fix bug

Azure · Feb 6, 2024 · eea1f11 · eea1f11
1 parent 17d0186
commit eea1f11
Show file tree

Hide file tree

Showing 4 changed files with 5 additions and 3 deletions.
diff --git a/.github/workflows/build-image.yaml b/.github/workflows/build-image.yaml
@@ -14,7 +14,7 @@ on:
 jobs:
   docker:
     name: Docker build ${{ matrix.name }}
-    runs-on: ubuntu-latest
+    runs-on: [self-hosted, linux, x64, cpu]
     timeout-minutes: 600
     permissions:
       contents: read

diff --git a/.github/workflows/unit-tests.yaml b/.github/workflows/unit-tests.yaml
@@ -41,9 +41,10 @@ jobs:
     - name: Install dependencies
       run: |
         export LD_LIBRARY_PATH="/usr/local/lib:$LD_LIBRARY_PATH"
+        export DEBIAN_FRONTEND=noninteractive
         python3 -m pip install --upgrade pip
-        cd ${{ matrix.dir }}/
         apt-get update && apt-get install -y python3-mpi4py
+        cd ${{ matrix.dir }}/
         python3 -m pip install .[test]
         make postinstall
     - name: Run code lint

diff --git a/dockerfile/torch1.14-cuda11.8.dockerfile b/dockerfile/torch1.14-cuda11.8.dockerfile
@@ -51,7 +51,7 @@ RUN cd third_party/msccl && \
 # cache TE build to save time in CI
 ENV MAX_JOBS=1
 RUN python3 -m pip install --upgrade pip && \
-    MAX_JOBS=1 python3 -m pip install flash-attn==1.0.9 git+https://github.com/NVIDIA/TransformerEngine.git@stable
+    python3 -m pip install git+https://github.com/NVIDIA/TransformerEngine.git@stable
 
 ADD . .
 RUN python3 -m pip install . && \

diff --git a/dockerfile/torch2.1-cuda12.2.dockerfile b/dockerfile/torch2.1-cuda12.2.dockerfile
@@ -49,6 +49,7 @@ RUN cd third_party/msccl && \
     -gencode=arch=compute_90,code=sm_90" && \
     make install
 # cache TE build to save time in CI
+ENV MAX_JOBS=1
 RUN python3 -m pip install --upgrade pip && \
     python3 -m pip install git+https://github.com/NVIDIA/TransformerEngine.git@stable