Skip to content

test cuda 12.8 #1861

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

Closed
t-vi wants to merge 1 commit into from
Closed

test cuda 12.8 #1861

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 2 additions & 4 deletions .azure/docker-build.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,10 +40,8 @@ jobs:
#maxParallel: "3"
matrix:
# CUDA 12.1
"cuda 12.1 | torch 2.5.1 | cudnn FE v1.5.2":
{ CUDA_VERSION: "12.1.1", TORCH_VERSION: "2.5.1", TRITON_VERSION: "3.1.0", CUDNN_FRONTEND_VERSION: "1.5.2" }
"cuda 12.1 | torch 2.5 /nightly | cudnn FE v1.5.2":
{ CUDA_VERSION: "12.1.1", TORCH_VERSION: "main", TORCH_INSTALL: "source", CUDNN_FRONTEND_VERSION: "1.5.2" }
"cuda 12.8 | torch 2.x /nightly | cudnn FE v1.10.0":
{ CUDA_VERSION: "12.8.0", TORCH_VERSION: "main", TORCH_INSTALL: "nightly", CUDNN_FRONTEND_VERSION: "1.10.0" }
#'cuda 12.1': # this version - '8.9.5.29-1+cuda12.1' for 'libcudnn8' was not found
# how much time to give 'run always even if cancelled tasks' before stopping them
cancelTimeoutInMinutes: "2"
Expand Down
103 changes: 55 additions & 48 deletions dockers/ubuntu-cuda/Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,18 +13,21 @@
# limitations under the License.

ARG UBUNTU_VERSION="22.04"
ARG CUDA_VERSION="12.1.0"
ARG CUDA_VERSION="12.8.0"
# select devel | runtime
ARG IMAGE_TYPE="devel"
#12.8.0-cudnn-devel-ubuntu24.04

FROM nvidia/cuda:${CUDA_VERSION}-${IMAGE_TYPE}-ubuntu${UBUNTU_VERSION}

ARG CUDNN_VERSION="9.1.0.70"
ARG CUDNN_FRONTEND_VERSION="1.5.2"
ARG PYTHON_VERSION="3.10"
ARG TORCH_VERSION="2.2.1"
ARG TRITON_VERSION="2.2.0"
ARG TORCH_INSTALL="stable"

FROM nvidia/cuda:${CUDA_VERSION}-cudnn-${IMAGE_TYPE}-ubuntu${UBUNTU_VERSION}

ARG CUDNN_VERSION="9.8.0.87-1"
ARG CUDNN_FRONTEND_VERSION="1.10.0"
ARG PYTHON_VERSION="3.12"
ARG TORCH_VERSION="main"
ARG TRITON_VERSION="3.1.0"
ARG TORCH_INSTALL="nightly"

