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
18.37k stars 1.28k forks source link

Feat/kto #1316

Open Erland366 opened 16 hours ago

Erland366 commented 16 hours ago

PR to further support KTO Unsloth Example Notebook

Main difference is just the column metrics score. For KTO, they don't have metrics/accuracies log on the output