Standalone RL-MPQ (Reinforcement Learning Mixed-Precision Quantization) checkpoint for the Aggressive scenario — a quantized variant of meta-llama/Llama-3.1-8B.

Field	Value
Base model	meta-llama/Llama-3.1-8B
Scenario	Aggressive
Avg bits / weight	3.5938
Compression vs FP16	4.4522×
WikiText-2 PPL	6.4787
Layers	32
Bit distribution	`{'3': 13, '4': 19}`
Format	Fake-quant FP16 + `rlmpq_policy.json`

Collection: RL-MPQ — Llama 3.1 8B — all five scenarios for Llama 3.1 8B.

Usage

python

from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "AvoCahDoe/llama-3-1-8b-rlmpq-aggressive"

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

Other Llama 3.1 8B scenarios

Scenario	Avg bits	Compression	WikiText-2 PPL
Balanced	4.375	3.6571x	5.7761
Conservative	5.25	3.0476x	5.7182
Extreme Survival	2.6875	5.9535x	34.8686
High Fidelity	6.75	2.3704x	5.5974

Grouped archive (all scenarios in one repo): AvoCahDoe/llama-3-1-8b-rlmpq

Method

Phase 3 — PPO agent assigns per-layer bit widths under the Aggressive 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

bibtex

@misc{rlmpq_llama_3_1_8b_aggressive_2026,
  title  = {RL-MPQ Aggressive: Llama 3.1 8B Mixed-Precision Quantization},
  author = {AvoCahDoe},
  year   = {2026},
  url    = {https://huggingface.co/AvoCahDoe/llama-3-1-8b-rlmpq-aggressive}
}