reciprocate/openllama-13b_rm_oasst-hh

OpenLLama-13B for reward modeling

• Dataset: https://huggingface.co/datasets/pvduy/rm_oa_hh
• Logs: https://wandb.ai/sorry/autocrit/runs/j05t4e97?workspace=user-sorry
• Code: https://github.com/CarperAI/autocrit/blob/main/train_reward_model.py

Usage:

from transformers import AutoModelForSequenceClassification, AutoTokenizer

ckpt = "reciprocate/openllama-13b_rm_oasst-hh"
model = AutoModelForSequenceClassification.from_pretrained(ckpt, load_in_4bit=True)
tokenizer = AutoTokenizer.from_pretrained(ckpt)

model(**tokenizer("ASSISTANT: This sentence is a lie.", return_tensors="pt"))[0].item()

Output:

-1.626953125