changjonathanc/minLoRA
A minimal ~100-line PyTorch library for applying LoRA (Low-Rank Adaptation) to any PyTorch model for efficient fine-tuning.
Not currently ranked — collecting fresh signals.
star history
minLoRA re-implements LoRA using PyTorch’s native parametrize utilities, enabling parameter-efficient fine-tuning without modifying model definitions. It supports adding LoRA to linear layers, embeddings, and other modules via a functional API. The library provides utilities for training with LoRA params, merging weights for inference, and loading multiple LoRA adapters for flexible model switching.
Frequently asked
- What is changjonathanc/minLoRA?
- A minimal ~100-line PyTorch library for applying LoRA (Low-Rank Adaptation) to any PyTorch model for efficient fine-tuning.
- Is minLoRA open source?
- Yes — changjonathanc/minLoRA is open source, released under the MIT license.
- What language is minLoRA written in?
- changjonathanc/minLoRA is primarily written in Jupyter Notebook.
- How popular is minLoRA?
- changjonathanc/minLoRA has 493 stars on GitHub.
- Where can I find minLoRA?
- changjonathanc/minLoRA is on GitHub at https://github.com/changjonathanc/minLoRA.