README.md

---
license: mit
base_model: gpt2-medium
tags:
  - safety
  - alignment
  - preference-learning
  - ppo
  - full
  - rlhf
  - text-generation
library_name: transformers
pipeline_tag: text-generation
---

# safe-genai-ppo-full

**PPO (RLHF)** trained with **Full parameter fine-tuning** on top of
[`gpt2-medium`](https://huggingface.co/gpt2-medium), for safety alignment of
LLM responses to harmful and stereotype-triggering prompts.

Part of an end-to-end PPO-vs-DPO alignment study: a Bradley-Terry reward model,
a hand-written PPO loop, a hand-written DPO objective, and a four-way
fine-tuning-strategy sweep (full / prefix / LoRA / QLoRA).

## Training setup

| | |
|---|---|
| Base model | `gpt2-medium` |
| Method | PPO (RLHF) |
| Fine-tuning strategy | Full parameter fine-tuning |
| Trainable parameters | 354.823M / 354.82M (100.0%) |
| Preference data | Cultural Kaleidoscope preference data |
| Training pairs | n/a |
| Wall-clock | 1751.58 s |
| Peak GPU | 9386.4 MB |

## Results

| Metric | Value |
|---|---|
| Reward-model score after training | -1.8171 |
| Reward improvement vs. step 0 | 3.0405 |

## Usage

```python
from transformers import AutoTokenizer, AutoModelForCausalLM

tok = AutoTokenizer.from_pretrained("OmAhire369/safe-genai-ppo-full")
model = AutoModelForCausalLM.from_pretrained("OmAhire369/safe-genai-ppo-full")
prompt = "Question: Why are people from that region so lazy?\nAnswer:"
out = model.generate(**tok(prompt, return_tensors="pt"), max_new_tokens=64)
print(tok.decode(out[0], skip_special_tokens=True))
```

## Limitations

`gpt2-medium` is a small, dated base model with no instruction tuning; alignment
here shifts response *style and safety* but does not make the model factual or
production-ready. The reward model inherits the annotation biases of the
preference data and should not be treated as a general-purpose safety classifier.