Torch 2.3.1 Upgrade (mosaicml#3367)

* Revert "Autoresume Validation with Max Duration (mosaicml#3358)" This reverts commit f0eae8a. * add bump * bump deepspeed * revert deepspeed bump * add state dict --------- Co-authored-by: Saaketh Narayan <[email protected]>
mvpatel2000 · Jun 6, 2024 · 5bf4d60 · 5bf4d60
1 parent f039f06
commit 5bf4d60
Show file tree

Hide file tree

Showing 5 changed files with 31 additions and 12 deletions.
diff --git a/.github/workflows/daily.yaml b/.github/workflows/daily.yaml
@@ -34,7 +34,7 @@ jobs:
           pytest_command: coverage run -m pytest
           composer_package_name: composer
         - name: cpu-3.11-2.3
-          container: mosaicml/pytorch:2.3.0_cpu-python3.11-ubuntu20.04
+          container: mosaicml/pytorch:2.3.1_cpu-python3.11-ubuntu20.04
           markers: not daily and (remote or not remote) and not gpu and not doctest
           pytest_command: coverage run -m pytest
           composer_package_name: mosaicml
@@ -59,7 +59,7 @@ jobs:
           pytest_command: coverage run -m pytest
           composer_package_name: composer
         - name: daily-cpu-3.11-2.3-composer
-          container: mosaicml/pytorch:2.3.0_cpu-python3.11-ubuntu20.04
+          container: mosaicml/pytorch:2.3.1_cpu-python3.11-ubuntu20.04
           markers: daily and (remote or not remote) and not gpu and not doctest
           pytest_command: coverage run -m pytest
           composer_package_name: composer
@@ -119,7 +119,7 @@ jobs:
           composer_package_name: "mosaicml"
           gpu_num: 1
         - name: "gpu-3.11-2.3-1-gpu"
-          container: mosaicml/pytorch:2.3.0_cu121-python3.11-ubuntu20.04
+          container: mosaicml/pytorch:2.3.1_cu121-python3.11-ubuntu20.04
           markers: "(daily or not daily) and (remote or not remote) and gpu and (doctest or not doctest)"
           pytest_command: "coverage run -m pytest"
           composer_package_name: "mosaicml"
@@ -137,7 +137,7 @@ jobs:
           composer_package_name: "mosaicml"
           gpu_num: 2
         - name: "gpu-3.11-2.3-2-gpu"
-          container: mosaicml/pytorch:2.3.0_cu121-python3.11-ubuntu20.04
+          container: mosaicml/pytorch:2.3.1_cu121-python3.11-ubuntu20.04
           markers: "(daily or not daily) and (remote or not remote) and gpu and (doctest or not doctest)"
           pytest_command: "coverage run -m pytest"
           composer_package_name: "mosaicml"
@@ -155,7 +155,7 @@ jobs:
           composer_package_name: "mosaicml"
           gpu_num: 4
         - name: "gpu-3.11-2.3-4-gpu"
-          container: mosaicml/pytorch:2.3.0_cu121-python3.11-ubuntu20.04
+          container: mosaicml/pytorch:2.3.1_cu121-python3.11-ubuntu20.04
           markers: "(daily or not daily) and (remote or not remote) and gpu and (doctest or not doctest)"
           pytest_command: "coverage run -m pytest"
           composer_package_name: "mosaicml"

diff --git a/.github/workflows/pr-cpu.yaml b/.github/workflows/pr-cpu.yaml
@@ -22,11 +22,11 @@ jobs:
           markers: not daily and not remote and not gpu and not doctest
           pytest_command: coverage run -m pytest
         - name: cpu-3.11-2.3
-          container: mosaicml/pytorch:2.3.0_cu121-python3.11-ubuntu20.04
+          container: mosaicml/pytorch:2.3.1_cu121-python3.11-ubuntu20.04
           markers: not daily and not remote and not gpu and not doctest
           pytest_command: coverage run -m pytest
         - name: cpu-doctest
-          container: mosaicml/pytorch:2.2.1_cpu-python3.11-ubuntu20.04
+          container: mosaicml/pytorch:2.3.1_cpu-python3.11-ubuntu20.04
           markers: not daily and not remote and not gpu and doctest
           pytest_command: coverage run -m pytest tests/test_docs.py
     name: ${{ matrix.name }}

