Erreur alphafold2

Bonjour,

J'ai l'erreur suivante lors de l'utilisation d'alphafold2 (version 2.1.1 et 2.2.3) en mode sbatch en suivant le modèle:
https://ifb-elixirfr.gitlab.io/cluster/doc/software/alphafold2/
Des drivers semblent manquer. Comment les ajouter?
Merci beaucoup pour votre aide,

Romain

srun: job 27758647 queued and waiting for resources
srun: job 27758647 has been allocated resources
I1206 14:12:20.364980 46987089804992 templates.py:857] Using precomputed obsolete pdbs /shared/bank/alphafold2/current/pdb_mmcif/obsolete.dat.
I1206 14:12:21.216653 46987089804992 xla_bridge.py:230] Unable to initialize backend 'tpu_driver': Not found: Unable to find driver in registry given worker: 
2022-12-06 14:12:21.218240: W external/org_tensorflow/tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcuda.so.1'; dlerror: libcuda.so.1: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/lib
2022-12-06 14:12:21.218287: W external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:269] failed call to cuInit: UNKNOWN ERROR (303)
I1206 14:12:21.223239 46987089804992 xla_bridge.py:230] Unable to initialize backend 'gpu': Failed precondition: No visible GPU devices.
I1206 14:12:21.223954 46987089804992 xla_bridge.py:230] Unable to initialize backend 'tpu': Invalid argument: TpuPlatform is not available.
W1206 14:12:21.224107 46987089804992 xla_bridge.py:235] No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)
/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/bin/run_alphafold.sh: line 3: 27488 Killed                  python /shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/bin/run_alphafold.py "$@"
slurmstepd: error: Detected 1 oom-kill event(s) in StepId=27758647.0 cgroup. Some of your processes may have been killed by the cgroup out-of-memory handler.
srun: error: cpu-node-37: task 0: Out Of Memory

Bonjour Romain,

Le messageNo GPU/TPU found, falling back to CPU est normal puisque le job s’exécute sur un nœud standard et non un nœud "GPU".
Pour accéder au nœud GPU, il faut en faire la demande au préalable (ici-même).
Je peux vous y autoriser mais avant je vous invite à demander la création d'un projet via le portail de gestion de compte: MyUi
En effet, le projet "fair_bioinfo_2021" est utile le temps de la formation mais par la suite nous demandons à créer vos propres projets.
Une fois ce projet validé, vous pourrez y copier vos données (de l'espace "fair_bioinfo_2021" vers votre nouveau espace projet).
Ce projet vous permets de bénéficier d'un espace de stockage (/shared/projects/) et permets de suivre l'utilisation des ressources (stockage, cpu, etc).
C'est une bonne pratique pour assurer une bonne gestion des données (plus d'infos: The concept of project - IFB Core Cluster Documentation).
Vous pouvez demander plusieurs projets si besoin.

L'erreur Out Of Memory indique un manque de mémoire.
Vous pouvez aussi réessayer en augmentant la quantité de mémoire demandé (en doublant par exemple: --mem=64G et si ca suffit pas, en re-doublant, etc)

Bonne journée

Bonjour,

Merci beaucoup pour votre retour
J'ai demandé la création d'un projet " tfheterodimer_3dpred " hier et j'ai eu le message la confirmation de création cet après-midi.
Cependant /shared/projects/tfheterodimer_3dpred n'apparaît pas encore sur le cluster (il y a peut etre un délai?)
J'ai utilisé l'espace "fair_bioinfo_2021" en attendant, pour tester l'outil. Mes excuses pour cela.
Puis-je avoir accès à un nœud GPU pour le projet tfheterodimer_3dpred ?

Bonne fin de journée,
Romain BM

Bonjour Romain,

Je viens de vous donner l'accès à la partition GPU pour le projet "tfheterodimer_3dpred".

Il y a eu petit couac lors de la création du projet. C'est corrigé et le dossier apparaît maintenant.

PS: la documentation pour AlphaFold va être mise à jour d'ici peu.

