paramtrize

.github/workflows/docker.yaml

@@ -20,15 +20,19 @@ jobs:
         - name: "2.3.1_cu121"
           base_image: mosaicml/pytorch:2.3.1_cu121-python3.11-ubuntu20.04
           dep_groups: "[all]"
+          te_commit: b5a7c9f
         - name: "2.3.1_cu121_aws"
           base_image: mosaicml/pytorch:2.3.1_cu121-python3.11-ubuntu20.04-aws
           dep_groups: "[all]"
+          te_commit: b5a7c9f
         - name: "2.4.0_cu124"
           base_image: mosaicml/pytorch:2.4.0_cu124-python3.11-ubuntu20.04
           dep_groups: "[all]"
+          te_commit: 901e5d2
         - name: "2.4.0_cu124_aws"
           base_image: mosaicml/pytorch:2.4.0_cu124-python3.11-ubuntu20.04-aws
           dep_groups: "[all]"
+          te_commit: 901e5d2
     steps:
 
     - name: Checkout
@@ -95,3 +99,4 @@ jobs:
           BRANCH_NAME=${{ github.head_ref || github.ref_name }}
           BASE_IMAGE=${{ matrix.base_image }}
           DEP_GROUPS=${{ matrix.dep_groups }}
+          TE_COMMIT=${{ matrix.te_commit }}

Dockerfile

@@ -6,6 +6,7 @@ FROM $BASE_IMAGE
 
 ARG BRANCH_NAME
 ARG DEP_GROUPS
+ARG TE_COMMIT
 
 ENV TORCH_CUDA_ARCH_LIST="8.0 8.6 8.7 8.9 9.0"
 
@@ -15,7 +16,7 @@ ADD https://raw.githubusercontent.com/mosaicml/llm-foundry/$BRANCH_NAME/setup.py
 RUN rm setup.py
 
 # Install TransformerEngine
-RUN NVTE_FRAMEWORK=pytorch CMAKE_BUILD_PARALLEL_LEVEL=4 MAX_JOBS=4 pip install git+https://github.com/NVIDIA/TransformerEngine.git@901e5d2
+RUN NVTE_FRAMEWORK=pytorch CMAKE_BUILD_PARALLEL_LEVEL=4 MAX_JOBS=4 pip install git+https://github.com/NVIDIA/TransformerEngine.git@$TE_COMMIT
 
 # Install and uninstall foundry to cache foundry requirements
 RUN git clone -b $BRANCH_NAME https://github.com/mosaicml/llm-foundry.git