huggingface / alignment-handbook

Robust recipes to align language models with human and AI preferences

Home Page:https://huggingface.co/HuggingFaceH4

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Is there anyway that I can use learning rate warm-up during the training ?

shamanez opened this issue · comments

I am using this repo to:

  1. Continual Pre-training
  2. SFT
  3. DPR

For stage 1, I want to use a learning rate warm-up.

Hello @shamanez the scripts in this repo are mostly focused on chat models, so for continual pretraining I recommend checking out the transformers script: https://github.com/huggingface/transformers/blob/main/examples/pytorch/language-modeling/run_clm.py

There you can specify e.g. warmup_ratio in TrainingArguments to get the warmup.