#99th of 15 in ML frameworks and libraries
diffusers
PyTorch library of diffusion pipelines and schedulers for image generation
- Stars
- 34.7k
- License
- Apache-2.0
- Last commit
- Oct 2026
- Last release
- Oct 2026
- Language
- Python
32,846 on 24 Mar 202634,717 on 13 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Diffusers is a Python library from Hugging Face for running and training diffusion models. It provides pipelines loaded with from_pretrained, interchangeable noise schedulers, and pretrained model classes that can be combined into custom systems. It covers text-to-image, image-to-image, inpainting, super resolution and unconditional generation, plus audio and 3D molecule structures.
Who it is for: Python developers building or fine-tuning diffusion-based generation
Strengths
- Loads any of 30,000+ Hub checkpoints with a few lines via from_pretrained
- Schedulers, models and pipelines are separate, so custom sampling loops are possible
- Docs include training and memory/speed optimization guides
- Used by InvokeAI, kohya_ss and BentoML, among 14,000+ repositories
Weaknesses
- Library only: no web UI or HTTP server included
- README documents PyTorch installs only
- README gives no VRAM or RAM figures per model
- Quickstart assumes CUDA; Apple Silicon is in a separate guide
What it needs
- GPU optional
- Docker
- Needs PyTorch
- Models: Stable Diffusion, DDPM, ControlNet, InstructPix2Pix, DeepFloyd IF
Also in ML frameworks and libraries
See all 15| Rank | Project | Score |
|---|---|---|
| 1 | transformersModel-definition library for text, vision, audio and multimodal models | 94 out of 100 |
| 2 | pytorchPython tensor library with GPU support and tape-based autograd for deep learning | 92 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 86 out of 100 |
| #4 | pytorch-image-modelsPyTorch library of image models, pretrained weights and reference training scripts | 82 out of 100 |
| #5 | tensorflowTensor and neural network framework with Python and C++ APIs | 80 out of 100 |