Domovská stránka Prehľadávať Pomoc
Prihlásiť sa
dongzenan
/
alpaca-lora
1
0
Fork 0
Súbory Issues 0 Pull requesty 0 Wiki

Žiadny popis

Strom: 63121244c8
Branche Tagy
alpaca-lora
Eric Wang 63121244c8 Licenses and whatnot 3 rokov pred
.gitignore 26f64780ad initial commit 3 rokov pred
DATA_LICENSE 63121244c8 Licenses and whatnot 3 rokov pred
LICENSE 63121244c8 Licenses and whatnot 3 rokov pred
README.md 63121244c8 Licenses and whatnot 3 rokov pred
alpaca_data.json 26f64780ad initial commit 3 rokov pred
conversion.py 26f64780ad initial commit 3 rokov pred
finetune.py 63121244c8 Licenses and whatnot 3 rokov pred
generate.py 63121244c8 Licenses and whatnot 3 rokov pred
iteration.ipynb 26f64780ad initial commit 3 rokov pred
lengths.ipynb 26f64780ad initial commit 3 rokov pred

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