Update CUDA_UPGRADE_GUIDE.MD #2060
Conversation
| | 12.4 | 8.9.7.29 | Latest CUDA Nightly | | ||
| | 11.8 | 9.1.0.70 | Legacy CUDA Release | | ||
| | 12.6 (.3) | 9.5.1.17 | Stable CUDA Release | | ||
| | 12.8 (.1) | 9.7.1.26 | Latest CUDA Nightly | |
There was a problem hiding this comment.
We are currently building with 12.8.0 (not .1) if I'm not mistaken: https://github.com/pytorch/pytorch/blob/562328501e167206dc7d4b16895b5ae538520e06/.github/scripts/generate_binary_build_matrix.py#L75-L88 matches: https://docs.nvidia.com/cuda/archive/12.8.0/cuda-toolkit-release-notes/index.html#id4
Also, it seems torch==2.7.0+cu128 uses cuDNN==9.7.1.26, but the nightly builds use 9.8? https://github.com/pytorch/pytorch/blob/562328501e167206dc7d4b16895b5ae538520e06/.github/scripts/generate_binary_build_matrix.py#L78
We might need to clarify versioned releases vs. nightly builds.
There was a problem hiding this comment.
Thanks for checking, indeed we are using 12.8.0 - https://github.com/pytorch/pytorch/blob/main/.ci/docker/common/install_cuda.sh#L224.
Wondering why for CUDNN versions it is still 9.7.1.26 for 2.7.0, the update to 9.8 is merged on 3/13 and should be in 2.7.0 branch cut? (pytorch/pytorch#148963)
|
Closing as this one is outdated. |
No description provided.