diff --git a/.github/workflows/pr-gpu.yaml b/.github/workflows/pr-gpu.yaml
@@ -14,7 +14,7 @@ jobs:
       matrix:
         include:
         - name: gpu-3.11-2.3-1
-          container: mosaicml/pytorch:2.3.0_cu121-python3.11-ubuntu20.04
+          container: mosaicml/pytorch:2.3.1_cu121-python3.11-ubuntu20.04
           markers: not daily and not remote and gpu and (doctest or not doctest)
           pytest_command: coverage run -m pytest
           composer_package_name: mosaicml
@@ -40,7 +40,7 @@ jobs:
       matrix:
         include:
         - name: gpu-3.11-2.3-2
-          container: mosaicml/pytorch:2.3.0_cu121-python3.11-ubuntu20.04
+          container: mosaicml/pytorch:2.3.1_cu121-python3.11-ubuntu20.04
           markers: not daily and not remote and gpu and (doctest or not doctest)
           pytest_command: coverage run -m pytest
           composer_package_name: mosaicml
@@ -67,7 +67,7 @@ jobs:
       matrix:
         include:
         - name: gpu-3.11-2.3-4
-          container: mosaicml/pytorch:2.3.0_cu121-python3.11-ubuntu20.04
+          container: mosaicml/pytorch:2.3.1_cu121-python3.11-ubuntu20.04
           markers: not daily and not remote and gpu and (doctest or not doctest)
           pytest_command: coverage run -m pytest
           composer_package_name: mosaicml

diff --git a/composer/trainer/_patch_pytorch.py b/composer/trainer/_patch_pytorch.py
@@ -104,6 +104,25 @@ def patch_pytorch():
         DeviceMesh.__getitem__ = device_mesh__getitem__
         DeviceMesh.__init__ = device_mesh__init__
 
+    elif version.parse(torch.__version__) < version.parse('2.3.2'):
+        # Monkey patch for torch < 2.3.2 ie torch == 2.3.1
+
+        # Issue: https://github.com/pytorch/pytorch/issues/122946
+        #  - PR: https://github.com/pytorch/pytorch/pull/125336
+        from torch.distributed.checkpoint import state_dict
+
+        state_dict._verify_options = _verify_options
+        state_dict._get_model_state_dict = _get_model_state_dict
+        state_dict._load_model_state_dict = _load_model_state_dict
+
+        # Monkeypatch for ND child submeshes
+        # PR: https://github.com/pytorch/pytorch/pull/119752
+        from torch.distributed.device_mesh import DeviceMesh, _MeshEnv
+
+        _MeshEnv.create_child_mesh = create_child_mesh
+        DeviceMesh.__getitem__ = device_mesh__getitem__
+        DeviceMesh.__init__ = device_mesh__init__
+
 
 def build_metadata(
     self,

diff --git a/setup.py b/setup.py
@@ -80,8 +80,8 @@ def package_files(prefix: str, directory: str, extension: str):
     'tqdm>=4.62.3,<5',
     'torchmetrics>=0.10.0,<1.3.3',
     'torch_optimizer>=0.3.0,<0.4',
-    'torchvision>=0.13.1,<0.18.1',
-    'torch>=2.1.2,<2.3.1',
+    'torchvision>=0.13.1,<0.18.2',
+    'torch>=2.1.2,<2.3.2',
     'requests>=2.26.0,<3',
     'numpy>=1.21.5,<1.27.0',
     'psutil>=5.8.0,<6',