Bonne journée

Bonjour,
Merci!
J'ai l'erreur suivante lorsque je lance sbatch my_fold.sh

sbatch: error: Batch job submission failed: Invalid account or account/partition combination specified

J'ai pourtant bien précisé
#SBATCH -A tfheterodimer_3dpred
dans mon script

J'ai peut-être raté une étape?

Romain

Bonjour Romain,

My Bad. J'ai fait une petite erreur lors de l'association (corrigé). Pouvez-vous réessayer ?

merci!
J'ai réessayé et la commande s'est lancée.
Mais elle ne va pas jusqu'au bout et j'ai l'erreur suivante.

I1208 12:50:18.265425 47972528590528 tpu_client.py:54] Starting the local TPU driver.
I1208 12:50:18.375782 47972528590528 xla_bridge.py:212] Unable to initialize backend 'tpu_driver': Not found: Unable to find driver in registry given worker: local://
I1208 12:50:19.362817 47972528590528 xla_bridge.py:212] Unable to initialize backend 'tpu': Invalid argument: TpuPlatform is not available

Les droits pour la partition GPU ne sont peut être pas passés?

C'est au-dessus de mes compétences. @team.alphafold une idée ?

PS: Le job s’exécute bien sur un noeud GPU.

Pouvez-vous nous donner tous les paramètres utilisés pour lancer le job ?

Oui,
je lance le code avec la commande sbatch my_fold_mult.sh

le code est:

#!/bin/bash
#
#SBATCH -A tfheterodimer_3dpred
#SBATCH -p gpu
#SBATCH --gres=gpu:1g.5gb:1
#SBATCH --cpus-per-task=10
#SBATCH --mem=50G

module load alphafold/2.2.3

mkdir -p /tmp/$USER_alphafold

srun run_alphafold.sh --fasta_paths=/shared/home/rblanc/tests/AlphaFold/LFY_UFO.fasta \
    --output_dir=/shared/projects/tfheterodimer_3dpred/AlphaFold \
    --model_preset=multimer \
    --db_preset=full_dbs \
    --data_dir=/shared/bank/alphafold2/current \
    --uniref90_database_path=/shared/bank/alphafold2/current/uniref90/uniref90.fasta \
    --mgnify_database_path=/shared/bank/alphafold2/current/mgnify/mgy_clusters_2018_12.fa \
    --template_mmcif_dir=/shared/bank/alphafold2/current/pdb_mmcif/mmcif_files \
    --max_template_date=2020-05-14 \
    --obsolete_pdbs_path=/shared/bank/alphafold2/current/pdb_mmcif/obsolete.dat \
    --bfd_database_path=/shared/bank/alphafold2/current/bfd/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt \
    --uniclust30_database_path=/shared/bank/alphafold2/current/uniclust30/uniclust30_2018_08/uniclust30_2018_08 \
    --use_gpu_relax=1 \
    --pdb_seqres_database_path=/shared/bank/alphafold2/current/pdb_seqres \
    --uniprot_database_path=/shared/bank/alphafold2/current/uniprot

et l'erreur complète est:

