Home

Awesome

LLMzip

This repository contains the code for our paper LLMZip: Lossless Text Compression using Large Language Models

Setup

This repository is identical to the [LLaMA repository] (https://github.com/facebookresearch/llama) with additional scripts to perform compression. The setup is identical to that of LLaMA. LLaMA Setup is included below for ease of access

Compression

The code below can be used for compressing any text file ($TEXT_FILE) using LLaMa and Arithmetic Coding , the resulting compressed file will be stored in a specified folder ($COMPRESSION_FOLDER). $TARGET_FOLDER is the folder with LLaMa weights and tokenizer.

torchrun --nproc_per_node 1 LLMzip_run.py --ckpt_dir $TARGET_FOLDER/model_size --tokenizer_path $TARGET_FOLDER/tokenizer.model --win_len 511 --text_file $TEXT_FILE --compression_folder $COMPRESSION_FOLDER 

torchrun --nproc_per_node 1 LLMzip_run.py --ckpt_dir $TARGET_FOLDER/model_size --tokenizer_path $TARGET_FOLDER/tokenizer.model --win_len 511 --text_file $TEXT_FILE --compression_folder $COMPRESSION_FOLDER --encode_decode 0

Arithmetic Coding

The arithmetic coding implementation is from Deep Zip repo , which is based of the implementation by Project Nayuki

Llama Setup

In order to download the checkpoints and tokenizer, fill this google form

Setup

In a conda env with pytorch / cuda available, run:

pip install -r requirements.txt

Then in this repository:

pip install -e .

Download

Once your request is approved, you will receive links to download the tokenizer and model files. Edit the download.sh script with the signed url provided in the email to download the model weights and tokenizer.

Inference

The provided example.py can be run on a single or multi-gpu node with torchrun and will output completions for two pre-defined prompts. Using TARGET_FOLDER as defined in download.sh:

torchrun --nproc_per_node MP example.py --ckpt_dir $TARGET_FOLDER/model_size --tokenizer_path $TARGET_FOLDER/tokenizer.model

Different models require different MP values:

ModelMP
7B1
13B2
33B4
65B8

FAQ

Reference

LLaMA: Open and Efficient Foundation Language Models -- https://arxiv.org/abs/2302.13971

@article{touvron2023llama,
  title={LLaMA: Open and Efficient Foundation Language Models},
  author={Touvron, Hugo and Lavril, Thibaut and Izacard, Gautier and Martinet, Xavier and Lachaux, Marie-Anne and Lacroix, Timoth{\'e}e and Rozi{\`e}re, Baptiste and Goyal, Naman and Hambro, Eric and Azhar, Faisal and Rodriguez, Aurelien and Joulin, Armand and Grave, Edouard and Lample, Guillaume},
  journal={arXiv preprint arXiv:2302.13971},
  year={2023}
}

Model Card

See MODEL_CARD.md

License

See the LICENSE file.