PyTorch pip installation not working

Trying to install PyTorch with the given command from the homepage with Stable, Windows, Pip, Python and Cuda 11.3.

The command I get is
pip3 install torch==1.10.0+cu113 torchvision==0.11.1+cu113 torchaudio===0.10.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html

I get the error

ERROR: Could not find a version that satisfies the requirement torch==1.10.0+cu113 (from versions: 0.1.2, 0.1.2.post1, 0.1.2.post2)
ERROR: No matching distribution found for torch==1.10.0+cu113

I have Python 3.10.0 installed, in 64 bit (same as my system) with Pip installed with it.
I have CUDA 11.5 installed from NVIDIA’s site.

I have looked multiple places and found no solution, please help.

The issue is with your python version. You need python 3.8 because the repositories aren’t yet available for python 3.10

3 Likes

Ah, alright. I didn’t know this. Will try again with Python 3.8. Thanks.

I am using Python 3.8 with CUDA 11.5 and still get the same error.

Try to update pip and rerun the command.

I tried upgrading pip and installing with python 3.7. But it did not work.

What kind of error are you seeing?

PyTorch 1.10 does not support CUDA 11.5. Lower your CUDA to 11.3.1

PyTorch 1.10 should support CUDA 11.5 in source builds. While the binaries are built with 10.2 and 11.3, the nightlies are also built with 11.5.
The local CUDA toolkit won’t be used unless you are building PyTorch from source or a custom CUDA extension, so @sohani wouldn’t need to downgrade the local CUDA toolkit unless one of the previous use cases is used.