AI/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2024-10-01 01:06:10 -04:00
Code Issues Packages Projects Releases Wiki Activity
gpt4all: an ecosystem of open-source chatbots trained on a massive collections of clean assistant data including code, stories and dialogue
76 Commits 491 Branches 56 Tags 46 MiB
C++ 49.1%
QML 31.6%
Python 8.3%
CMake 5.4%
JavaScript 3.8%
Other 1.6%
393e0f951d
Go to file
Andriy Mulyar 393e0f951d
Update README.md
2023-03-28 17:06:28 -04:00
chat Added binaries to chat 2023-03-28 16:35:39 -04:00
configs Update generate.yaml 2023-03-28 13:59:26 -07:00
eval_data started eval script and added eval data 2023-03-27 21:50:08 +00:00
figs Merge branch 'main' of https://github.com/nomic-ai/gpt4all into main 2023-03-28 20:23:42 +00:00
peft@098962fa65 feat: peft submodule 2023-03-25 16:23:14 +00:00
transformers@cae78c46d6 feat: transformers submodule, gitignore 2023-03-25 16:16:11 +00:00
.gitignore chore: gitignore 2023-03-27 16:33:00 +00:00
.gitmodules feat: peft submodule 2023-03-25 16:23:14 +00:00
clean.py fix: naming 2023-03-27 17:30:33 +00:00
data.py fix: just read from watermark file 2023-03-27 17:30:44 +00:00
env.yaml feat: env for conda, pip 2023-03-25 16:16:40 +00:00
eval_figures.py updated eval 2023-03-28 20:22:48 +00:00
eval_self_instruct.py added eval code 2023-03-28 18:47:38 +00:00
generate.py metrics run on configs now 2023-03-28 00:09:47 +00:00
gpt4all-lora-demo.gif GIF 2023-03-28 15:54:44 -04:00
read.py feat: train and clean data 2023-03-25 16:17:48 +00:00
README.md Update README.md 2023-03-28 17:06:28 -04:00
requirements.txt feat: env for conda, pip 2023-03-25 16:16:40 +00:00
train.py fix: append first_epoch 2023-03-27 17:30:24 +00:00
TRAINING_LOG.md Update TRAINING_LOG.md 2023-03-28 13:47:22 -07:00

README.md

GPT4All

Demo, data and code to train an assistant-style large language model with ~800k GPT-3.5-Turbo Generations

📗 Technical Report

gpt4all-lora-demo Run on M1 Mac (not sped up!)

Try it yourself

Clone this repository down and download the CPU quantized gpt4all model.

Place the quantized model in the chat directory and start chatting by running:

  • ./chat/gpt4all-lora-quantized-OSX-m1 on M1 Mac/OSX
  • ./chat/gpt4all-lora-quantized-linux-x86 on Windows/Linux

To compile for custom hardware, see our fork of the Alpaca C++ repo.

Note: the full model on GPU (16GB of RAM required) performs much better in our qualitative evaluations.

Reproducibility

Trained LoRa Weights:

Raw Data:

We are not distributing a LLaMa 7B checkpoint.

You can reproduce our trained model by doing the following:

Setup

Clone the repo

git clone --recurse-submodules git@github.com:nomic-ai/gpt4all.git

git submodule configure && git submodule update

Setup the environment

python -m pip install -r requirements.txt

cd transformers
pip install -e . 

cd ../peft
pip install -e .

Training

accelerate launch --dynamo_backend=inductor --num_processes=8 --num_machines=1 --machine_rank=0 --deepspeed_multinode_launcher standard --mixed_precision=bf16  --use_deepspeed --deepspeed_config_file=configs/deepspeed/ds_config.json train.py --config configs/train/finetune-7b.yaml

Generate

python generate.py --config configs/generate/generate.yaml --prompt "Write a script to reverse a string in Python

If you utilize this reposistory, models or data in a downstream project, please consider citing it with:

@misc{gpt4all,
  author = {Yuvanesh Anand and Zach Nussbaum and Brandon Duderstadt and Benjamin Schmidt and Andriy Mulyar},
  title = {GPT4All: Training an Assistant-style Chatbot with Large Scale Data Distillation from GPT-3.5-Turbo},
  year = {2023},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/nomic-ai/gpt4all}},
}