jetmoe-8b-rlmpq-conservative

JetMoE 8B — RL-MPQ Conservative

Standalone RL-MPQ (Reinforcement Learning Mixed-Precision Quantization) checkpoint for the Conservative scenario — a quantized variant of jetmoe/jetmoe-8b.

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

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "AvoCahDoe/jetmoe-8b-rlmpq-conservative"

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

Other JetMoE 8B scenarios

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

Method

1. Phase 3 — PPO agent assigns per-layer bit widths under the Conservative reward target.
2. Phase 4 — Policy replayed on real weights; WikiText-2 perplexity validates quality.
3. Export — Fake-quantized FP16 weights compatible with Hugging Face Transformers.

Files

Citation

@misc{rlmpq_jetmoe_8b_conservative_2026,
  title  = {RL-MPQ Conservative: JetMoE 8B Mixed-Precision Quantization},
  author = {AvoCahDoe},
  year   = {2026},
  url    = {https://huggingface.co/AvoCahDoe/jetmoe-8b-rlmpq-conservative}
}