# PyTorch 1.6 - "Tesla T4 with CUDA capability sm\_75 is not compatible" **URL:** \ **Category:** Uncategorized\ **Created:** [July 29, 2020, 8:49pm UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003 "2020-07-29T20:49:23Z")\ **Posts on this page:** 20\ **Page:** 1
**Author:** ![sniklaus](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/sniklaus/32/3390_2.png) [@sniklaus](https://discuss.pytorch.org/u/sniklaus)\ **Post date:** [July 29, 2020, 8:49pm UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/1 "2020-07-29T20:49:23Z")
I installed PyTorch 1.6 with pip as follows. ```auto pip install --no-cache-dir torch==1.6.0+cu101 torchvision==0.7.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html ``` However, PyTorch prints the following now. ```auto Tesla T4 with CUDA capability sm_75 is not compatible with the current PyTorch installation. The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_70. ``` How come the pre-built PyTorch binaries for 1.6 do not support the newest CUDA capability? Is this intentional? How can I fix my setup? Thanks! ---
**Author:** ![kl\_divergence](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/kl_divergence/32/5069_2.png) [@kl\_divergence](https://discuss.pytorch.org/u/kl_divergence)\ **Post date:** [July 30, 2020, 5:28am UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/2 "2020-07-30T05:28:28Z")
Even I’m facing this issue with `Quadro RTX 8000`. The official recommendation is to build from source. But I don’t get it, why is not supported ? ---
**Author:** ![ptrblck](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ptrblck/32/1823_2.png) [@ptrblck](https://discuss.pytorch.org/u/ptrblck)\ **Post date:** [July 30, 2020, 6:41am UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/3 "2020-07-30T06:41:05Z")
@seemethere was `sm_75` dropped accidentally, since the release notes only mention `6.1` and `PTX for 3.7`? However, [this PR](https://github.com/pytorch/builder/commit/8d064bfcfe2e73d89a067dd5f9311828c1792f90) seems to have also removed `7.5`. ---
**Author:** ![ilovescience](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/i/e495f1/32.png) [@ilovescience](https://discuss.pytorch.org/u/ilovescience)\ **Post date:** [July 30, 2020, 7:05am UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/4 "2020-07-30T07:05:26Z")
I also have the same warning on Colab Tesla T4 GPU. But it doesn’t seem to actually affect anything I can still use CUDA. ---
**Author:** ![kl\_divergence](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/kl_divergence/32/5069_2.png) [@kl\_divergence](https://discuss.pytorch.org/u/kl_divergence)\ **Post date:** [July 30, 2020, 8:27am UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/5 "2020-07-30T08:27:11Z")
I’m able to use but not sure whether all CUDA functionalities work as intended because of this error. This is a strange behavior. Building Pytorch isn’t easy thing on my restricted server. ---
**Author:** ![tom](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/tom/32/3162_2.png) [@tom](https://discuss.pytorch.org/u/tom)\ **Post date:** [July 30, 2020, 9:12am UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/6 "2020-07-30T09:12:14Z")
I think it’s there for cuda 10.2. (But my poor 1080 Ti is dropped.) ---
**Author:** ![bsun0802](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/bsun0802/32/21730_2.png) [@bsun0802](https://discuss.pytorch.org/u/bsun0802)\ **Post date:** [July 30, 2020, 11:40am UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/7 "2020-07-30T11:40:08Z")
Facing the same issue here. Card is GeForce RTX 2080 Ti. What does this warning mean? Inside torch, torch.cuda.is\_available() returns True and I can train with GPU. Do I need to and how can I fix it? Thanks in advance. ```auto /home/zbxs/miniconda3/lib/python3.7/site-packages/torch/cuda/ __init__.py:125: UserWarning: GeForce RTX 2080 Ti with CUDA capability sm_75 is not compatible with the current PyTorch installation. The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_70. If you want to use the GeForce RTX 2080 Ti GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/ warnings.warn(incompatible_device_warn.format(device_name, capability, " ".join(arch_list), device_name)) ``` Python=3.7.7 cudatoolkit 10.1.243 h6bb024c\_0 defaults Pytorch 1.6.0. ---
**Author:** ![ptrblck](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ptrblck/32/1823_2.png) [@ptrblck](https://discuss.pytorch.org/u/ptrblck)\ **Post date:** [July 30, 2020, 5:29pm UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/8 "2020-07-30T17:29:30Z")
You are right. I missed the `7.5` argument for CUDA10.2 @bsun0802 as @tom mentioned, `sm_75` is still available in the CUDA10.2 binary, so you might want to install this particular version. ---
**Author:** ![sniklaus](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/sniklaus/32/3390_2.png) [@sniklaus](https://discuss.pytorch.org/u/sniklaus)\ **Post date:** [July 30, 2020, 5:39pm UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/9 "2020-07-30T17:39:18Z")
What is the rationale behind removing `sm_75` for the binaries that target CUDA 10.1? I am afraid that I cannot upgrade to CUDA 10.2 due to dependency and driver issues. What are the effects of using the PyTorch binaries that do not support `sm_75` on a graphics card that is supporting `sm_75`? ---
