Gemma 2 9B — RL-MPQ Conservative
Standalone RL-MPQ (Reinforcement Learning Mixed-Precision Quantization) checkpoint for the Conservative scenario — a quantized variant of google/gemma-2-9b.
| Field | Value |
|---|---|
| Base model | google/gemma-2-9b |
| Scenario | Conservative |
| Avg bits / weight | 5.1429 |
| Compression vs FP16 | 3.1111× |
| WikiText-2 PPL | 116.5244 |
| Layers | 42 |
| Bit distribution | {'4': 30, '8': 12} |
| Format | Fake-quant FP16 + rlmpq_policy.json |
Collection: RL-MPQ — Gemma 2 9B — all five scenarios for Gemma 2 9B.
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "AvoCahDoe/gemma-2-9b-rlmpq-conservative"
model = AutoModelForCausalLM.from_pretrained(repo, torch_dtype="float16")
tokenizer = AutoTokenizer.from_pretrained(repo)
Other Gemma 2 9B scenarios
| Scenario | Avg bits | Compression | WikiText-2 PPL |
|---|---|---|---|
| Aggressive | 3.6667 | 4.3636x | 162.8437 |
| Balanced | 4.2857 | 3.7333x | 127.0798 |
| Extreme Survival | 2.7857 | 5.7436x | 424.7991 |
| High Fidelity | 7.0476 | 2.2703x | 104.8098 |
Grouped archive (all scenarios in one repo): AvoCahDoe/gemma-2-9b-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_gemma_2_9b_conservative_2026,
title = {RL-MPQ Conservative: Gemma 2 9B Mixed-Precision Quantization},
author = {AvoCahDoe},
year = {2026},
url = {https://huggingface.co/AvoCahDoe/gemma-2-9b-rlmpq-conservative}
}