---
license: llama2
base_model: meta-llama/Llama-2-7b-hf
tags:
  - rl-mpq
  - mixed-precision
  - quantization
library_name: transformers
---

# LLAMA-2-7B — Aggressive

Subfolder of [`AvoCahDoe/llama-2-7b-rlmpq`](https://huggingface.co/AvoCahDoe/llama-2-7b-rlmpq) (all RL-MPQ scenarios for this model).

Fake-quantized weights for **Aggressive** applied to
[meta-llama/Llama-2-7b-hf](https://huggingface.co/meta-llama/Llama-2-7b-hf) (per-layer asymmetric group-wise quant, group_size=128).

## Load

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "AvoCahDoe/llama-2-7b-rlmpq"
model = AutoModelForCausalLM.from_pretrained(repo, subfolder="Aggressive", torch_dtype="float16")
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder="Aggressive")
```

## Metrics

| Metric | Value |
|--------|-------|
| Scenario | Aggressive |
| Average bits per weight | 3.5938 |
| Compression ratio (vs FP16) | 4.4522x |
| WikiText-2 perplexity | 5.2614 |
| Layers | 32 |
| Bit distribution | {'3': 13, '4': 19} |

## Policy (bits per layer)

```
[4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4]
```

Generated: 2026-06-11T15:31:57.857429