README.md

---
base_model: meta-llama/Llama-3.1-8B-Instruct
library_name: peft
model_name: margin_reg_baseline_code_20260328_121558
tags:
- base_model:adapter:meta-llama/Llama-3.1-8B-Instruct
- lora
- reward-trainer
- transformers
- trl
licence: license
---

# Model Card for margin_reg_baseline_code_20260328_121558

This model is a fine-tuned version of [meta-llama/Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct).
It has been trained using [TRL](https://github.com/huggingface/trl).

## Quick start

```python
import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification
from peft import PeftModel, PeftConfig

# -----------------------------
# 1. Define PEFT model ID & Checkpoint (Epoch)
# -----------------------------
peft_model_id = "xxccho/margin_reg_baseline_code"

# [ Checkpoints to Epochs Mapping ]
# Epoch 1  : "checkpoint-246"
# Epoch 2  : "checkpoint-492"
# Epoch 3  : "checkpoint-738"
# Epoch 4  : "checkpoint-984"
# Epoch 5  : "checkpoint-1230"
# Epoch 6  : "checkpoint-1476"
# Epoch 7  : "checkpoint-1722"
# Epoch 8  : "checkpoint-1968"
# Epoch 9  : "checkpoint-2214"
# Epoch 10 : "checkpoint-2460"

# 예시: 5 Epoch 체크포인트를 사용하려면 "checkpoint-1230" 할당. None일 시 최종(10x) 모델 로드
checkpoint = None  

# 2. Load the PEFT config
config = PeftConfig.from_pretrained(peft_model_id, subfolder=checkpoint) if checkpoint else PeftConfig.from_pretrained(peft_model_id)

# 3. Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
if tokenizer.pad_token is None:
    tokenizer.pad_token = tokenizer.eos_token

# 4. Load base model
base_model = AutoModelForSequenceClassification.from_pretrained(
    config.base_model_name_or_path,
    num_labels=1,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

# 5. Apply LoRA adapter
model = PeftModel.from_pretrained(base_model, peft_model_id, subfolder=checkpoint) if checkpoint else PeftModel.from_pretrained(base_model, peft_model_id)
model.config.pad_token_id = tokenizer.pad_token_id
model.eval()

# Example Usage
text = "User: Write a python code for calculating fibonacci sequence.\nAssistant: Here is the code..."
inputs = tokenizer(text, return_tensors="pt").to(model.device)

with torch.no_grad():
    outputs = model(**inputs)
    reward_score = outputs.logits.squeeze().item()

print(f"[Code] Reward Score: {reward_score:.4f}")

```

## Training procedure

[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/changhee1016-seoul-national-university/reward-model-ood-robustness/runs/ogc45zau) 


This model was trained with Reward.

### Framework versions

- PEFT 0.18.0
- TRL: 0.26.1
- Transformers: 4.57.3
- Pytorch: 2.9.0
- Datasets: 4.4.1
- Tokenizers: 0.22.1

## Citations



Cite TRL as:
    
```bibtex
@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 Gallou{\'e}dec},
	year         = 2020,
	journal      = {GitHub repository},
	publisher    = {GitHub},
	howpublished = {\url{https://github.com/huggingface/trl}}
}
```