Skip to content
OpenTrain AIFor AI Companies

Optimize Weight Rounding via Signed Gradient Descent for the Quantization of LLMs

Published Sep 1, 2023
arXiv PDF
Researcher verdict
Starting point
Use as implementation starting point
Benchmark evidence
Missing
Not verified yet
Time to first repro
A few hours
Fast first run
Risk flags
0
None detected

Results and benchmarks

Freshness tier: cold
Optimize Weight Rounding via Signed Gradient Descent for the Quantization of LLMs presents a quantization method.

Implementation

Best maintained implementation now

Recommended
Confidence: High
Reproducibility: Strong
intel/neural-compressor

SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime

2,704 stars · 320 forks · Last push Aug 26, 2026 · Apache-2.0 license

  • License
  • CI
  • Dependencies
  • Docker

Official implementation from Papers with Code · Repository link is mentioned in the paper metadata · Community adoption signal (2704 stars)

Why this implementation
Confidence: high

intel/neural-compressor is the strongest maintained implementation based on ranking signals. CI workflows are present. License is declared (Apache-2.0).

Open intel/neural-compressor
Reproduction risks
  • No repository-level red flags were detected, but paper-specific preprocessing and hyperparameter details may still be under-specified.
  • Selected intel/neural-compressor as the strongest maintained implementation for new work.
  • Includes CI workflow signals.
  • Includes dependency/environment manifest signals.
  • Repository activity is within the last 24 months.

Compare implementation paths

Compare maintenance quality, reproducibility coverage, and evidence confidence before choosing a reproduction baseline.

intel/neural-compressor
best maintained
Maintenance
Active
Confidence
High
Reproducibility
Strong
Stars
2,704
Last push
Aug 26, 2026 (2d)

Official implementation from Papers with Code · Repository link is mentioned in the paper metadata

  • No Docker setup
intel/auto-round
historical official
Maintenance
Active
Confidence
High
Reproducibility
Strong
Stars
1,588
Last push
Aug 27, 2026 (1d)

Official implementation from Papers with Code · Repository link is mentioned in the paper metadata

  • No Docker setup
vllm-project/vllm
alternative
Maintenance
Active
Confidence
Low
Reproducibility
Strong
Stars
90,156
Last push
Aug 27, 2026 (1d)

Community adoption signal (90156 stars)

  • No Docker setup
  • Low confidence match

Reproduction readiness

Time to first repro: hours
Last checked: Aug 27, 2026

Ready to run

Ready to reproduce

  • Clone intel/neural-compressor and install dependencies from pyproject.toml.
  • CI pipeline detected, so automated tests are in place.
  • Last updated 2 days ago.
Open intel/neural-compressor

Quick start

git clone https://github.com/intel/neural-compressor.git
pip install -e .

Framework baselines

Repositories and ecosystem

No additional verified repositories beyond the primary recommendation.

Hugging Face artifacts

No trustworthy direct or curated related Hugging Face artifacts were found yet. Use targeted searches to quickly locate candidate models, datasets, and demos.

Search models Search datasets Search spaces

Tip: start with models, then check datasets and spaces if you need evaluation data or demos.

Research context

Tasks

None detected

Methods

Quantization, Stochastic optimization

Domains

None detected

Evaluation and human feedback data

Open this paper in HFEPX to review benchmark signals, evaluation modes, and human-feedback protocol context.

Open in HFEPX
Explore similar papers

Jump to Paper2Code search queries derived from this paper's research context.

Quantization Stochastic optimization

Data includes links from Papers with Code ( CC-BY-SA-4.0 ).