-value instead.
         # Treated as a sequence of tensors of float-type.
         return onnx.helper.make_sequence_type_proto(
-            onnx.helper.make_tensor_type_proto(TensorProto.FLOAT, None)
+            onnx.helper.make_tensor_type_proto(onnx.TensorProto.FLOAT, None)
         )
     if isinstance(val, numbers.Number):
         nparray = np.array(val)
@@ -102,50 +99,3 @@ def values_to_value_infos(name_values):
         for (name, val) in name_values
         if val is not None
     ]
-
-
-def make_model_from_function_proto(
-    function_proto: FunctionProto,
-    function_opset_version: int,
-    input_value_infos: Sequence[ValueInfoProto],
-    output_value_infos: Sequence[ValueInfoProto],
-    **attrs: Any,
-) -> ModelProto:
-    """Creates a model containing a single call to a given
-    function with input and output value_infos, etc.
-
-    Args:
-        function_proto (FunctionProto): function proto
-            representing a single call
-        function_opset_version (int):  function_proto's version
-        input_value_infos (list of ValueInfoProto): function's input
-        output_value_infos (list of ValueInfoProto): function's output
-        **attrs (dict): the attributes of the node for the function
-
-    Returns:
-        ModelProto
-    """
-
-    input_names = [vi.name for vi in input_value_infos]
-    output_names = [vi.name for vi in output_value_infos]
-    node = onnx.helper.make_node(
-        function_proto.name,
-        input_names,
-        output_names,
-        domain=function_proto.domain,
-        **attrs,
-    )
-    graph = onnx.helper.make_graph([node], "node_graph", input_value_infos, output_value_infos)
-    model_proto_opset: Iterable[onnx.OperatorSetIdProto] = function_proto.opset_import
-    if all(o.domain != function_proto.domain for o in model_proto_opset):
-        model_proto_opset = [
-            *model_proto_opset,
-            onnx.helper.make_opsetid(function_proto.domain, function_opset_version),
-        ]
-    model = onnx.helper.make_model(
-        graph,
-        functions=[function_proto],
-        producer_name="onnxscript",
-        opset_imports=model_proto_opset,
-    )
-    return model

onnxscript/_thirdparty/README.md

@@ -0,0 +1,3 @@
+# Third-party
+
+Third-party vendored libraries. License in respective packages.