search

hpcaitech / FastFold

Optimizing AlphaFold Training and Inference on GPU Clusters
Apache License 2.0
556 stars 84 forks source link

Invalid resolution format: ['?'] #171

Closed HF06 closed 1 year ago

HF06 commented 1 year ago

Hi, when i run the multimer inference, i have encountered "Invalid resolution format: ['?']" error. could anybody help me to deal with the error issues? [03/30/23 09:19:58] INFO colossalai - root - INFO: Found an exact template match 3buk_B.
INFO colossalai - root - INFO: Reading PDB entry from data/pdb_mmcif/mmcif_files/1bnd.cif. Query:
MSILFYVIFLAYLRGIQGNNMDQRSLPEDSLNSLIIKLIQADILKNKLSKQMVDVKENYQSTLPKAEAPREPERGGPAKSAFQPVIAMDTELLRQQRRYNSPRVLLSDSTPLEPPPLYL MEDYVGSPVVANRTSRRKRYAEHKSHRGEYSVCDSESLWVTDKSSAIDIRGHQVTVLGEIKTGNSPVKQYFYETRCKEARPVKNGCRGIDDKHWNSQCKTSQTYVRALTSENNKLVGWR WIRIDTSCVCALSRKIGRT, template:
YAEHKSHRGEVSVCDSESLWVTDKSSAIDIRGHQVTVLGEIKTQNSPVKQYFYETRCKEARPVKNGCRGIDDKHWNSQCKTSQTYVRALTSENNKLVGWRWIRIDTSCVCALSRKIGRT INFO colossalai - root - INFO: Invalid resolution format: ['?']
INFO colossalai - root - INFO: Found an exact template match 1bnd_B.

And this error leads to another error "torch.multiprocessing.spawn.ProcessExitedException: process 1 terminated with signal SIGSEGV" in inference_multimer_model (inference.py line293: torch.multiprocessing.spawn(inference_model, nprocs=args.gpus, args=(args.gpus, result_q, batch, args)))

Gy-Lu commented 1 year ago

Hi, as I know, this error is usually raised when the input is not valid or the losing part of databases.

HF06 commented 1 year ago

Hi, as I know, this error is usually raised when the input is not valid or the losing part of databases.

Thank you for your reply. I wonder what format of input should be. My input is like following:

AAA63231.1 neurotrophin-3 [Homo sapiens] MSILFYVIFLAYLRGIQGNNMDQRSLPEDSLNSLIIKLIQADILKNKLSKQMVDVKENYQSTLPKAEAPR EPERGGPAKSAFQPVIAMDTELLRQQRRYNSPRVLLSDSTPLEPPPLYLMEDYVGSPVVANRTSRRKRYA EHKSHRGEYSVCDSESLWVTDKSSAIDIRGHQVTVLGEIKTGNSPVKQYFYETRCKEARPVKNGCRGIDD KHWNSQCKTSQTYVRALTSENNKLVGWRWIRIDTSCVCALSRKIGRT AAA63231.1 neurotrophin-3 [Homo sapiens] MSILFYVIFLAYLRGIQGNNMDQRSLPEDSLNSLIIKLIQADILKNKLSKQMVDVKENYQSTLPKAEAPR EPERGGPAKSAFQPVIAMDTELLRQQRRYNSPRVLLSDSTPLEPPPLYLMEDYVGSPVVANRTSRRKRYA EHKSHRGEYSVCDSESLWVTDKSSAIDIRGHQVTVLGEIKTGNSPVKQYFYETRCKEARPVKNGCRGIDD KHWNSQCKTSQTYVRALTSENNKLVGWRWIRIDTSCVCALSRKIGRT

Are there any problems about my input. Inference of monomer is ok. It may not be due to the losing part of databases.

Gy-Lu commented 1 year ago

I recommend you add > before the tagline and not using special char in the tag, such as:

>AAA63231_1
MSILFYVIFLAYLRGIQGNNMDQRSLPEDSLNSLIIKLIQADILKNKLSKQMVDVKENYQSTLPKAEAPR
EPERGGPAKSAFQPVIAMDTELLRQQRRYNSPRVLLSDSTPLEPPPLYLMEDYVGSPVVANRTSRRKRYA
EHKSHRGEYSVCDSESLWVTDKSSAIDIRGHQVTVLGEIKTGNSPVKQYFYETRCKEARPVKNGCRGIDD
KHWNSQCKTSQTYVRALTSENNKLVGWRWIRIDTSCVCALSRKIGRT
>AAA63231_2
MSILFYVIFLAYLRGIQGNNMDQRSLPEDSLNSLIIKLIQADILKNKLSKQMVDVKENYQSTLPKAEAPR
EPERGGPAKSAFQPVIAMDTELLRQQRRYNSPRVLLSDSTPLEPPPLYLMEDYVGSPVVANRTSRRKRYA
EHKSHRGEYSVCDSESLWVTDKSSAIDIRGHQVTVLGEIKTGNSPVKQYFYETRCKEARPVKNGCRGIDD
KHWNSQCKTSQTYVRALTSENNKLVGWRWIRIDTSCVCALSRKIGRT

By the way, using the same tag would cause some undefined behavior. If you got other problems, plz let me know.

HF06 commented 1 year ago

Thanks. I have resolved my problems. It seemed to be the environment problems. Yes, using the same tag results to an output of monomer maybe because of deletion of the alignments.

HF06 commented 1 year ago

BTW, I have run the training script on PDB dataset, it seems lacking a required argument: --obsolete_pdbs_file_path.