Closed RuiWang1998 closed 2 years ago
This pull request reduces the GRAM requirements for the OmegaFold
It runs sequences as long as 4096 on a single NVIDIA A100 card with 80 GB memory without hitting the limit with --subbatch_size set to 128.
--subbatch_size
This pull request reduces the GRAM requirements for the OmegaFold
It runs sequences as long as 4096 on a single NVIDIA A100 card with 80 GB memory without hitting the limit with
--subbatch_sizeset to 128.