excessive RAM use after splitting partitions

60 views
Skip to first unread message

jeremy le luyer

unread,
Feb 28, 2019, 9:09:11 PM2/28/19
to EVidenceModeler-users
Hello,

I tested EVM to compile different genes prediction, however even after partitioning I end up with a very large memory usage (>200Go), which looks excessive. 
"""
partition_EVM_inputs.pl --genome "$GENOME" \
 --gene_predictions denovo.test.gff3 --protein_alignments homology.test.gff3 \
 --transcript_alignments transcript.test.gff3 \
--segmentSize 100000 --overlapSize 10000 --partition_listing partitions_list.out

write_EVM_commands.pl --genome "$GENOME" --weights "$WEIGHT" \
 --gene_predictions denovo.test.gff3 --protein_alignments homology.test.gff3 \
--transcript_alignments transcript.test.gff3 \
  --output_file_name "$OUTPUT" --partitions partitions_list.out >commands.list
"""

I tried running a single command to check how much RAM would be necessary:

"""
evidence_modeler.pl --genome sspace.final.scaffolds.fasta --gene_predictions denovo.gff3 --weights weight.txt --transcript_alignments transcript.gff3  --protein_alignments homology.gff3  --exec_dir scaffold1size567472/scaffold1size567472_450001-550000 > scaffold1size567472/scaffold1size567472_450001-550000/EVM_output.txt 2> scaffold1size567472/scaffold1size567472_450001-550000/EVM_output.txt.lo
"""

"""
=>> PBS: job killed: mem 211812812kb exceeded limit 209715200kb
"""

the files included in this partition have a reasonable size:

3.8 Feb 28 23:33 scaffold1size567472/scaffold1size567472_450001-550000/denovo.test.gff3
0 Mar  1 01:29 scaffold1size567472/scaffold1size567472_450001-550000/EVM_output.txt
171 Mar  1 01:29 scaffold1size567472/scaffold1size567472_450001-550000/EVM_output.txt.log
0 Feb 28 23:33 scaffold1size567472/scaffold1size567472_450001-550000/homology.test.gff3
100K Feb 28 23:33 scaffold1size567472/scaffold1size567472_450001-550000/sspace.final.scaffolds.fasta
17K Feb 28 23:33 scaffold1size567472/scaffold1size567472_450001-550000/transcript.test.gff3

Do you have any idea of what I might be doing wrong here ?

Thank you

J.

Brian Haas

unread,
Mar 1, 2019, 6:52:06 AM3/1/19
to jeremy le luyer, EVidenceModeler-users
Hi,

Can you send me the inputs that are causing this?   I'll figure out what the issue is.


best,

~brian

--
You received this message because you are subscribed to the Google Groups "EVidenceModeler-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to evidencemodeler-...@googlegroups.com.
To post to this group, send email to evidencemo...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/evidencemodeler-users/ca71bb37-3291-4831-a57f-59be82462577%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.


--
--
Brian J. Haas
The Broad Institute
http://broadinstitute.org/~bhaas

 

Brian Haas

unread,
Mar 1, 2019, 6:38:02 PM3/1/19
to jeremy le luyer, EVidenceModeler-users
Hi Jeremy,

Thanks for sending one of the examples.   I ran it like so on my laptop:

~/GITHUB/EVidenceModeler/evidence_modeler.pl --genome sspace.final.scaffolds.fasta --gene_predictions denovo.test.gff3 --weights weight.txt --transcript_alignments transcript.test.gff3 > evm.out

and it ran through in a few seconds w/o any trouble.

If you do the same, does it work for you?

It could be that this is not the small partition that's causing trouble. If there's a way to see which inputs are driving the RAM up, that's the one that we want to tackle.

best,

~brian


linz li

unread,
Jun 28, 2020, 8:56:28 AM6/28/20
to EVidenceModeler-users
Dear Brian, 

I meet a same problem when I run EVM, I split each file into very small file. but it showed >300Gb RAM, and got out of memory error. Did you guys fixed it?

Best
Linzhou

Brian Haas

unread,
Jun 28, 2020, 9:05:56 AM6/28/20
to linz li, EVidenceModeler-users
Hi,

Do you have a small example you can share with me that reproduces the problem?  My guess is that there's something off with the coordinates.  If I can nail it down, I can patch the code to prevent it in the future.

You can send files to:  bh...@broadinstitute.org

thx

--
You received this message because you are subscribed to the Google Groups "EVidenceModeler-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to evidencemodeler-...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages