Updated legacy checkpoint wrapper to used pickle+gzip format
Merge request reports
Activity
Looks pretty slow still, on the FRGC test I didn't manage yet to finish any
write_scorestask. Currentlywrite_scorestook 60% of the execution time (33% for thecompute_scores), and growing.Of the
write_scorestask, we spend most of it in theuncompress_unpicklefunction, which I guess makes sense. But still it's weird that it's so slow as I don't remember it ever being an issue before. I am wondering if it's linked to one of my library versions, my env does not havebob-develin it so I might have some unpinned stuff that is playing tricks.mentioned in commit e6b0ccd0
mentioned in issue bob#269 (closed)
Please register or sign in to reply