accelerate

Accelerate

accelerate Permissive license Apache Active ★ 9,817 v1.14.0 released 2026-06-11

Install

accelerate on PyPI

pip

pip install accelerate

uv

uv add accelerate

poetry

poetry add accelerate

Package facts

License	Apache (permissive)
Python support	supports the current Python release (>=3.10.0)
Install friction	low — pure-Python wheel
Runtime dependencies	7 — numpy, packaging, psutil, pyyaml, torch, huggingface_hub, safetensors
Maintenance	actively maintained — 63 days since the last release
Last repo commit	2026-08-10
First released	2020-10-30
Popularity	one of the top 1,000 most-downloaded packages on PyPI (30-day window, as of 2026-08-13)
Known vulnerabilities	none known (OSV.dev, checked 2026-08-13)

Evidence: accelerate-1.14.0-py3-none-any.whl

Keywords: deep, learning

Development Status :: 5 - Production/StableIntended Audience :: DevelopersIntended Audience :: EducationIntended Audience :: Science/ResearchLicense :: OSI Approved :: Apache Software LicenseOperating System :: OS IndependentProgramming Language :: Python :: 3Programming Language :: Python :: 3.10Topic :: Scientific/Engineering :: Artificial Intelligence

About accelerate

from the package's own PyPI description — quoted content, verbatim

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. -->

<a href="https://github.com/huggingface/accelerate/blob/main/LICENSE"><img alt="License" src="https://img.shields.io/github/license/huggingface/accelerate.svg?color=blue"></a> <a...

Read as markdown · JSON record · Source repository · Homepage

AI interpretation — verify before relying

AI-generated interpretation of the package facts above; every digit, version, license, or vulnerability id it cites is grounded in the facts already shown on this page

Accelerate abstracts away distributed training boilerplate for PyTorch, letting you run the same training script unchanged across single/multi-GPU, TPU, and mixed-precision configurations.

Low friction: pure Python wheel with well-maintained dependencies. Active maintenance with 63 days since last release signals strong community backing.

Apache permissive license allows commercial and private use with minimal restrictions; you must include a copy of the license and state significant changes, but no copyleft obligations apply.

Usage

pip install accelerate

from accelerate import Accelerator

accelerator = Accelerator()
model, optimizer, data = accelerator.prepare(model, optimizer, data)
accelerator.backward(loss)

Requires Python ≥3.10.0 and torch; GPU/TPU hardware optional but needed to benefit from distributed acceleration.

Verdict: Production-ready library for painless PyTorch multi-device training. No known vulnerabilities, permissive license, low install friction, and active maintenance make it a solid choice for anyone writing custom PyTorch training loops.

Needs verification

Whether all seven runtime dependencies (numpy, packaging, psutil, pyyaml, torch, huggingface_hub, safetensors) are truly required at runtime or some are optional.
Performance overhead or compatibility gotchas when using accelerate with custom CUDA kernels or non-standard model architectures.
Specific use cases where the CLI tool (accelerate config/launch) provides measurable advantage over direct torch.distributed.run.

pytorch distributed trainingmulti-gpu training wrappermixed precision pytorchpytorch training accelerationtpu pytorch trainingpytorch device abstractiongradient accumulation pytorch