pythia-70m_utility_reward

TRL Model

This is a TRL language model that has been fine-tuned with reinforcement learning to guide the model outputs according to a value, function, or human feedback. The model can be used for text generation. This was used as a test model in the reward interpretability study at https://arxiv.org/abs/2310.08164.

Usage

To use this model for inference, first install the TRL library:

python -m pip install trl

You can then generate text as follows:

from transformers import pipeline

generator = pipeline("text-generation", model="amirabdullah19852020//tmp/tmplozr85hs/amirabdullah19852020/pythia-70m_utility_reward")
outputs = generator("Hello, my llama is cute")

If you want to use the model for training or to obtain the outputs from the value head, load the model as follows:

from transformers import AutoTokenizer
from trl import AutoModelForCausalLMWithValueHead

tokenizer = AutoTokenizer.from_pretrained("amirabdullah19852020//tmp/tmplozr85hs/amirabdullah19852020/pythia-70m_utility_reward")
model = AutoModelForCausalLMWithValueHead.from_pretrained("amirabdullah19852020//tmp/tmplozr85hs/amirabdullah19852020/pythia-70m_utility_reward")

inputs = tokenizer("Hello, my llama is cute", return_tensors="pt")
outputs = model(**inputs, labels=inputs["input_ids"])