jwohlwend
Misgenerate the filename
Dear Developer,
I have two workstations. I set up the calculation on the first computer and transferred the data and conditions to the second computer(Generated requirements.txt at the first computer and transferred to the second computer) . I tried to generate protein-ligand interactions on Boltz-1 under Python in WSL.
The calculations work well on both computers. However, the second computer generates the results with filenames like 250219_H5_model_0.cif.
The first computer generated files named 250217_005_047_model_0.cif, as I intended in my scripts.
I suspect that during communication with the MSA server (specifically in the directory D:\python\boltz\data\results\250219_007\boltz_results_250212_003_001\msa\H5_unpaired_tmp_env), the H5_unpaired_tmp_env file appears, and the "H5" part of the name is being incorporated into the prediction result filename.
I also tried to generate results without scripts using these commands. (boltz predict /mnt/d/python/boltz/data/yaml/250212_001/250212_003_001.yaml --out_dir /mnt/d/python/boltz/data/results/250219_002/ --use_msa_server --msa_server_url https://api.colabfold.com ) But the result is the same. It generates H5 or H9 H** file names.
I'm really confused if the computer spec. effects the results.
The first computer : Windows 11 Pro, Ryzen7 5600G, RTX3060 12GB The second computer : Windows server 2025, Xeon 4410Y, Quadro A4000 16GB
I am experiencing the same issue with strange names or predictions that include H*. Additionally, it seems to be overwriting some files. I originally had 266 YAML files in the directory, but I only received 21 predictions, even though the progress bar indicated that all 266 samples were processed.
This issue appeared recently, likely due to a recent PR. A few weeks ago, everything was working fine. I don’t believe it is related to MSA, as I am using the same MSA as before.
I think I found the problem.
Sample name is defined here: https://github.com/jwohlwend/boltz/blob/main/src/boltz/data/parse/schema.py#L868
Which should come from the function parameter here: https://github.com/jwohlwend/boltz/blob/main/src/boltz/data/parse/schema.py#L459
But recently someone added this code here: https://github.com/jwohlwend/boltz/blob/main/src/boltz/data/parse/schema.py#L654
Which overwrites variable name
I think I found the problem. Sample name is defined here: https://github.com/jwohlwend/boltz/blob/main/src/boltz/data/parse/schema.py#L868 Which should come from the function parameter here: https://github.com/jwohlwend/boltz/blob/main/src/boltz/data/parse/schema.py#L459 But recently someone added this code here: https://github.com/jwohlwend/boltz/blob/main/src/boltz/data/parse/schema.py#L654 Which overwrites variable
name
Thank you for your reply. I understand this situation. Because the Boltz-1 on the 1st computer cloned on Janauary, and the 2nd computer was cloned last week. I guess the code has been changed. Thank you for your kind help. I wil wait the code would be updated.
It was fixed here: https://github.com/jwohlwend/boltz/commit/bed74738cd5cc74327c51ee7368e942c2e234bf6