Unsloth
AI
model finetuning library
SHOW HN2024
45points
6comments
#1,358of 3,515 in 2024
Top 39%of 2024 by points
How HN took it
QuietFewer than 5 comments about the project
0 positive, 2 neutral or asking, 0 mixed, 0 critical, out of 2 comments about the project.
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
No accuracy loss from the speedups, since no approximations are used.
From 1 comment
No catch at all!! There's 0 approximations, so everything is exact! We just have a custom backprop engine, rewrite everything into OpenAI's Triton language, and do all the…
danielhanchen on HN
Requests
Support for full parallel finetuning, not just 16bit and 4bit.
From 1 comment
Any news on when Unsloth's parallel full tuning will be available?
popinman322 on HN
Questions
Is there a training speed or VRAM penalty for the speedups?
From 1 comment
Hard to read on mobile, but if you don’t mind me asking, what’s the catch? Is there a penalty to training faster and less VRAM?
syntaxing on HN
Alternatives named
Hugging Face with Flash Attention 2.
From 2 comments
Hey! If you're interested to try out finetuning Llama-3 8B (Meta's new 15 trillion token!! model), made a Colab to finetune Llama-3 2x faster and use 60% less VRAM, and supports…
danielhanchen on HNNo catch at all!! There's 0 approximations, so everything is exact! We just have a custom backprop engine, rewrite everything into OpenAI's Triton language, and do all the…
danielhanchen on HN