Is this a memory error?

[katemichie]

A large number (half?) of my AF2 runs running on standard docker install on AWS looking at complexes/multimers aren't working anymore when it used to be only 1 in 10 that failed. They aren't massive in size (total of 900 aa) but it seems maybe there is a memory problem with the sequence alignments. The instance simply shuts down and the log file isn't super helpful (although I only speak 'basic' computer talk). The errors vary but it is always at the MSA step. For example, below: Does this make sense to anyone?

I0227 06:20:39.545079 140374311003968 run_docker.py:255] I0227 06:20:39.545193 140374311003968 run_docker.py:255] - 06:18:32.515 INFO: Realigning 14726 HMM-HMM alignments using Maximum Accuracy algorithm I0227 06:20:39.545316 140374311003968 run_docker.py:255] I0227 06:20:39.545434 140374311003968 run_docker.py:255] - 06:20:38.284 ERROR: Error in /tmp/hh-suite/src/hhalignment.cpp:3539: MergeMasterSlave: I0227 06:20:39.545548 140374311003968 run_docker.py:255] I0227 06:20:39.545661 140374311003968 run_docker.py:255] - 06:20:38.284 ERROR: did not find 511 match states in sequence 1 of UniRef100_Q0F0S0. Sequence: I0227 06:20:39.545774 140374311003968 run_docker.py:255]

[jowodo]

I encounter a similar error. Did you find anything to circumvent this error?

[jowodo]

This could be related to https://github.com/soedinglab/hh-suite/issues/277

[jdmontenegro]

Any news on this front? I am running into the same issue with my predictions. Using alphafold from singularity. v 2.3.2