auto-round
Repository of AutoRound: Advanced Weight-Only Quantization Algorithm for LLMs
Decision gist · record as of 2026-08-14
Yes. AutoRound is actively maintained, has no known vulnerabilities, installs with low friction, and is permissively licensed. It is well-suited for anyone needing to compress LLMs or VLMs for inference—whether for research, edge deployment, or cost reduction. Start with the CLI recipes (auto-round-best, auto-round-rtn) if you want quick results, or use the Python API for fine-grained control.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires Python 3.10+.
- Quantization is GPU-accelerated; CPU-only quantization is possible but slower.
- Needs torch and transformers installed.
License · maintenance · safety
Apache 2.0 (permissive) — Apache 2.0 permissive license allows commercial and private use with minimal restrictions—suitable for most production and research contexts.
last release 2026-07-13 (32 days) · last repo commit 2026-08-14 · 1,566 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 300,133 downloads/mo, #7,847 on PyPI
Alternatives
Verify before relying
pip install auto-round
from auto_round import AutoRound
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "meta-llama/Llama-2-7b"
model = AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)
auto_round = AutoRound(model, tokenizer, dataset="wikitext2")
quantized_model = auto_round.quantize()- Exact memory overhead for mixed-precision scheme generation (stated as 1.1X–1.5X BF16 RAM but not verified independently)
- Actual quantization time for 7B models on single GPU (stated as ~10 minutes but hardware-dependent)
- Support matrix for all 10+ VLMs mentioned (list not provided in fact sheet)
What it is and what it does
AutoRound is a quantization toolkit that compresses large language models and vision-language models to ultra-low bit widths (2–4 bits) while preserving accuracy. It uses sign-gradient descent to find optimal quantization parameters with minimal tuning overhead. The package integrates with popular inference frameworks (vLLM, SGLang, Transformers) and supports multiple export formats (AutoRound, AutoAWQ, AutoGPTQ, GGUF), making quantized models portable across different deployment environments.
The toolkit offers several quantization recipes—from fast round-to-nearest (RTN) baseline to more accurate iterative methods—and includes utilities for multi-GPU quantization, mixed-precision schemes, and multiple calibration datasets. It targets both researchers optimizing model accuracy at low bits and practitioners seeking to reduce model size and inference latency for deployment.
Use it for
- Compress a 7B LLM to 2–3 bits for edge deployment or cost-effective cloud inference
- Export a quantized model to GGUF format for use with llama.cpp or other C++ inference engines
- Quantize a vision-language model for efficient multimodal inference on resource-constrained hardware
- Generate a mixed-precision quantization scheme automatically to balance accuracy and model size
- Integrate quantized models into vLLM or SGLang for fast batch inference with reduced memory footprint
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes.
AutoRound is actively maintained, has no known vulnerabilities, installs with low friction, and is permissively licensed. It is well-suited for anyone needing to compress LLMs or VLMs for inference—whether for research, edge deployment, or cost reduction. Start with the CLI recipes (auto-round-best, auto-round-rtn) if you want quick results, or use the Python API for fine-grained control.
Install
auto-round on PyPI
Before you install
Low friction install with a pure-Python wheel. The package is actively maintained (last commit 2026-08-14, 32 days since release) and depends on standard ML libraries (torch, transformers, datasets, numpy, accelerate, tqdm, py-cpuinfo, pydantic). Requires Python 3.10 or later.
Requires Python 3.10+. Quantization is GPU-accelerated; CPU-only quantization is possible but slower. Needs torch and transformers installed.
License in practice
Apache 2.0 permissive license allows commercial and private use with minimal restrictions—suitable for most production and research contexts.
Quickstart
pip install auto-round
from auto_round import AutoRound
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "meta-llama/Llama-2-7b"
model = AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)
auto_round = AutoRound(model, tokenizer, dataset="wikitext2")
quantized_model = auto_round.quantize()Verify before relying
- Exact memory overhead for mixed-precision scheme generation (stated as 1.1X–1.5X BF16 RAM but not verified independently)
- Actual quantization time for 7B models on single GPU (stated as ~10 minutes but hardware-dependent)
- Support matrix for all 10+ VLMs mentioned (list not provided in fact sheet)
Package facts
| License | Apache 2.0 permissive |
| Python support | Supports the current Python release >=3.10.0 |
| Install friction | Low. Pure-Python wheel |
| Runtime dependencies | 8 packagesacceleratedatasetsnumpypy-cpuinfotorchtqdmtransformerspydantic |
| Maintenance | Actively maintained 32 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 300,133 / month, #7,847 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | Intended Audience :: Science/ResearchLicense :: OSI Approved :: Apache Software LicenseProgramming Language :: Python :: 3Topic :: Scientific/Engineering :: Artificial Intelligence |
Evidence: auto_round-0.14.2-py3-none-any.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “LLM quantization low bit”
- auto-roundAutoRound quantizes large language models and vision-language models…
- diffqDiffQ performs differentiable quantization of PyTorch models using…
- flashinfer-pythonFlashInfer provides optimized GPU kernels for LLM inference…
Give your agent the search over MCP, or paste the wish link into any chat.
More Artificial Intelligence packages
LiteLLM provides a unified Python interface to call 100+ LLM providers (OpenAI, Anthropic, Gemini, Bedrock, Azure, and others) using OpenAI-compatible API format, available as both a Python SDK and a self-hosted AI Gateway proxy server.
Install it if you need to work with multiple LLM providers or want to centralize LLM routing in your organization.
Client library and CLI tool for downloading, uploading, and managing models, datasets, and repositories on the Hugging Face Hub platform.
Install it if you work with Hugging Face Hub models or datasets.
LangChain provides a framework for building agents and LLM-powered applications by composing language models, tools, and memory through a unified API that abstracts over multiple model providers.
hf-xet provides chunk-based deduplication and efficient file transfer for the Hugging Face Hub, enabling faster uploads and downloads of large files with local disk caching.
Tokenizers converts raw text into token sequences for NLP models, with support for training custom vocabularies and using pre-built tokenizers (BPE, WordPiece) optimized for speed via Rust.
Transformers provides a unified framework for loading, fine-tuning, and running state-of-the-art pretrained models across text, vision, audio, video, and multimodal tasks using PyTorch, JAX, or TensorFlow.
Install it if you need to run or train any transformer-based model for NLP, vision, audio, or multimodal tasks.
See also llmcompressor · auto-gptq · vllm · vllm-cpu · torchao · ipex-llm · ai-edge-quantizer · nvidia-modelopt · compressed-tensors · sgl-kernel