test

.github/workflows/release.yaml

@@ -77,8 +77,10 @@ jobs:
         TAG_NAME=$(echo "${BRANCH_NAME}" | sed 's/\//_/g')
         echo "BRANCH_NAME=${BRANCH_NAME}" >> $GITHUB_ENV
 
-        echo "DOCKER_TAG=mosaicml/llm-foundry:release_${TAG_NAME}" >> $GITHUB_ENV
-        echo "AWS_DOCKER_TAG=mosaicml/llm-foundry:release_${TAG_NAME}_aws" >> $GITHUB_ENV
+        echo "DOCKER_TAG=mosaicml/llm-foundry:test_release_${TAG_NAME}" >> $GITHUB_ENV
+        echo "AWS_DOCKER_TAG=mosaicml/llm-foundry:test_release_${TAG_NAME}_aws" >> $GITHUB_ENV
+        # echo "LATEST_TAG=mosaicml/llm-foundry:release-latest" >> $GITHUB_ENV
+        # echo "AWS_LATEST_TAG=mosaicml/llm-foundry:release_aws-latest" >> $GITHUB_ENV
 
 
     - name: Build and push AWS Docker image
@@ -89,6 +91,7 @@ jobs:
         push: true
         tags: |
           ${{ env.AWS_DOCKER_TAG }}
+          ${{ env.AWS_LATEST_TAG }}
         build-args: |
           BASE_IMAGE=mosaicml/pytorch:2.4.0_cu124-python3.11-ubuntu20.04-aws
           BRANCH_NAME=${{ env.BRANCH_NAME }}
@@ -104,6 +107,7 @@ jobs:
         push: true
         tags: |
           ${{ env.DOCKER_TAG }}
+          ${{ env.LATEST_TAG }}
         build-args: |
           BASE_IMAGE=mosaicml/pytorch:2.4.0_cu124-python3.11-ubuntu20.04
           BRANCH_NAME=${{ env.BRANCH_NAME }}

Dockerfile

@@ -1,46 +1,30 @@
+# Copyright 2022 MosaicML LLM Foundry authors
+# SPDX-License-Identifier: Apache-2.0
+
 ARG BASE_IMAGE
 FROM $BASE_IMAGE
 
 ARG BRANCH_NAME
 ARG DEP_GROUPS
 ARG TE_COMMIT
-ARG KEEP_FOUNDRY="false"
+ARG KEEP_FOUNDRY=false
 
 ENV TORCH_CUDA_ARCH_LIST="8.0 8.6 8.7 8.9 9.0"
 
-RUN echo "Starting build with BRANCH_NAME=${BRANCH_NAME}, DEP_GROUPS=${DEP_GROUPS}, TE_COMMIT=${TE_COMMIT}, KEEP_FOUNDRY=${KEEP_FOUNDRY}"
-
 # Check for changes in setup.py.
+# If there are changes, the docker cache is invalidated and a fresh pip installation is triggered.
 ADD https://raw.githubusercontent.com/mosaicml/llm-foundry/$BRANCH_NAME/setup.py 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@$TE_COMMIT
 
-RUN echo "Cloning llm-foundry repository"
+# Install and uninstall foundry to cache foundry requirements
 RUN git clone -b $BRANCH_NAME https://github.com/mosaicml/llm-foundry.git
-
-RUN echo "Installing llm-foundry"
 RUN pip install --no-cache-dir "./llm-foundry${DEP_GROUPS}"
 
-RUN echo "KEEP_FOUNDRY value: ${KEEP_FOUNDRY}"
-
 # Conditionally uninstall llm-foundry and remove its directory
 RUN if [ "$KEEP_FOUNDRY" != "true" ]; then \
-      echo "Uninstalling llm-foundry"; \
       pip uninstall -y llm-foundry && \
       rm -rf llm-foundry; \
-    else \
-      echo "Keeping llm-foundry installed"; \
     fi
-
-# Add a final check to see if llm-foundry is installed
-RUN pip list | grep llm-foundry || echo "llm-foundry not found in pip list"
-
-# Print the contents of the directory where llm-foundry was cloned
-RUN ls -la /llm-foundry || echo "llm-foundry directory not found"
-
-# Try to import llm-foundry in Python
-RUN python -c "import llm_foundry; print('llm-foundry successfully imported')" || echo "Failed to import llm-foundry"
-
-RUN echo "Build process completed"