hiyouga/FastEdit

Health v2 · maintenance only

Adoption not part of the score

1366 stars · 103 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

FastEdit is a Python library and CLI tool for editing large language models, injecting fresh or customized factual knowledge into pretrained models in seconds using the Rank-One Model Editing (ROME) algorithm. It supports models such as GPT-J, LLaMA/LLaMA-2, BLOOM, Falcon, Baichuan, and InternLM via Hugging Face Transformers.

Use cases

• inject new facts into a llama model without retraining
• update outdated knowledge in a large language model
• edit a model's answer to a factual question
• quickly patch factual errors in an llm
• run rank-one model editing on gpt-j or falcon
• customize chatbot knowledge with a single command

When to choose

• you need to update specific factual knowledge in an open-source LLM without full fine-tuning
• you want a fast, single-command model editing workflow with ROME
• you work with supported architectures like LLaMA, BLOOM, Falcon, or Baichuan and have a 24GB+ GPU

When to avoid

• you need broad capability changes rather than targeted factual edits
• you require frequent large-scale knowledge updates, which fine-tuning or RAG handles better
• your model architecture is not among the supported ones
• you need actively developed tooling, as the project has seen limited recent releases

Facets

library · maturity maintenance

llm-training machine-learning cli large-language-models machine-learning python model-editing knowledge-injection rome transformers pytorch llama factual-editing natural-language-processing gpu linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="hiyouga/FastEdit")