How to reduce RAM usage of "herro inference"?

lbcb-sci / herro

HERRO is a highly-accurate, haplotype-aware, deep-learning tool for error correction of Nanopore R10.4.1 or R9.4.1 reads (read length of >= 10 kbps is recommended).

Other

136 stars 9 forks source link

How to reduce RAM usage of "herro inference"? #40

Open shelkmike opened 3 weeks ago

shelkmike commented 3 weeks ago

Can I reduce RAM (not VRAM) usage of the error-correction stage of HERRO by tweaking some of its parameters?

ItokawaK commented 2 weeks ago

I am also wandering this. Inputing 110 Gb fastq and their alignments eventually exceeded RAM usage 256 GB and stopped after ~80% of the reads were done..

pgonzale60 commented 1 week ago

Out of curiosity, are you using the scripts/create_batched_alignments.sh? I would expect that to tackle this issue.