I1208 12:50:17.492459 47972528590528 templates.py:857] Using precomputed obsolete pdbs /shared/bank/alphafold2/current/pdb_mmcif/obsolete.dat.
I1208 12:50:18.265425 47972528590528 tpu_client.py:54] Starting the local TPU driver.
I1208 12:50:18.375782 47972528590528 xla_bridge.py:212] Unable to initialize backend 'tpu_driver': Not found: Unable to find driver in registry given worker: local://
I1208 12:50:19.362817 47972528590528 xla_bridge.py:212] Unable to initialize backend 'tpu': Invalid argument: TpuPlatform is not available.
I1208 12:50:22.875974 47972528590528 run_alphafold.py:376] Have 25 models: ['model_1_multimer_v2_pred_0', 'model_1_multimer_v2_pred_1', 'model_1_multimer_v2_pred_2', '
model_1_multimer_v2_pred_3', 'model_1_multimer_v2_pred_4', 'model_2_multimer_v2_pred_0', 'model_2_multimer_v2_pred_1', 'model_2_multimer_v2_pred_2', 'model_2_multimer_
v2_pred_3', 'model_2_multimer_v2_pred_4', 'model_3_multimer_v2_pred_0', 'model_3_multimer_v2_pred_1', 'model_3_multimer_v2_pred_2', 'model_3_multimer_v2_pred_3', 'mode
l_3_multimer_v2_pred_4', 'model_4_multimer_v2_pred_0', 'model_4_multimer_v2_pred_1', 'model_4_multimer_v2_pred_2', 'model_4_multimer_v2_pred_3', 'model_4_multimer_v2_p
red_4', 'model_5_multimer_v2_pred_0', 'model_5_multimer_v2_pred_1', 'model_5_multimer_v2_pred_2', 'model_5_multimer_v2_pred_3', 'model_5_multimer_v2_pred_4']
I1208 12:50:22.876127 47972528590528 run_alphafold.py:393] Using random seed 8481300240475865 for the data pipeline
I1208 12:50:22.876272 47972528590528 run_alphafold.py:161] Predicting LFY_UFO
I1208 12:50:22.882764 47972528590528 pipeline_multimer.py:210] Running monomer pipeline on chain A: sp|Q00958|LFY_ARATH Protein LEAFY OS=Arabidopsis thaliana OX=3702 G
N=LFY PE=1 SV=2
I1208 12:50:22.882892 47972528590528 jackhmmer.py:133] Launching subprocess "/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/bin/jackhmmer -o /dev/null -A /tm
p/tmpgqarqlop/output.sto --noali --F1 0.0005 --F2 5e-05 --F3 5e-07 --incE 0.0001 -E 0.0001 --cpu 8 -N 1 /tmp/tmpmlv03i3c.fasta /shared/bank/alphafold2/current/uniref90
/uniref90.fasta"
I1208 12:50:22.911369 47972528590528 utils.py:36] Started Jackhmmer (uniref90.fasta) query
I1208 12:57:00.120888 47972528590528 utils.py:40] Finished Jackhmmer (uniref90.fasta) query in 397.209 seconds
I1208 12:57:00.135301 47972528590528 jackhmmer.py:133] Launching subprocess "/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/bin/jackhmmer -o /dev/null -A /tm
p/tmpghsvvxfe/output.sto --noali --F1 0.0005 --F2 5e-05 --F3 5e-07 --incE 0.0001 -E 0.0001 --cpu 8 -N 1 /tmp/tmpmlv03i3c.fasta /shared/bank/alphafold2/current/mgnify/m
gy_clusters_2018_12.fa"
I1208 12:57:00.146211 47972528590528 utils.py:36] Started Jackhmmer (mgy_clusters_2018_12.fa) query
I1208 13:01:44.694318 47972528590528 utils.py:40] Finished Jackhmmer (mgy_clusters_2018_12.fa) query in 284.506 seconds
I1208 13:01:45.024384 47972528590528 hmmbuild.py:121] Launching subprocess ['/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/bin/hmmbuild', '--hand', '--amino
', '/tmp/tmp_ypfr_ks/output.hmm', '/tmp/tmp_ypfr_ks/query.msa']
I1208 13:01:45.080546 47972528590528 utils.py:36] Started hmmbuild query
I1208 13:01:45.510279 47972528590528 hmmbuild.py:128] hmmbuild stdout:
# hmmbuild :: profile HMM construction from multiple sequence alignments
# HMMER 3.3.2 (Nov 2020); http://hmmer.org/
# Copyright (C) 2020 Howard Hughes Medical Institute.
# Freely distributed under the BSD open source license.
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# input alignment file:             /tmp/tmp_ypfr_ks/query.msa
# output HMM file:                  /tmp/tmp_ypfr_ks/output.hmm
# input alignment is asserted as:   protein
# model architecture construction:  hand-specified by RF annotation
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

