TORCH_CUDA_ARCH_LIST="5.0;6.0;7.0;7.5;8.0;8.6"
case ${CUDA_VERSION} in
12.8)
TORCH_CUDA_ARCH_LIST="7.5;8.0;8.6;9.0;10.0;12.0+PTX" #removing sm_50-sm_70 as these architectures are deprecated in CUDA 12.8 and will be removed in future releases
EXTRA_CAFFE2_CMAKE_FLAGS+=("-DATEN_NO_TEST=ON")
;;
12.6)
TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST};9.0"
EXTRA_CAFFE2_CMAKE_FLAGS+=("-DATEN_NO_TEST=ON")
;;
and open webui change the PyTorch version accordingly. I think for many people still using 'old' cards (only 8 years for V100 with cuda capability 7.0) a separate docker tag will be convenient.
Desired Solution you'd like
the only thing needs to change is the github action, just copy the current cuda tag workflow, and add a build arg: --build-arg="USE_CUDA_VER=cu126" and name it with a new tag e.g. cuda-cu126
Alternatives Considered
none
Additional Context
No response
Originally created by @ER-EPR on GitHub (Jun 2, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/14586
### Check Existing Issues
- [x] I have searched the existing issues and discussions.
related to https://github.com/open-webui/open-webui/discussions/13654#discussion-8291452
### Problem Description
Since PyTorch drop support for cuda capability 7.0 and below in their 2.7-cu128 release, in https://github.com/pytorch/pytorch/blob/v2.7.0/.ci/manywheel/build_cuda.sh
```
TORCH_CUDA_ARCH_LIST="5.0;6.0;7.0;7.5;8.0;8.6"
case ${CUDA_VERSION} in
12.8)
TORCH_CUDA_ARCH_LIST="7.5;8.0;8.6;9.0;10.0;12.0+PTX" #removing sm_50-sm_70 as these architectures are deprecated in CUDA 12.8 and will be removed in future releases
EXTRA_CAFFE2_CMAKE_FLAGS+=("-DATEN_NO_TEST=ON")
;;
12.6)
TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST};9.0"
EXTRA_CAFFE2_CMAKE_FLAGS+=("-DATEN_NO_TEST=ON")
;;
```
and open webui change the PyTorch version accordingly. I think for many people still using 'old' cards (only 8 years for V100 with cuda capability 7.0) a separate docker tag will be convenient.
### Desired Solution you'd like
the only thing needs to change is the github action, just copy the current cuda tag workflow, and add a build arg: ` --build-arg="USE_CUDA_VER=cu126" ` and name it with a new tag e.g. cuda-cu126
### Alternatives Considered
none
### Additional Context
_No response_
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @ER-EPR on GitHub (Jun 2, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/14586
Check Existing Issues
related to https://github.com/open-webui/open-webui/discussions/13654#discussion-8291452
Problem Description
Since PyTorch drop support for cuda capability 7.0 and below in their 2.7-cu128 release, in https://github.com/pytorch/pytorch/blob/v2.7.0/.ci/manywheel/build_cuda.sh
and open webui change the PyTorch version accordingly. I think for many people still using 'old' cards (only 8 years for V100 with cuda capability 7.0) a separate docker tag will be convenient.
Desired Solution you'd like
the only thing needs to change is the github action, just copy the current cuda tag workflow, and add a build arg:
--build-arg="USE_CUDA_VER=cu126"and name it with a new tag e.g. cuda-cu126Alternatives Considered
none
Additional Context
No response
@tjbck commented on GitHub (Jun 2, 2025):
PR welcome.