@@ -355,7 +355,7 @@ jobs:
355355 group : GZ_BD-CPU
356356 env :
357357 TASK : fleet-ci-paddle-build-whl-${{ github.event.pull_request.number }}
358- docker_image : " ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:cuda129-coverage-test "
358+ docker_image : " ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:ubuntu24- cuda129-dev "
359359 steps :
360360 - name : Check docker image and run container
361361 env :
@@ -458,13 +458,13 @@ jobs:
458458 steps :
459459 - name : Determine the runner
460460 run : |
461- gpu_id=$(( $(echo $PWD | awk -F'/' '{print $3}' | awk -F'-' '{print $2}') + 3 ))
461+ gpu_id=$(( $(echo $PWD | awk -F'/' '{print $3}' | awk -F'-' '{print $2}') - 1 ))
462462 echo GPU_DEVICES="$gpu_id" >> $GITHUB_ENV
463463
464464 - name : Check docker image and run container
465465 env :
466466 GPU_DEVICES : ${{ env.GPU_DEVICES }}
467- docker_image : " ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:cuda129-coverage-test "
467+ docker_image : " ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:ubuntu24- cuda129-dev "
468468 run : |
469469 container_name=${TASK}-$(date +%Y%m%d-%H%M%S)
470470 echo "container_name=${container_name}" >> ${{ github.env }}
@@ -500,19 +500,21 @@ jobs:
500500 docker exec -t ${{ env.container_name }} /bin/bash -ce '
501501 rm -rf * .[^.]*
502502 source /root/proxy
503- pip install uv coverage==7.6.1 bce-python-sdk==0.8.74 wrapt
503+ export LD_LIBRARY_PATH=/usr/local/cuda-12.9/targets/x86_64-linux/lib:/usr/local/cuda/lib64
504+ pip install uv coverage==7.6.1 bce-python-sdk==0.8.74 wrapt pytest matplotlib parameterized
504505 wget -q --tries=5 --no-proxy --no-check-certificate https://paddle-github-action.cdn.bcebos.com/PR/paddlefleet/${PR_ID}/${COMMIT_ID}/paddldfleet.tar.gz
505506 tar -xf paddldfleet.tar.gz --strip-components=1
506507 git config --global --add safe.directory /paddle
507508 pip install dist/paddlefleet-0.0.0-cp310-cp310-linux_x86_64.whl --extra-index-url=https://www.paddlepaddle.org.cn/packages/stable/cu129/ --extra-index-url=https://www.paddlepaddle.org.cn/packages/nightly/cu129/
508- echo "paddlefleet commit:"
509- python -c "import paddlefleet; print(paddlefleet.version.commit)"
509+ wget https://github.com/mikefarah/yq/releases/latest/download/yq_linux_amd64 -O /usr/local/bin/yq
510+ chmod +x /usr/local/bin/yq
510511 '
511512
512513 - name : Download paddle.tar.gz and install paddle whl
513514 run : |
514515 docker exec -t ${{ env.container_name }} /bin/bash -c '
515516 set -e
517+ export LD_LIBRARY_PATH=/usr/local/cuda-12.9/targets/x86_64-linux/lib:/usr/local/cuda/lib64
516518 mkdir -p /PaddlePaddle
517519 cd /PaddlePaddle
518520 echo "Downloading Paddle.tar.gz from cfs"
@@ -522,12 +524,15 @@ jobs:
522524 export UV_HTTP_TIMEOUT=300
523525 pip uninstall paddlepaddle-gpu -y
524526 pip install paddlepaddle_gpu-0.0.0-cp310-cp310-linux_x86_64.whl --force-reinstall --extra-index-url=https://www.paddlepaddle.org.cn/packages/nightly/cu129/
527+ echo "paddlefleet commit:"
528+ python -c "import paddlefleet; print(paddlefleet.version.commit)"
525529 '
526530
527531 - name : Single card test
528532 run : |
529533 docker exec -t ${{ env.container_name }} /bin/bash -xce '
530534 pwd
535+ export LD_LIBRARY_PATH=/usr/local/cuda-12.9/targets/x86_64-linux/lib:/usr/local/cuda/lib64
531536 if [ "${BRANCH}" != "develop" ]; then
532537 git checkout $fleet_branch
533538 echo "Checked out fleet branch: $fleet_branch"
@@ -539,7 +544,7 @@ jobs:
539544 export UV_NO_SYNC=1 # This environment variable prevents uv sync from being executed when running un run.