# idx name                  nseq  alen  mlen eff_nseq re/pos description
#---- -------------------- ----- ----- ----- -------- ------ -----------
1     query                  593  1232   420     3.16  0.590 

# CPU time: 0.42u 0.00s 00:00:00.42 Elapsed: 00:00:00.42


stderr:


I1208 13:01:45.510342 47972528590528 utils.py:40] Finished hmmbuild query in 0.429 seconds
I1208 13:01:45.511947 47972528590528 hmmsearch.py:103] Launching sub-process ['/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/bin/hmmsearch', '--noali', '--c
pu', '8', '--F1', '0.1', '--F2', '0.1', '--F3', '0.1', '--incE', '100', '-E', '100', '--domE', '100', '--incdomE', '100', '-A', '/tmp/tmppjrtdui6/output.sto', '/tmp/tm
ppjrtdui6/query.hmm', '/shared/bank/alphafold2/current/pdb_seqres']
I1208 13:01:45.539009 47972528590528 utils.py:36] Started hmmsearch (pdb_seqres) query
I1208 13:01:45.546266 47972528590528 utils.py:40] Finished hmmsearch (pdb_seqres) query in 0.007 seconds
Traceback (most recent call last):
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/bin/run_alphafold.py", line 422, in <module>
    app.run(main)
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/lib/python3.8/site-packages/absl/app.py", line 312, in run
    _run_main(main, args)
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/lib/python3.8/site-packages/absl/app.py", line 258, in _run_main
    sys.exit(main(argv))
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/bin/run_alphafold.py", line 398, in main
    predict_structure(
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/bin/run_alphafold.py", line 172, in predict_structure
    feature_dict = data_pipeline.process(
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/lib/python3.8/site-packages/alphafold/data/pipeline_multimer.py", line 264, in process
    chain_features = self._process_single_chain(
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/lib/python3.8/site-packages/alphafold/data/pipeline_multimer.py", line 212, in _process_single_chain
    chain_features = self._monomer_data_pipeline.process(
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/lib/python3.8/site-packages/alphafold/data/pipeline.py", line 185, in process
    pdb_templates_result = self.template_searcher.query(msa_for_templates)
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/lib/python3.8/site-packages/alphafold/data/tools/hmmsearch.py", line 79, in query
    return self.query_with_hmm(hmm)
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.2.3/lib/python3.8/site-packages/alphafold/data/tools/hmmsearch.py", line 112, in query_with_hmm
    raise RuntimeError(
RuntimeError: hmmsearch failed:
stdout:


stderr:
Fatal exception (source file esl_buffer.c, line 1608):
failed to slurp /shared/bank/alphafold2/current/pdb_seqres

srun: error: gpu-node-01: task 0: Exited with exit code 1

Bonjour,
Il me semble que les partitions GPU 1g.5gb sont un peu limitées pour alphafold, surtout si le sujet n'est pas un tutoriel/essai.
Je suggère de passer avec 3g.20gb et voir si ça marche
J.C.H

Bonjour,
Merci beaucoup pour votre aide.
Le script a bien fonctionné pour prédire la structure d'une protéine avec la version 2.1.1 d'alphafold.
Par contre j'ai l'erreur suivante lorsque que j'essaie de prédire la structure de deux protéines ensembles (option multimer d'alphafold):

