Downloads · 30 days
23
30% of all-time downloads
detllm-qwen3-0.6b-int8
nathanbarry/detllm-qwen3-0.6b-int8
detllm-qwen3-0.6b-int8 is a machine learning model from nathanbarry. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Prepared artifact for nathanrs/detllm: an inference pipeline where the entire forward pass is exact integer arithmetic, producing bit-identical logits across runs, batch compositions, prefill/decode splits, and hardwa…
All-time downloads
76
Public
Parameters
763M
781 MB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors781 MB · 100%
Parameter types
How the weights are stored.
I8752M · 98%
Base models
Model card
From the Hugging Face model README
detllm — deterministic integer-only Qwen3-0.6B (int8 artifact)
Prepared artifact for nathanrs/detllm: an inference pipeline where the entire forward pass is exact integer arithmetic, producing bit-identical logits across runs, batch compositions, prefill/decode splits, and hardware backends (NVIDIA tensor cores vs. pure-CPU reference).
Derived from Qwen/Qwen3-0.6B (Apache-2.0) by one-time offline preparation (per-channel int8 weight quantization with folded norms and analytic smoothing, per-row-quantized embedding, int16 RoPE tables, dyadic scale constants). WikiText2 perplexity 20.72 vs the fp16 baseline's 20.95.
Integrity: runs on different machines are only comparable when
sha256(model.safetensors) matches — the repo's demo prints it on startup.
Usage:
git clone https://github.com/nathanrs/detllm && cd detllm && uv sync
hf download nathanbarry/detllm-qwen3-0.6b-int8 --local-dir artifacts/qwen3-0.6b-int8
uv run python scripts/demo.py # the determinism demo
- License
- apache-2.0
- Created
- Aug 9, 2026
- Updated
- Aug 9, 2026