Torch=1.1.0 is not compatible with cuda 10.2

Hi
I need to have torch==1.1.0 and before I used

# CUDA 10.0
conda install pytorch==1.1.0 torchvision==0.3.0 cudatoolkit=10.0 -c pytorch

But now I want to install it again in another virtual environment and get compatibility error. This error says that my cuda version is 10.2.

Also, I tried

pip install torch==1.1.0 torchvision==0.3.0 -f https://download.pytorch.org/whl/cu90/torch_stable.html

but I recieved this error:

ERROR: Could not find a version that satisfies the requirement torch==1.1.0 (from versions: 0.1.2, 0.1.2.post1, 0.1.2.post2, 1.4.0, 1.5.0, 1.5.1, 1.6.0)
ERROR: No matching distribution found for torch==1.1.0

I can not find the solution.
Is there any suggestions?

Best Regards

I think the PyTorch 1.1.0 binaries were only built with CUDA9.0 and CUDA10.0 as seen here (and cpu of course).
Why do you need this particular PyTorch version and would you be able to install the latest one (1.6) with CUDA10.2?

Thanks for your reply.
I want to run an open source code. One of its requirements is torch==1.1.0.

What should I do now?

If you need to run version 1.1.0, you would need to install it with CUDA9.0 or 10.0.
Alternatively, if you need PyTorch 1.1.0 and CUDA10.2 you would need to build form source.

How should I downgrade my cuda version from 10.2 to 10?

Also, I could not understand what is the meaning of source in the second suggestion.

You can find the install commands in my link from the previous post.

I checked it out but I think it relates to install torch in different versions of cuda.
I can not find anything about downgrading version of cuda, am I wrong?

Yes, install the previous version of CUDA

Thanks.
Please guide me how I do it.

Could you post the error message you are seeing after running:

conda install pytorch==1.1.0 torchvision==0.3.0 cudatoolkit=10.0 -c pytorch