Zhixin Dou
Zhixin Dou
> Monomer supports bf16, but multimer doesn't yet. Maybe we will support it in the future. How to use bf16 or fp32 to reduce memory use in GPU calculation? Thanks...
I met the same problem.
In case of any further issue like this one, here is the solution: Pay attention to your cudatoolkit path, by "which nvcc", if it use the other version of cudatoolkit....
find you cudatoolkit location, which nvcc. Make sure you are calling the cudatoolkit=11.3
感谢你的博客，带我认识了一个新的世界一角。
Here is an example code for resumable prediction if anyone need it. Put this code block after 134 line in scripts/fold.py. ``` if os.path.exists(args.pdb): predicted_proteins = set() for filename in...