issues at runtime - --use_msa_server

[YogBN]

Dear Jeremy,
First of all please accept my gratitude for your fine work. In terms of the issue, I went ahead and installed Boltz-1 in a fresh virtual environment with conda. I then tried to model an multimer with RNA with a test.yaml but when I attempted to predict the structure of the multimer, the runtime failed saying "Error: No such option--use_msa_server". In the yaml file, I have commented out the msa lines for both protein and RNA, since I want the MSA from MMSeqs2 web server.

Here is the command I used:

Boltz predict --devices 1 --recycling_steps 10 --diffussion_samples 10 --cache "path to cache directory" --out_dir "path to output directory" --accelerator gpu --use_msa_server True --msa_server_url https://api.colabfold.com --msa_pairing_strategy greedy

This results in the aforementioned error. The yaml file is as follows:

version: 1
sequences:
  - protein:
      id: [A,B,C,D,E,F,G,H]
      sequence: GDTPSNPLRPIADDTIDHASHTPGSVSSAFILEAMVNVISGPKVLMKQIPIWLPLGVADQKTYSFDSTTAAIMLASYTITHFGKATNPLVRVNRLGPGIPDHPLRLLRIGNQAFLQEFVLPPVQLPQYFTFDLTALKLITQPLPAATWTDDTPTGSN

     msa:

  - rna:
      id: [I,J,K,L,M,N,O,P]
      sequence:UACAUUCCCAGCCUUUGUAGUGUUUUCGCCAAGCA

     msa: 

Can you please provide me with some guidance why this standard use case is failing?

I look forward to hearing from you.

With warm regards,
Yogi

[sangyeon-hits]

Are you using the most recent commit of the Boltz code? The error looks trivial that the option is not considered valid.
And also, the option --use_msa_server is a flag, meaning that you only have to command like

boltz predict [args ...] --use_msa_server [other_options ...]

not like --use_msa_server True.

[zhuowenzhao]

I can attest --use_msa_server True is not one of the valid options for boltz==0.2.0.

[zhuowenzhao]

However, git pull the latest commit and build from there works!