I1209 10:56:24.864889 47671831510720 templates.py:857] Using precomputed obsolete pdbs /shared/bank/alphafold2/current/pdb_mmcif/obsolete.dat.
I1209 10:56:25.454231 47671831510720 xla_bridge.py:230] Unable to initialize backend 'tpu_driver': Not found: Unable to find driver in registry given worker: 
I1209 10:56:26.377464 47671831510720 xla_bridge.py:230] Unable to initialize backend 'tpu': Invalid argument: TpuPlatform is not available.
I1209 10:56:32.566188 47671831510720 run_alphafold.py:384] Have 5 models: ['model_1_multimer', 'model_2_multimer', 'model_3_multimer', 'model_4_multimer', 'model_5_mul
timer']
I1209 10:56:32.566587 47671831510720 run_alphafold.py:397] Using random seed 1552155542698198349 for the data pipeline
I1209 10:56:32.567203 47671831510720 run_alphafold.py:150] Predicting LFY_UFO
I1209 10:56:32.573152 47671831510720 pipeline_multimer.py:210] Running monomer pipeline on chain A: sp|Q00958|LFY_ARATH Protein LEAFY OS=Arabidopsis thaliana OX=3702 G
N=LFY PE=1 SV=2
I1209 10:56:32.573355 47671831510720 jackhmmer.py:130] Launching subprocess "/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/bin/jackhmmer -o /dev/null -A /tm
p/tmp2mni4345/output.sto --noali --F1 0.0005 --F2 5e-05 --F3 5e-07 --incE 0.0001 -E 0.0001 --cpu 8 -N 1 /tmp/tmp1nf7ec2z.fasta /shared/bank/alphafold2/current/uniref90
/uniref90.fasta"
I1209 10:56:32.615034 47671831510720 utils.py:36] Started Jackhmmer (uniref90.fasta) query
I1209 11:01:37.167584 47671831510720 utils.py:40] Finished Jackhmmer (uniref90.fasta) query in 304.552 seconds
I1209 11:01:37.177792 47671831510720 jackhmmer.py:130] Launching subprocess "/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/bin/jackhmmer -o /dev/null -A /tm
p/tmp2rrb0bi9/output.sto --noali --F1 0.0005 --F2 5e-05 --F3 5e-07 --incE 0.0001 -E 0.0001 --cpu 8 -N 1 /tmp/tmp1nf7ec2z.fasta /shared/bank/alphafold2/current/mgnify/m
gy_clusters_2018_12.fa"
I1209 11:01:37.228314 47671831510720 utils.py:36] Started Jackhmmer (mgy_clusters_2018_12.fa) query
I1209 11:11:13.217339 47671831510720 utils.py:40] Finished Jackhmmer (mgy_clusters_2018_12.fa) query in 575.989 seconds
I1209 11:11:13.272267 47671831510720 hmmbuild.py:121] Launching subprocess ['/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/bin/hmmbuild', '--hand', '--amino
', '/tmp/tmp5g9z70hb/output.hmm', '/tmp/tmp5g9z70hb/query.msa']
I1209 11:11:13.322670 47671831510720 utils.py:36] Started hmmbuild query
I1209 11:11:13.866998 47671831510720 hmmbuild.py:128] hmmbuild stdout:
# hmmbuild :: profile HMM construction from multiple sequence alignments
# HMMER 3.3.2 (Nov 2020); http://hmmer.org/
# Copyright (C) 2020 Howard Hughes Medical Institute.
# Freely distributed under the BSD open source license.
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# input alignment file:             /tmp/tmp5g9z70hb/query.msa
# output HMM file:                  /tmp/tmp5g9z70hb/output.hmm
# input alignment is asserted as:   protein
# model architecture construction:  hand-specified by RF annotation
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

# idx name                  nseq  alen  mlen eff_nseq re/pos description
#---- -------------------- ----- ----- ----- -------- ------ -----------
1     query                  593  1232   420     3.16  0.590 

# CPU time: 0.53u 0.01s 00:00:00.54 Elapsed: 00:00:00.54


stderr:


