Model Card for llm-course-hw2-reward-model

This model is a fine-tuned version of HuggingFaceTB/SmolLM-135M-Instruct on the HumanLLMs/Human-Like-DPO-Dataset dataset. It has been trained using TRL.

This model is a reward model used for training efromomr/llm-course-hw2-ppo.

train_loss: 0.07687531913222831

##Usage example

DEVICE = torch.device('cuda')

tokenizer = AutoTokenizer.from_pretrained(llm-course-hw2-reward-model)
reward_model = AutoModelForSequenceClassification.from_pretrained(llm-course-hw2-reward-model, num_labels = 1)
reward_model.config.pad_token_id = tokenizer.pad_token_id
reward_model = reward_model.to(DEVICE)
reward_model.eval()

inputs_chosen = tokenizer.apply_chat_template('Any text'], tokenize=False)
inputs_chosen = tokenizer(inputs_chosen, return_tensors="pt").to(DEVICE)


score_chosen = reward_model(**inputs_chosen).logits[0].cpu().detach()
print(score_chosen)

#0.223

Framework versions

• TRL: 0.15.2
• Transformers: 4.48.3
• Pytorch: 2.5.1+cu124
• Datasets: 3.3.2
• Tokenizers: 0.21.0

Citations

Cite TRL as:

@misc{vonwerra2022trl,
    title        = {{TRL: Transformer Reinforcement Learning}},
    author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}