🦙🌲🤏 Alpaca-LoRA: Low-Rank LLaMA Instruct-Tuning

The code in this repo is not yet fully tested. I'm still in the process of retraining the model with the outputs included, and I make no guarantees about the results of running generate.py.

This repository contains code for reproducing the Stanford Alpaca results using low-rank adaptations (LoRAs). The goal is to provide an open Instruct model of similar quality to text-davinci-003 that can run on most consumer GPUs with 8-bit quantization.

Users will need to be ready to fork Huggingface transformers to access Jason Phang's LLaMA implementation. For fine-tuning LoRAs we use Huggingface's PEFT. Included also is code to download the LLaMA foundation model from the Huggingface model hub (for research). Once I've finished running the finetuning code myself, I'll put the LoRA on the Hub as well, and the code in generate.py should work as expected.

Setup

1. Install dependencies (install zphang's transformers fork)

pip install -q datasets loralib sentencepiece

pip uninstall transformers
pip install -q git+https://github.com/zphang/transformers@c3dc391

pip install -q git+https://github.com/huggingface/peft.git

2. Install bitsandbytes from source.

Inference (generate.py)

See generate.py. This file reads the decapoda-research/llama-7b-hf model from the Huggingface model hub and the LoRA weights from tloen/alpaca-lora-7b, and runs inference on a specified input. Users should treat this as example code for the use of the model, and modify it as needed.

Training (finetune.py)

Under construction. If you're impatient, note that this file contains a set of hardcoded hyperparameters you should feel free to modify. PRs adapting this code to multi-GPU setups and larger models are always welcome.

To do

• Hyperparameter tuning
• Documentation for notebook
• Support for 13b, 30b, 65b
• Train a version that doesn't waste tokens on the prompt header
• Inference CLI and evaluation
• Better disclaimers about why using LLaMA without permission is very bad!