---
license: llama2
base_model: meta-llama/Llama-2-7b-hf
pipeline_tag: text-generation
language:
  - en
tags:
  - llama
  - text-generation
  - rl-mpq
  - mixed-precision
  - quantization
  - fake-quantization
  - llama-2
library_name: transformers
datasets:
  - wikitext
---

# Llama 2 7B — RL-MPQ Quantized (Thesis Release)

**Quantized variant of [meta-llama/Llama-2-7b-hf](https://huggingface.co/meta-llama/Llama-2-7b-hf)** using **RL-MPQ**
(Reinforcement Learning Mixed-Precision Quantization): per-layer bit-width policies
trained with PPO, validated on WikiText-2 perplexity.

This repo ships **five compression scenarios** as subfolders — from near-FP16 fidelity
to aggressive survival mode — so you can pick the bits-vs-quality trade-off for your thesis experiments.

| | |
|---|---|
| **Base model** | [meta-llama/Llama-2-7b-hf](https://huggingface.co/meta-llama/Llama-2-7b-hf) |
| **Method** | RL-MPQ (PPO per-layer bit policy) |
| **Format** | Fake-quant FP16 weights + `rlmpq_policy.json` |
| **Recommended start** | `subfolder="Balanced"` |

## Scenarios

- [`High_Fidelity/`](./High_Fidelity/) — load with `subfolder="High_Fidelity"`
- [`Conservative/`](./Conservative/) — load with `subfolder="Conservative"`
- [`Balanced/`](./Balanced/) — load with `subfolder="Balanced"`
- [`Aggressive/`](./Aggressive/) — load with `subfolder="Aggressive"`
- [`Extreme_Survival/`](./Extreme_Survival/) — load with `subfolder="Extreme_Survival"`

## Results (WikiText-2)

| Scenario | Avg bits | Compression vs FP16 | Perplexity |
|----------|----------|---------------------|------------|
| [High_Fidelity](./High_Fidelity/) | 6.5 | 2.4615x | 4.9808 |
| [Conservative](./Conservative/) | 5.125 | 3.122x | 5.0276 |
| [Balanced](./Balanced/) | 4.375 | 3.6571x | 5.0437 |
| [Aggressive](./Aggressive/) | 3.5938 | 4.4522x | 5.2614 |
| [Extreme_Survival](./Extreme_Survival/) | 2.9688 | 5.3895x | 10.9577 |

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "AvoCahDoe/llama-2-7b-rlmpq"
scenario = "Balanced"  # High_Fidelity | Conservative | Aggressive | Extreme_Survival

model = AutoModelForCausalLM.from_pretrained(repo, subfolder=scenario, torch_dtype="float16")
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=scenario)
```

> **Important:** Always pass `subfolder=<scenario>`. Root `config.json` describes the
> collection; weights and tokenizer live inside each scenario folder.

## Method (thesis summary)

1. **Phase 3** — PPO agent selects per-layer bit widths under scenario-specific reward targets.
2. **Phase 4** — Policies replayed on real weights; WikiText-2 PPL measures quality retention.
3. **Export** — Fake-quantized FP16 checkpoints (compatible with Hugging Face Transformers).

## Citation

```bibtex
@misc{rlmpq2026,
  title  = {RL-MPQ: Reinforcement Learning Mixed-Precision Quantization},
  author = {AvoCahDoe},
  year   = {2026},
  url    = {https://huggingface.co/AvoCahDoe/llama-2-7b-rlmpq}
}
```

Part of the **RL-NMP-Model-Quantasation** thesis framework.