Open ParisNeo opened 1 year ago
Hi, not sure why it fails with conda. Are you getting this error on other versions of CUDA as well? I'm currently traveling and have a slow internet, so will check this later.
Take your time. The thing is this works fine on all my windows PCs but when I need it on my server, it fails and I'm obliged to use another binding or use it in CPU mode which is a big loss of performance.
By the way, Is it possible to have a function that shows the model information (model size in memory, vram occupation etc)? And it also would be cool to have a function that returns the seed that was used for the current message (for example if I set seed=-1 I want to be able to get the actual seed that was used so that I can make the generation repeatable).
Hi there. I have installed cuda toolkit 11.7 using conda. Also gcc (tested multiple versions). Here with 9.4.0 I get this error:
Is there any prevompiled version I can try to install directly? Or do you have an idea why this fails?