About how to train reward model #61
Description
I am fine-tuning a reward model (single-label regression) using the helpsteer2 data. Should I use PEFT to fine-tune the attention layer or attention+ffn? Looking forward to your reply.
{{ message }}
I am fine-tuning a reward model (single-label regression) using the helpsteer2 data. Should I use PEFT to fine-tune the attention layer or attention+ffn? Looking forward to your reply.