multimer installation weird behaviour

[eguarnera]

Thanks for the amazing work you put together. I'm trying to run openfold multimer branch with the following command

python3 run_pretrained_openfold.py
./FASTA
data/pdb_mmcif/mmcif_files/
--uniref90_database_path data/uniref90/uniref90.fasta
--mgnify_database_path data/mgnify/mgy_clusters_2022_05.fa
--pdb_seqres_database_path data/pdb_seqres/pdb_seqres.txt
--uniref30_database_path data/uniref30/UniRef30_2021_03
--uniprot_database_path data/uniprot/uniprot.fasta
--bfd_database_path data/bfd/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt
--jackhmmer_binary_path /SW/python/miniconda3/x86_64/envs/openfold_env/bin/jackhmmer
--hhblits_binary_path /SW/python/miniconda3/x86_64/envs/openfold_env/bin/hhblits
--hmmsearch_binary_path /SW/python/miniconda3/x86_64/envs/openfold_env/bin/hmmsearch
--hmmbuild_binary_path /SW/python/miniconda3/x86_64/envs/openfold_env/bin/hmmbuild
--kalign_binary_path /SW/python/miniconda3/x86_64/envs/openfold_env/bin/kalign
--config_preset "model_1_multimer_v3"
--model_device "cuda:0"
--output_dir ./

I'm observing a strange behaviour, the input fasta file contains three chains, the alignments are produced but saved into ./alignments/alignments/ and later on I get an error as the alignments cannot be found in ./alignments/

Any idea?

[2024-01-24 10:42:29,197] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect)
INFO:/cdata/TECHNO/OPENFOLD/multimer/openfold/openfold/utils/script_utils.py:Successfully loaded JAX parameters at openfold/resources/pa
rams/params_model_1_multimer_v3.npz...
INFO:/cdata/TECHNO/OPENFOLD/multimer/openfold/run_pretrained_openfold.py:Generating alignments for R1.1_HC...
INFO:/cdata/TECHNO/OPENFOLD/multimer/openfold/run_pretrained_openfold.py:Generating alignments for R1.1_LC...
INFO:/cdata/TECHNO/OPENFOLD/multimer/openfold/run_pretrained_openfold.py:Generating alignments for CDXXX...
Traceback (most recent call last):
File "/cdata/TECHNO/OPENFOLD/multimer/openfold/run_pretrained_openfold.py", line 473, in
main(args)
File "/cdata/TECHNO/OPENFOLD/multimer/openfold/run_pretrained_openfold.py", line 282, in main
feature_dict = generate_feature_dict(
File "/cdata/TECHNO/OPENFOLD/multimer/openfold/run_pretrained_openfold.py", line 154, in generate_feature_dict
feature_dict = data_processor.process_fasta(
File "/cdata/TECHNO/OPENFOLD/multimer/openfold/openfold/data/data_pipeline.py", line 1250, in process_fasta
chain_features = self._process_single_chain(
File "/cdata/TECHNO/OPENFOLD/multimer/openfold/openfold/data/data_pipeline.py", line 1166, in _process_single_chain
raise ValueError(f"Alignments for {chain_id} not found...")
ValueError: Alignments for R1.1_HC not found...

[jnwei]

Hi, thanks for your interest in our work and for raising this issue.

This error arises when a local path is specified for the output_dir rather than a full path. We'll have a patch to fix this issue in the next update.

For your current work, you can reuse the pre-computed alignments by using the --use_precomputed_alignments flag for the command, and pointing it to the directory with your computed alignments.