Qwen 2.5 7B — RL-MPQ Conservative
Standalone RL-MPQ (Reinforcement Learning Mixed-Precision Quantization) checkpoint for the
Conservative scenario — a quantized variant of
Qwen/Qwen2.5-7B.
| Field | Value |
|---|
| Base model | Qwen/Qwen2.5-7B |
| Scenario | Conservative |
| Avg bits / weight | 3.6786 |
| Compression vs FP16 | 4.3495× |
| WikiText-2 PPL | 8.4114 |
| Layers | 28 |
| Bit distribution | {'3': 9, '4': 19} |
| Format | Fake-quant FP16 + rlmpq_policy.json |
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "AvoCahDoe/qwen2-5-7b-rlmpq-conservative"
model = AutoModelForCausalLM.from_pretrained(repo, torch_dtype="float16")
tokenizer = AutoTokenizer.from_pretrained(repo)
Other Qwen 2.5 7B scenarios
Grouped archive (all scenarios in one repo):
AvoCahDoe/qwen2-5-7b-rlmpq
Method
- Phase 3 — PPO agent assigns per-layer bit widths under the Conservative reward target.
- Phase 4 — Policy replayed on real weights; WikiText-2 perplexity validates quality.
- Export — Fake-quantized FP16 weights compatible with Hugging Face Transformers.
Files
| File | Description |
|---|
config.json | Llama architecture + RL-MPQ metadata |
model.safetensors | Fake-quantized weights |
rlmpq_policy.json | Per-layer bit-width policy |
rlmpq_metrics.json | Validation & PPL summary |
Citation
@misc{rlmpq_qwen2_5_7b_conservative_2026,
title = {RL-MPQ Conservative: Qwen 2.5 7B Mixed-Precision Quantization},
author = {AvoCahDoe},
year = {2026},
url = {https://huggingface.co/AvoCahDoe/qwen2-5-7b-rlmpq-conservative}
}