[cuDNN] cuDNN to 9.7.1.26 for CUDA 12.8#146957
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/146957
Note: Links to docs will display an error until the docs builds have been completed. ⏳ No Failures, 3 PendingAs of commit 1a64bc6 with merge base f50d359 ( UNSTABLE - The following jobs are marked as unstable, possibly due to flakiness on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
| if [[ ${CUDA_VERSION:0:4} == "12.8" ]]; then | ||
| CUDNN_NAME="cudnn-linux-x86_64-9.7.1.26_cuda12-archive" | ||
| elif [[ ${CUDA_VERSION:0:4} == "12.6" ]]; then | ||
| CUDNN_NAME="cudnn-linux-x86_64-9.5.1.17_cuda12-archive" |
There was a problem hiding this comment.
12.6 is supported by the latest CUDNN too, why not build against there as well?
There was a problem hiding this comment.
good point, issue is that the AMI testing for cuda 12.8 is in process, if we update cuda 12.6 now in this PR it will need to rebuild win AMI (it's the same AMI for various cuda versions). Let's update 12.6 cudnn next week as a follow-up.
There was a problem hiding this comment.
CUDNN 9.6 fixes some hopper bugs for fp8 matmuls so we should update it to the latest version too.
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
follow up for pytorch/pytorch#146957 use 9.7.1.26 for CUDA 12.6 too to test the windows AMI cc @atalman
|
@tinglvv Why didn't we update 12.6 at the same time? |
rebasing for #146717
cc @atalman @malfet @eqy @ptrblck @nWEIdia