Have you tried: import torch print(torch.cuda.is_available())
Look at the market share of video cards with more than 100GB of Vram.
Have you tried: import torch print(torch.cuda.is_available())