SHELL ["/bin/bash", "-c"]
# https://techoverflow.net/2019/05/18/how-to-fix-configuring-tzdata-interactive-input-when-building-docker-images/
Expand Down Expand Up @@ -62,7 +65,6 @@ RUN \
add-apt-repository ppa:deadsnakes/ppa && \
apt-get install -y \
python${PYTHON_VERSION} \
python${PYTHON_VERSION}-distutils \
python${PYTHON_VERSION}-dev \
&& \
update-alternatives --install /usr/bin/python${PYTHON_VERSION%%.*} python${PYTHON_VERSION%%.*} /usr/bin/python${PYTHON_VERSION} 1 && \
Expand All @@ -83,22 +85,22 @@ RUN if [ "${TORCH_INSTALL}" == "source" ]; then \
CUDNN_BASE_VER=${CUDNN_VERSION%%.*} && \
CUDA_VERSION_M=${CUDA_VERSION%%.*} && \
apt update -qq --fix-missing && \
if [ "${CUDNN_BASE_VER}" == "9" ]; then \
CUDNN_PACKAGE_NAME="${CUDNN_VERSION}-1" && \
apt upgrade -y --allow-downgrades --allow-change-held-packages && \
apt install -y libcudnn9-cuda-${CUDA_VERSION_M}=${CUDNN_PACKAGE_NAME} \
libcudnn9-dev-cuda-${CUDA_VERSION_M}=${CUDNN_PACKAGE_NAME} \
nlohmann-json3-dev ; \
else \
CUDA_VERSION_MM=${CUDA_VERSION%.*} && \
# There are some test failures from cuDNN 12.1, so 'upgrade' requests for 12.1 to 12.2. \
CUDA_VERSION_MM="${CUDA_VERSION_MM/12.1/12.2}" && \
CUDNN_PACKAGE_NAME="${CUDNN_VERSION}-1+cuda${CUDA_VERSION_MM}" && \
apt upgrade -y --allow-downgrades --allow-change-held-packages && \
apt install -y libcudnn8=${CUDNN_PACKAGE_NAME} \
libcudnn8-dev=${CUDNN_PACKAGE_NAME} \
nlohmann-json3-dev ; \
fi && \
#if [ "${CUDNN_BASE_VER}" == "9" ]; then \
# CUDNN_PACKAGE_NAME="${CUDNN_VERSION}-1" && \
# apt upgrade -y --allow-downgrades --allow-change-held-packages && \
# apt install -y libcudnn9-cuda-${CUDA_VERSION_M}=${CUDNN_PACKAGE_NAME} \
# libcudnn9-dev-cuda-${CUDA_VERSION_M}=${CUDNN_PACKAGE_NAME} \
# nlohmann-json3-dev ; \
#else \
# CUDA_VERSION_MM=${CUDA_VERSION%.*} && \
# # There are some test failures from cuDNN 12.1, so 'upgrade' requests for 12.1 to 12.2. \
# CUDA_VERSION_MM="${CUDA_VERSION_MM/12.1/12.2}" && \
# CUDNN_PACKAGE_NAME="${CUDNN_VERSION}-1+cuda${CUDA_VERSION_MM}" && \
# apt upgrade -y --allow-downgrades --allow-change-held-packages && \
# apt install -y libcudnn8=${CUDNN_PACKAGE_NAME} \
# libcudnn8-dev=${CUDNN_PACKAGE_NAME} \
# nlohmann-json3-dev ; \
#fi && \
rm -rf /root/.cache && \
rm -rf /var/lib/apt/lists/*; \
fi
Expand All @@ -111,29 +113,23 @@ ENV \
ARG TORCH_INSTALL
ENV TORCH_USE_CUDA_DSA=1

RUN \
if [ "${TORCH_INSTALL}" == "source" ]; then \
# building pytorch from source
git clone --recursive https://github.com/pytorch/pytorch && \
cd pytorch && \
git checkout "${TORCH_VERSION}" && \
git submodule sync && \
git submodule update --init --recursive && \
pip install . && \
pip install "pytorch-triton==$(cat .ci/docker/triton_version.txt)" --index-url="https://download.pytorch.org/whl/nightly/" && \
cd .. && \
rm -rf pytorch; \
elif [ "${TORCH_INSTALL}" == "test" ]; then \
# installing pytorch from wheels
CUDA_VERSION_MM=${CUDA_VERSION%.*} && \
pip install "torch==${TORCH_VERSION}" "triton==${TRITON_VERSION}" \
--index-url="https://download.pytorch.org/whl/test/cu${CUDA_VERSION_MM//'.'/''}"; \
else \
# installing pytorch from wheels \
CUDA_VERSION_MM=${CUDA_VERSION%.*} && \
pip install "torch==${TORCH_VERSION}" "triton==${TRITON_VERSION}" \
--index-url="https://download.pytorch.org/whl/cu${CUDA_VERSION_MM//'.'/''}"; \
fi
RUN pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128
# if [ "${TORCH_INSTALL}" == "source" ]; then \
# # building pytorch from source
# git clone --recursive https://github.com/pytorch/pytorch && \
# cd pytorch && \
# git checkout "${TORCH_VERSION}" && \
# git submodule sync && \
# git submodule update --init --recursive && \
# pip install . && \
## pip install "pytorch-triton==$(cat .ci/docker/triton_version.txt)" --index-url="https://download.pytorch.org/whl/nightly/" && \
# cd .. && \
# rm -rf pytorch; \
# elif [ "${TORCH_INSTALL}" == "nightly" ]; then \
# # installing pytorch from wheels
# pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128 \
# fi


ARG TORCH_INSTALL

Expand All @@ -142,6 +138,17 @@ RUN \
# building nvFuser from source
git clone https://github.com/NVIDIA/Fuser.git && \
cd Fuser && \
git checkout segment_validate_speedup && \
git submodule update --init --recursive && \
pip install -r requirements.txt && \
python setup.py install --no-test --no-benchmark && \
cd .. && \
rm -rf Fuser ; \
elif [ "${TORCH_INSTALL}" == "nightly" ]; then \
# building nvFuser from source
git clone https://github.com/NVIDIA/Fuser.git && \
cd Fuser && \
git checkout segment_validate_speedup && \
git submodule update --init --recursive && \
pip install -r requirements.txt && \
python setup.py install --no-test --no-benchmark && \
Expand Down
Loading