Encountering Rate Limit Despite MSA Server Not Being Used

[drb450]

Expected Behavior

I expect structures to be predicted without a rate limit.

Current Behavior

I have provided .a3m files already with the expectation I can avoid connecting to the server and getting a rate limit. This seems to work fine for the first 25 sequences and then I am provided a rate limit again of ~ 90 seconds between structures.

Steps to Reproduce (for bugs)

colabfold_batch --num-recycle 0 --num-models 1 fasta_file.fasta output_dir_with_a3m_files

ColabFold Output (for bugs)

2025-01-13 14:33:42,599 Sleeping for 8s. Reason: RATELIMIT... etc.

[punit-jha123]

I think your run is trying to get the templates from the server? you have to turn off the templates when folding