search

KoslickiLab / YACHT

A mathematically characterized hypothesis test for organism presence/absence in a metagenome
MIT License
28 stars 7 forks source link

Yacht training takes ~90min for GTDB r214 representative #91

Closed ShaopengLiu1 closed 7 months ago

ShaopengLiu1 commented 8 months ago

In my initial run for readme, I wrote ~15min for yacht train on r207 data. This step now takes 1.5h.

Adam found it took him ~1.5h to do so, I didn't keep the original runlog file. But this time I got the same result as him (~1.5h). Will update the readme and strongly suggest downloading the pre-trained file from zenodo.

ShaopengLiu1 commented 7 months ago

The main reason for this ~10x time difference is MEM usage (I didn't find where it come from)

  1. the initial version used 50GB MEM with ~15min and the current version only used 5GB.
  2. while we don't have parameter for MEM, this seems come from the default performance
ShaopengLiu1 commented 7 months ago

Readme updated