Hi, I am trying to install pytorch with the specific CUDA version of 11.7 and I can only use pip for now. But the recommended command on the official website for Linux-pip-python-cuda 11.7 is a general command “pip3 install torch torchvision torchaudio” and this command always gave me CUDA version of 11.4 which is not what I want. However, when I tried to specify the version based on other commands suggested from this forum or other forum (e,g, github, StackOverflow etc.), it always said something about the 403 forbidden error and no matching distribution found for torch==1.13.1+cu117. I have attached a screenshot as the example. Could anybody please help me?
Thank you so much for the reply. I followed your suggestion, created a clean environment and executed pip install torch only, it still gave me CUDA version of 11.4 (also it seems like I installed torch-1.13.1+computecanada in which computecanada is a totally strange word to me; I don’t know if this could have something to do with my struggling). I attached screenshots for your convenience.
It seems you might have added a computecanada mirror to your pip environment?
I don’t know what exactly computecanada is but it seems they are hosting some wheels as shown e.g. here and here.
Check what your ~/.config/pip/pip.conf or ~/.pip/pip.conf shows and if these index urls were added there.
I’m not familiar with these custom wheels by computecanada and thus cannot recommend installing and using them.
To install our binaries, visit our website, select the desired config from the install matrix, and run the created command in your setup.
my pip version:
pip 22.3.1
Below is my command to install cu11.7:
pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu117
