undefined | Better HN

0 pointsvaline3y ago0 comments

Fine tuning is moderately intense. Going from base LLaMA to Vicuna 13B costs about $300.

0 comments

1 comments · 1 top-level

The compute requirements are reasonable, but the memory requirements are extremely high, even with some upcoming breakthroughs like 4 bit bitsandbytes.

j / k navigate · click thread line to collapse