search

huggingface / alignment-handbook

Robust recipes to align language models with human and AI preferences
https://huggingface.co/HuggingFaceH4
Apache License 2.0
4.2k stars 357 forks source link

Make SFT script consistent with DPO script #86

Closed NielsRogge closed 6 months ago

NielsRogge commented 6 months ago

The SFT script currently doesn't pass the num_proc argument as in the DPO script, which I've added.

I've also added remove_columns and desc similar to the DPO script.

nathan-az commented 6 months ago

make style quality should correct the checks :)