Open bzhousd opened 6 months ago
you can set chunk_size to 512 in https://github.com/TencentAI4S/tfold/blob/master/projects/tfold_ag/predict.py#L67 This can save a lot of memory at the expense of computation time
We will expose this interface to users in the next update (This month)
Thank you for your brilliant work. I have a question about the size limitation of antigen in tFold-Ag. I got an error message, "CUDA out of memory." when I used a long antigen sequence. I tried to reduce the length of the sequence, I found that the limitation is 350-400 AAs when I use Nvidia A100, which has 40GB GPU memory. I have ~30 protein with long antigen, I wonder if you have any suggestions to reduce the GPU memory usage. Thanks