Dev Tools
ms-swift
Fine-tune and deploy 600+ LLMs and 300+ multimodal models with PEFT or full-parameter training.
★ 15.8k starsApache-2.0
View guide →Fine-tune huge models on small hardware with LoRA and other efficient methods
peft is an open-source dev tools project by huggingface (huggingface/peft on GitHub). It has 21,769 GitHub stars and 2,560 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 8 October 2026.
Pick the model to adapt and a method such as LoRA that fits your memory budget.
Install the peft package from PyPI - cloning is only needed to work on the library itself.
Wrap your model with a PEFT config (for example LoraConfig) using the README quickstart.
Train on your dataset, check the trainable-parameter count, and save the small adapter.
Implements LoRA and a family of parameter-efficient methods behind one consistent API.
The README's example trains about 0.12% of a 3B model's parameters with LoRA instead of the full model.
Use PEFT models directly through the Transformers integration for training and inference.
Apply parameter-efficient tuning to diffusion models as well as language models.
Works with Accelerate for distributed/offloaded training and with TRL for RLHF-style workflows.
Combines with quantization so large models can be adapted on modest GPUs.
Get the source code from GitHub and enter the project directory.
git clone https://github.com/huggingface/peft.git
cd peftInstall from PyPI alongside Transformers and Accelerate.
pip install peftThe README quickstart loads a model, applies a LoraConfig, and prints the trainable parameter share.
from peft import LoraConfig, get_peft_model
# load a model, then:
# model = get_peft_model(model, LoraConfig(...)) - see README quickstartIt is maintained by Hugging Face and contributors on GitHub.
Yes - PEFT is Apache-2.0 licensed and free to use. Base models you adapt carry their own licenses.
No API key is needed: training and inference run locally on your hardware with open models.
Compared with full fine-tuning, PEFT trains a fraction of the parameters (often well under 1%), cutting compute and storage; compared with prompt-only methods, it actually updates model behavior.
Run pip install peft and follow the README Quickstart to wrap a model with LoRA and print its trainable parameters.
Fine-tune and deploy 600+ LLMs and 300+ multimodal models with PEFT or full-parameter training.
Unified, efficient fine-tuning of 100+ LLMs and VLMs with a CLI and Web UI — ACL 2024
A coding-agent harness optimized for smaller local LLMs — small models, serious coding work
MCPJam Inspector: chat with, inspect, debug, and evaluate MCP servers across clients and models.
A free course to get into large language models, with roadmaps and runnable Colab notebooks
Hugging Face's library for loading and processing datasets for AI models in one line
The most-starred other Dev Tools projects in the LibreAI catalogue — the realistic shortlist if peft is not the right fit.
The open source coding agent
The model-definition framework behind modern AI — thousands of pretrained transformers for text, vision, and audio.
Anthropic's agentic coding tool that lives in your terminal and works across your whole codebase
Cross-platform desktop all-in-one manager for Claude Code, Codex, OpenCode and other coding agents
OpenAI's lightweight coding agent that runs locally in your terminal
Free and open — grab the source, star the repo, and start creating.