540545 export UV_HTTP_TIMEOUT=300
541546 python -c "import paddle; print(paddle.version.commit)"
542- timeout 10m bash ci/single_card_test.sh
547+ timeout 40m bash ci/single_card_test.sh
543548 single_card_exit_code=$?
544549 if [[ "$single_card_exit_code" != "0" ]]; then
545550 echo -e "::error:: \033[31mSingle card test failed.\033[0m"
@@ -565,7 +570,7 @@ jobs:
565570 env :
566571 PIP_CACHE_DIR : /root/.cache/pip
567572 TASK : paddle-fleet-CI-${{ github.event.pull_request.number }}-multi-card_test
568- docker_image : " ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:cuda129-coverage-test "
573+ docker_image : " ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:ubuntu24- cuda129-dev "
569574 steps :
570575 - name : Check docker image and run container
571576 run : |
@@ -604,19 +609,21 @@ jobs:
604609 docker exec -t ${{ env.container_name }} /bin/bash -ce '
605610 rm -rf * .[^.]*
606611 source /root/proxy
607- pip install uv coverage==7.6.1 bce-python-sdk==0.8.74 wrapt
612+ export LD_LIBRARY_PATH=/usr/local/cuda-12.9/targets/x86_64-linux/lib:/usr/local/cuda/lib64
613+ pip install uv coverage==7.6.1 bce-python-sdk==0.8.74 wrapt pytest matplotlib parameterized
608614 wget -q --tries=5 --no-proxy --no-check-certificate https://paddle-github-action.cdn.bcebos.com/PR/paddlefleet/${PR_ID}/${COMMIT_ID}/paddldfleet.tar.gz
609615 tar -xf paddldfleet.tar.gz --strip-components=1
610616 git config --global --add safe.directory /paddle
611617 pip install dist/paddlefleet-0.0.0-cp310-cp310-linux_x86_64.whl --extra-index-url=https://www.paddlepaddle.org.cn/packages/stable/cu129/ --extra-index-url=https://www.paddlepaddle.org.cn/packages/nightly/cu129/
612- echo "paddlefleet commit:"
613- python -c "import paddlefleet; print(paddlefleet.version.commit)"
618+ wget https://github.com/mikefarah/yq/releases/latest/download/yq_linux_amd64 -O /usr/local/bin/yq
619+ chmod +x /usr/local/bin/yq
614620 '
615621
616622 - name : Download paddle.tar.gz and install paddle whl
617623 run : |
618624 docker exec -t ${{ env.container_name }} /bin/bash -c '
619625 set -e
626+ export LD_LIBRARY_PATH=/usr/local/cuda-12.9/targets/x86_64-linux/lib:/usr/local/cuda/lib64
620627 mkdir -p /PaddlePaddle
621628 cd /PaddlePaddle
622629 echo "Downloading Paddle.tar.gz from cfs"
@@ -626,12 +633,15 @@ jobs:
626633 export UV_HTTP_TIMEOUT=300
627634 pip uninstall paddlepaddle-gpu -y
628635 pip install paddlepaddle_gpu-0.0.0-cp310-cp310-linux_x86_64.whl --force-reinstall --extra-index-url=https://www.paddlepaddle.org.cn/packages/nightly/cu129/
636+ echo "paddlefleet commit:"
637+ python -c "import paddlefleet; print(paddlefleet.version.commit)"
629638 '
630639
631640 - name : Multi-card test
632641 run : |
633642 docker exec -t ${{ env.container_name }} /bin/bash -ce '
634643 export PYTHONPATH=$(pwd)
644+ export LD_LIBRARY_PATH=/usr/local/cuda-12.9/targets/x86_64-linux/lib:/usr/local/cuda/lib64
635645 python -c "import paddle; print(paddle.version.commit)"
636646 if [ "${BRANCH}" != "develop" ]; then
637647 git checkout $fleet_branch
0 commit comments