8000 Add libtorch nightly build for CUDA 12.8 (#146265) · pytorch/pytorch@fe100c3 · GitHub
[go: up one dir, main page]
Skip to content

Commit fe100c3

Browse files
tinglvvpytorchmergebot
tinglvv
authored and
pytorchmergebot
committed
Add libtorch nightly build for CUDA 12.8 (#146265)
Try removing sm50 and sm60 to shrink binary size, and resolve the ld --relink error "Architecture support for Maxwell, Pascal, and Volta is considered feature-complete and will be frozen in an upcoming release." from 12.8 release note. Also updating the runner for cuda 12.8 test to g4dn (T4, sm75) due to the drop in sm50/60 support. #145570 Pull Request resolved: #146265 Approved by: https://github.com/atalman
1 parent ba214ab commit fe100c3

6 files changed

+77
-12
lines changed

‎.ci/manywheel/build_cuda.sh

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -54,7 +54,7 @@ cuda_version_nodot=$(echo $CUDA_VERSION | tr -d '.')
5454
TORCH_CUDA_ARCH_LIST="5.0;6.0;7.0;7.5;8.0;8.6"
 8000
5555
case ${CUDA_VERSION} in
5656
12.8)
57-
TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST};9.0;10.0;12.0+PTX" #Ripping out 5.0 and 6.0 due to ld error
57+
TORCH_CUDA_ARCH_LIST="7.0;7.5;8.0;8.6;9.0;10.0;12.0+PTX" #Ripping out 5.0 and 6.0 due to ld error
5858
EXTRA_CAFFE2_CMAKE_FLAGS+=("-DATEN_NO_TEST=ON")
5959
;;
6060
12.6)

‎.github/scripts/generate_binary_build_matrix.py

Lines changed: 0 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -246,9 +246,6 @@ def generate_libtorch_matrix(
246246
if os == "linux":
247247
arches += CUDA_ARCHES
248248
arches += ROCM_ARCHES
249-
# skip CUDA 12.8 builds for libtorch
250-
if "12.8" in arches:
251-
arches.remove("12.8")
252249
elif os == "windows":
253250
arches += CUDA_ARCHES
254251
if libtorch_variants is None:

‎.github/templates/linux_binary_build_workflow.yml.j2

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -111,7 +111,10 @@ jobs:
111111
ALPINE_IMAGE: "docker.io/s390x/alpine"
112112
{%- elif config["gpu_arch_type"] == "rocm" %}
113113
runs_on: linux.rocm.gpu
114-
{%- elif config["gpu_arch_type"] == "cuda" %}
114+
{%- elif config["gpu_arch_type"] == "cuda" and config["gpu_arch_version"] == "12.8" %}
115+
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
116+
runs_on: linux.g4dn.4xlarge.nvidia.gpu # 12.8 build needs sm_70+ runner
117+
{%- elif config["gpu_arch_type"] == "cuda" and config["gpu_arch_version"] != "12.8"%}
115118
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
116119
runs_on: linux.4xlarge.nvidia.gpu
117120
{%- else %}

‎.github/workflows/generated-linux-binary-libtorch-cxx11-abi-nightly.yml

Lines changed: 65 additions & 0 deletions
Some generated files are not rendered by default. Learn more about customizing how changed files appear on GitHub.

‎.github/workflows/generated-linux-binary-manywheel-main.yml

Lines changed: 1 addition & 1 deletion
Some generated files are not rendered by default. Learn more about customizing how changed files appear on GitHub.

‎.github/workflows/generated-linux-binary-manywheel-nightly.yml

Lines changed: 6 additions & 6 deletions
Some generated files are not rendered by default. Learn more about customizing how changed files appear on GitHub.

0 commit comments

Comments
 (0)
0