I1209 11:11:13.867122 47671831510720 utils.py:40] Finished hmmbuild query in 0.544 seconds
I1209 11:11:13.867875 47671831510720 hmmsearch.py:103] Launching sub-process ['/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/bin/hmmsearch', '--noali', '--c
pu', '8', '--F1', '0.1', '--F2', '0.1', '--F3', '0.1', '--incE', '100', '-E', '100', '--domE', '100', '--incdomE', '100', '-A', '/tmp/tmpo3cu9fkw/output.sto', '/tmp/tm
po3cu9fkw/query.hmm', '/shared/bank/alphafold2/current/pdb_seqres']
I1209 11:11:13.895746 47671831510720 utils.py:36] Started hmmsearch (pdb_seqres) query
I1209 11:11:13.904050 47671831510720 utils.py:40] Finished hmmsearch (pdb_seqres) query in 0.008 seconds
Traceback (most recent call last):
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/bin/run_alphafold.py", line 427, in <module>
    app.run(main)
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/lib/python3.8/site-packages/absl/app.py", line 312, in run
    _run_main(main, args)
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/lib/python3.8/site-packages/absl/app.py", line 258, in _run_main
    sys.exit(main(argv))
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/bin/run_alphafold.py", line 403, in main
    predict_structure(
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/bin/run_alphafold.py", line 166, in predict_structure
    feature_dict = data_pipeline.process(
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/lib/python3.8/site-packages/alphafold/data/pipeline_multimer.py", line 266, in process
    chain_features = self._process_single_chain(
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/lib/python3.8/site-packages/alphafold/data/pipeline_multimer.py", line 212, in _process_single_chain
    chain_features = self._monomer_data_pipeline.process(
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/lib/python3.8/site-packages/alphafold/data/pipeline.py", line 176, in process
    pdb_templates_result = self.template_searcher.query(msa_for_templates)
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/lib/python3.8/site-packages/alphafold/data/tools/hmmsearch.py", line 79, in query
    return self.query_with_hmm(hmm)
  File "/shared/ifbstor1/software/miniconda/envs/alphafold-2.1.1/lib/python3.8/site-packages/alphafold/data/tools/hmmsearch.py", line 112, in query_with_hmm
    raise RuntimeError(
RuntimeError: hmmsearch failed:
stdout:


stderr:
Fatal exception (source file esl_buffer.c, line 1608):
failed to slurp /shared/bank/alphafold2/current/pdb_seqres

srun: error: gpu-node-02: task 0: Exited with exit code 1

la commande est:

#!/bin/bash
#
#SBATCH -A tfheterodimer_3dpred
#SBATCH -p gpu
#SBATCH --gres=gpu:3g.20gb:1
#SBATCH --cpus-per-task=10
#SBATCH --mem=50G

module load alphafold/2.1.1

mkdir -p /tmp/$USER_alphafold

srun run_alphafold.sh --fasta_paths=/shared/home/rblanc/tests/AlphaFold/LFY_UFO.fasta \
    --output_dir=/shared/projects/tfheterodimer_3dpred/AlphaFold \
    --model_preset=multimer \
    --db_preset=full_dbs \
    --data_dir=/shared/bank/alphafold2/current \
    --uniref90_database_path=/shared/bank/alphafold2/current/uniref90/uniref90.fasta \
    --mgnify_database_path=/shared/bank/alphafold2/current/mgnify/mgy_clusters_2018_12.fa \
    --template_mmcif_dir=/shared/bank/alphafold2/current/pdb_mmcif/mmcif_files \
    --max_template_date=2020-05-14 \
    --obsolete_pdbs_path=/shared/bank/alphafold2/current/pdb_mmcif/obsolete.dat \
    --bfd_database_path=/shared/bank/alphafold2/current/bfd/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt \
    --uniclust30_database_path=/shared/bank/alphafold2/current/uniclust30/uniclust30_2018_08/uniclust30_2018_08 \
    --pdb_seqres_database_path=/shared/bank/alphafold2/current/pdb_seqres \
    --uniprot_database_path=/shared/bank/alphafold2/current/uniprot

Romain

Bonjour,
Il y a eu des tests courant octobre avec alphafold pour des gros ensembles de données ~4.5K acides aminés et en mode multimer, il me semble. Mais c'était avec Alphafold 2.2.3.
De plus, 2 instances 3g.20gb ont été utilisées (je ne sais pas si le logiciel a réellement pu en faire usage mais le calcul s'est terminé correctement à priori). Par ailleurs, ce calcul a duré très longtemps.
Vous pouvez essayer ces 2 options : la version plus récente ou demander 2 GPU.

Bonne journée,
J.C.H