#5858th of 75 in ML frameworks and libraries
accelerate
Thin PyTorch wrapper that runs your training loop on CPU, GPU, TPU or multi-node
- Stars
- 9.9k
- License
- Apache-2.0
- Last commit
- Oct 2026
- Last release
- Sep 2026
- Language
- Python
9,514 on 31 Mar 20269,924 on 20 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Accelerate adds an Accelerator class to a plain PyTorch training script, handling device placement, distributed setup and mixed precision (fp16, bf16, fp8) with about five changed lines. It includes an optional `accelerate config` and `accelerate launch` CLI, a notebook launcher, and plugins for DeepSpeed, FSDP and Megatron-LM, which the README marks experimental. It is not a high-level trainer: you still write the loop yourself.
Who it is for: PyTorch users who write their own training loops and need multi-device scaling
Strengths
- Whole API is one Accelerator class; the rest of the training loop stays unchanged
- Same script runs on CPU, single or multi-GPU, multi-node and TPU
- Optional CLI wraps torchrun, MPI and DeepSpeed launch config; plain python also works
- Mixed precision includes fp8 via Transformer Engine or MS-AMP
Weaknesses
- No built-in trainer; you must write and maintain your own training loop
- DeepSpeed, FSDP and Megatron-LM support are labeled experimental in the README
- PyTorch only; needs Python 3.8+ and PyTorch 1.10.0+ installed separately
- With DeepSpeed you still handle gradient accumulation yourself
What it needs
- GPU optional
- Docker
- Needs PyTorch, DeepSpeed (optional), Open MPI (optional), Transformer Engine or MS-AMP (optional, fp8)
- Models: PyTorch models, Hugging Face Transformers
Also in ML frameworks and libraries
See all 75| Rank | Project | Score |
|---|---|---|
| 1 | transformersPython library defining open models for text, vision, audio and multimodal tasks | 94 out of 100 |
| 2 | pytorchPython tensor library with GPU support and tape-based autograd for deep learning | 94 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 92 out of 100 |
| #4 | pytorch-image-modelsPyTorch library of image models, pretrained weights and reference training scripts | 91 out of 100 |
| #5 | supervisionPython toolkit for loading, annotating and counting computer vision detections | 90 out of 100 |