I notice on the PyTorch installation page that there is only an option to select CUDA 11.1 – due to infrastructure constraints we only have access to CUDA 11.0. Will this cause any issues with PyTorch 1.9.0?
If you install a binary package of PyTorch (e.g., from pip, conda), it should use its own version of CUDA distinct from the install on your system. (Only the driver version of your install would be used.) So there shouldn’t be any issues here.
@eqy I have heard that if your Docker or package CUDA version is greater than the system version, there could be problems. Is that true?
If the drivers are compatible, then there shouldn’t be any issues. Of course, if you see something unexpected, please report the problem here.
Well, I would like to report a problem.
Here is the error:
ERROR: Could not find a version that satisfies the requirement torch==1.9.0+cu11 1 (from versions: 0.4.1, 0.4.1.post2, 1.0.0, 1.0.1, 1.0.1.post2, 1.1.0, 1.2.0, 1 .3.0, 1.3.1, 1.4.0, 1.5.0, 1.5.1, 1.6.0, 1.7.0, 1.7.1, 1.8.0, 1.8.1, 1.9.0, 1.9. 1, 1.10.0)
ERROR: No matching distribution found for torch==1.9.0+cu111
While I am installing from my labmates given conda env .yml file. Which means that he already has this combination installed. Then why I am getting this error.
It’s working for me:
pip install torch==1.9.0+cu111 -f https://download.pytorch.org/whl/cu111/torch_stable.html Looking in indexes: https://pypi.org/simple, https://pypi.ngc.nvidia.com Looking in links: https://download.pytorch.org/whl/cu111/torch_stable.html Collecting torch==1.9.0+cu111 Downloading https://download.pytorch.org/whl/cu111/torch-1.9.0%2Bcu111-cp38-cp38-linux_x86_64.whl (2041.3 MB) |████████████████████████████████| 2041.3 MB 87 kB/s [...]
I was trying to create conda env by .yml and having that error. Then I installed it directly by command and it worked. I don’t know the reason but the error was unexpected.