Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Updated CI gpu image version #3097

Merged
merged 3 commits into from
Oct 17, 2023
Merged

Updated CI gpu image version #3097

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
a40dbf7
Updated CI gpu image version
vfdev-5 Oct 16, 2023
a0d3fd7
Fixed horovod installation to fix c++14 issue
vfdev-5 Oct 17, 2023
3847460
Merge branch 'master' into ci-update-gpu-image-version
vfdev-5 Oct 17, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
18 changes: 14 additions & 4 deletions .github/workflows/gpu-hvd-tests.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@ jobs:
pytorch-channel: [pytorch, ]
fail-fast: false
env:
DOCKER_IMAGE: "pytorch/conda-builder:cuda11.8"
DOCKER_IMAGE: "pytorch/conda-builder:cuda12.1"
REPOSITORY: ${{ github.repository }}
PR_NUMBER: ${{ github.event.pull_request.number }}
runs-on: linux.8xlarge.nvidia.gpu
Expand Down Expand Up @@ -102,9 +102,9 @@ jobs:

# Install PyTorch
if [ "${{ matrix.pytorch-channel }}" == "pytorch" ]; then
pip install --upgrade torch torchvision --index-url https://download.pytorch.org/whl/cu118
pip install --upgrade torch torchvision --index-url https://download.pytorch.org/whl/cu121
else
pip install --upgrade --pre torch torchvision --index-url https://download.pytorch.org/whl/nightly/cu118
pip install --upgrade --pre torch torchvision --index-url https://download.pytorch.org/whl/nightly/cu121
fi

python -c "import torch; print(torch.__version__, ', CUDA is available: ', torch.cuda.is_available()); exit(not torch.cuda.is_available())"
Expand All @@ -125,7 +125,17 @@ jobs:

set -xe

HOROVOD_GPU_OPERATIONS=NCCL HOROVOD_WITH_PYTORCH=1 pip install horovod[pytorch]
# Can't build Horovod with recent pytorch due to pytorch required C++17 standard
# and horovod is still using C++14
# HOROVOD_GPU_OPERATIONS=NCCL HOROVOD_WITH_PYTORCH=1 pip install horovod[pytorch]
# Using a similar hack as described here:
# https://github.com/horovod/horovod/issues/3941#issuecomment-1732505345
git clone --recursive https://github.com/horovod/horovod.git /horovod
cd /horovod
sed -i "s/CMAKE_CXX_STANDARD 14/CMAKE_CXX_STANDARD 17/g" CMakeLists.txt
sed -i "s/CMAKE_CXX_STANDARD 14/CMAKE_CXX_STANDARD 17/g" horovod/torch/CMakeLists.txt
HOROVOD_GPU_OPERATIONS=NCCL HOROVOD_WITH_PYTORCH=1 python setup.py install

horovodrun --check-build
pip list

Expand Down
6 changes: 3 additions & 3 deletions .github/workflows/gpu-tests.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@ jobs:
pytorch-channel: [pytorch, pytorch-nightly]
fail-fast: false
env:
DOCKER_IMAGE: "pytorch/conda-builder:cuda11.8"
DOCKER_IMAGE: "pytorch/conda-builder:cuda12.1"
REPOSITORY: ${{ github.repository }}
PR_NUMBER: ${{ github.event.pull_request.number }}
runs-on: linux.8xlarge.nvidia.gpu
Expand Down Expand Up @@ -102,9 +102,9 @@ jobs:

# Install PyTorch
if [ "${{ matrix.pytorch-channel }}" == "pytorch" ]; then
pip install --upgrade torch torchvision --index-url https://download.pytorch.org/whl/cu118
pip install --upgrade torch torchvision --index-url https://download.pytorch.org/whl/cu121
else
pip install --upgrade --pre torch torchvision --index-url https://download.pytorch.org/whl/nightly/cu118
pip install --upgrade --pre torch torchvision --index-url https://download.pytorch.org/whl/nightly/cu121
fi

python -c "import torch; print(torch.__version__, ', CUDA is available: ', torch.cuda.is_available()); exit(not torch.cuda.is_available())"
Expand Down