Home Verkennen Help
Inloggen
dongzenan
/
alpaca-lora
1
0
Vork 0
Bestanden Issues 0 Pull-aanvragen 0 Wiki

Geen omschrijving

Boom: 63121244c8
Aftakkingen Labels
alpaca-lora
Eric Wang 63121244c8 Licenses and whatnot 3 jaren geleden
.gitignore 26f64780ad initial commit 3 jaren geleden
DATA_LICENSE 63121244c8 Licenses and whatnot 3 jaren geleden
LICENSE 63121244c8 Licenses and whatnot 3 jaren geleden
README.md 63121244c8 Licenses and whatnot 3 jaren geleden
alpaca_data.json 26f64780ad initial commit 3 jaren geleden
conversion.py 26f64780ad initial commit 3 jaren geleden
finetune.py 63121244c8 Licenses and whatnot 3 jaren geleden
generate.py 63121244c8 Licenses and whatnot 3 jaren geleden
iteration.ipynb 26f64780ad initial commit 3 jaren geleden
lengths.ipynb 26f64780ad initial commit 3 jaren geleden

README.md

alpaca-lora

This repository contains code for reproducing the Stanford Alpaca results. Users will need to have LLaMA weights on hand and be ready to fork transformers.

  1. Install dependencies

    pip install -q bitsandbytes datasets accelerate loralib

pip install -q git+https://github.com/zphang/transformers@llama_push
pip install -q git+https://github.com/huggingface/peft.git\

  2. Convert weights

    python conversion.py --input_dir [LLAMA_DIR]/LLaMA --model_size 7B --output_dir ./7B

  3. Modify hyperparams in finetune.py

    MICRO_BATCH_SIZE = 12
BATCH_SIZE = 36
EPOCHS = 3
LEARNING_RATE = 2e-5

  4. Run experiments

    python finetune.py