**Author:** ![smth](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/smth/32/13_2.png) [@smth](https://discuss.pytorch.org/u/smth)\ **Post date:** [July 30, 2020, 7:17pm UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/10 "2020-07-30T19:17:48Z")
tagging @seemethere, we need to fix this bug and re-push the CUDA 10.1 binaries. ---
**Author:** ![smth](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/smth/32/13_2.png) [@smth](https://discuss.pytorch.org/u/smth)\ **Post date:** [July 30, 2020, 7:49pm UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/11 "2020-07-30T19:49:38Z")
okay, I got an update. @seemethere is fixing the issue and re-uploading the cuda 10.1 binaries asap, like in ~3 hours or so ---
**Author:** ![sniklaus](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/sniklaus/32/3390_2.png) [@sniklaus](https://discuss.pytorch.org/u/sniklaus)\ **Post date:** [July 30, 2020, 8:14pm UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/12 "2020-07-30T20:14:21Z")
Awesome, thank you for taking care of this so quickly! ---
**Author:** ![sniklaus](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/sniklaus/32/3390_2.png) [@sniklaus](https://discuss.pytorch.org/u/sniklaus)\ **Post date:** [July 31, 2020, 5:05am UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/13 "2020-07-31T05:05:29Z")
Just a heads-up for everyone following along, the issue has been [fixed](https://github.com/pytorch/builder/pull/472) underneath the hood but the binaries have unfortunately not been update yet: [https://anaconda.org/pytorch/pytorch/files](https://anaconda.org/pytorch/pytorch/files) ---
**Author:** ![sniklaus](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/sniklaus/32/3390_2.png) [@sniklaus](https://discuss.pytorch.org/u/sniklaus)\ **Post date:** [July 31, 2020, 7:53pm UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/14 "2020-07-31T19:53:18Z")
The new binaries seem to be available now, at least for `pip`. Make sure to use `--no-cache-dir` to not use a locally cached binary when `pip` installing PyTorch again. ```auto pip install --no-cache-dir torch==1.6.0+cu101 torchvision==0.7.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html ``` ---
**Author:** ![smth](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/smth/32/13_2.png) [@smth](https://discuss.pytorch.org/u/smth)\ **Post date:** [July 31, 2020, 10:47pm UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/15 "2020-07-31T22:47:18Z")
conda binaries shouldn’t have been broken, only the wheels were broken, and they are fixed now. thanks for patience 🙂 ---
**Author:** ![sniklaus](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/sniklaus/32/3390_2.png) [@sniklaus](https://discuss.pytorch.org/u/sniklaus)\ **Post date:** [August 1, 2020, 1:36am UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/16 "2020-08-01T01:36:25Z")
Thank you for chiming in! I can confirm, the new binaries showed up on pip a few hours or so ago. 👍 ---
**Author:** ![pesenti](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/p/a3d4f5/32.png) [@pesenti](https://discuss.pytorch.org/u/pesenti)\ **Post date:** [August 2, 2020, 2:36pm UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/17 "2020-08-02T14:36:07Z")
My code (on Colab Tesla T4 GPU) seems to work despite the warning. What is the impact of this bug? ---
**Author:** ![ptrblck](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ptrblck/32/1823_2.png) [@ptrblck](https://discuss.pytorch.org/u/ptrblck)\ **Post date:** [August 3, 2020, 6:46am UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/18 "2020-08-03T06:46:00Z")
The `sm_70` argument _should_ add the sass and ptx to the binaries. The ptx would make the code executable for your T4 (compute capability 7.5), but might not yield the optimal performance. ---
**Author:** ![malfet](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/malfet/32/27485_2.png) [@malfet](https://discuss.pytorch.org/u/malfet)\ **Post date:** [August 4, 2020, 8:45pm UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/19 "2020-08-04T20:45:19Z")
To the best of my knowledge, there are no sm\_70 ptx in pytorch-1.6 binaries, but sm\_70 ISA is compatible with sm\_75, although some kernels might not launch, if I’m understating this correctly (from [1. Turing Compatibility — Turing Compatibility Guide 12.3 documentation](https://docs.nvidia.com/cuda/turing-compatibility-guide/index.html) ): > The Turing architecture is based on Volta’s Instruction Set Architecture ISA 7.0, extending it with new instructions. As a consequence, any binary that runs on Volta will be able to run on Turing (forward compatibility), but a Turing binary will not be able to run on Volta. Please note that Volta kernels using more than 64KB of shared memory (via the explicit opt-in, see CUDA C++ Programming Guide) will not be able to launch on Turing, as they would exceed Turing’s shared memory capacity. ---
**Author:** ![luiszeni](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/luiszeni/32/34535_2.png) [@luiszeni](https://discuss.pytorch.org/u/luiszeni)\ **Post date:** [February 10, 2021, 12:33pm UTC](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003/20 "2021-02-10T12:33:16Z")
This removed the warnings for me. Tks [Next page](https://discuss.pytorch.org/t/pytorch-1-6-tesla-t4-with-cuda-capability-sm-75-is-not-compatible/91003.md?page=2)