unslothai / unsloth

Finetune Llama 3.2, Mistral, Phi, Qwen 2.5 & Gemma LLMs 2-5x faster with 80% less memory
https://unsloth.ai
Apache License 2.0
18.68k stars 1.31k forks source link

Can I use Unsloth / TRL to PEFT embeddings and rerankers? #343

Open l4b4r4b4b4 opened 7 months ago

l4b4r4b4b4 commented 7 months ago

I have done that with peft in the past with good and not so good results ;)

shimmyshimmer commented 7 months ago

Unsure sadly we've never tried it, but I'm assuming it works

shimmyshimmer commented 1 month ago

Will be supported in the future