Demo, data, and code to train open-source assistant-style large language model based on GPT-J and LLaMa

📗 Technical Report 2: GPT4All-J

📗 Technical Report 1: GPT4All

🐍 Official Python Bindings

💻 Official Typescript Bindings

💬 Official Web Chat Interface

💬 Official Chat Interface

🦜️🔗 Official Langchain Backend

Discord

GPT4All is made possible by our compute partner Paperspace.

Run on an M1 Mac (not sped up!)

Installs a native chat-client with auto-update functionality that runs on your desktop with the GPT4All-J model baked into it.

Mac/OSX

Windows

Ubuntu

These files are not yet cert signed by Windows/Apple so you will see security warnings on initial installation. We did not want to delay release while waiting for their process to complete.

Find the most up-to-date information on the GPT4All Website

ggml Model Download Link

Note this model is only compatible with the C++ bindings found here. It will not work with any existing llama.cpp bindings as we had to do a large fork of llama.cpp. GPT4All will support the ecosystem around this new C++ backend going forward.

Python bindings are imminent and will be integrated into this repository. Stay tuned on the GPT4All discord for updates.

Please see GPT4All-J Technical Report for details.

• We are releasing the curated training data for anyone to replicate GPT4All-J here: GPT4All-J Training Data
  • Atlas Map of Prompts
  • Atlas Map of Responses

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_gptj.json train.py --config configs/train/finetune_gptj.yaml

Run on M1 Mac (not sped up!)

Here's how to get started with the CPU quantized GPT4All model checkpoint:

1. Download the gpt4all-lora-quantized.bin file from Direct Link or [Torrent-Magnet].
2. Clone this repository, navigate to chat, and place the downloaded file there.
3. Run the appropriate command for your OS:
   • M1 Mac/OSX: cd chat;./gpt4all-lora-quantized-OSX-m1
   • Linux: cd chat;./gpt4all-lora-quantized-linux-x86
   • Windows (PowerShell): cd chat;./gpt4all-lora-quantized-win64.exe
   • Intel Mac/OSX: cd chat;./gpt4all-lora-quantized-OSX-intel

For custom hardware compilation, see our llama.cpp fork.

Find all compatible models in the GPT4All Ecosystem section.

Secret Unfiltered Checkpoint - [Torrent]

This model had all refusal to answer responses removed from training. Try it with:

• M1 Mac/OSX: cd chat;./gpt4all-lora-quantized-OSX-m1 -m gpt4all-lora-unfiltered-quantized.bin
• Linux: cd chat;./gpt4all-lora-quantized-linux-x86 -m gpt4all-lora-unfiltered-quantized.bin
• Windows (PowerShell): cd chat;./gpt4all-lora-quantized-win64.exe -m gpt4all-lora-unfiltered-quantized.bin
• Intel Mac/OSX: cd chat;./gpt4all-lora-quantized-OSX-intel -m gpt4all-lora-unfiltered-quantized.bin

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

To run GPT4All in python, see the new official Python bindings.

The old bindings are still available but now deprecated. They will not work in a notebook environment. To get running using the python client with the CPU interface, first install the nomic client using pip install nomic Then, you can use the following script to interact with GPT4All:

from nomic.gpt4all import GPT4All
m = GPT4All()
m.open()
m.prompt('write me a story about a lonely computer')

There are two ways to get up and running with this model on GPU. The setup here is slightly more involved than the CPU model.

1. clone the nomic client repo and run pip install .[GPT4All] in the home dir.
2. run pip install nomic and install the additional deps from the wheels built here

Once this is done, you can run the model on GPU with a script like the following:

from nomic.gpt4all import GPT4AllGPU
m = GPT4AllGPU(LLAMA_PATH)
config = {'num_beams': 2,
          'min_new_tokens': 10,
          'max_length': 100,
          'repetition_penalty': 2.0}
out = m.generate('write me a story about a lonely computer', config)
print(out)

Where LLAMA_PATH is the path to a Huggingface Automodel compliant LLAMA model. Nomic is unable to distribute this file at this time. We are working on a GPT4All that does not have this limitation right now.

You can pass any of the huggingface generation config params in the config.

Edge models in the GPT4All Ecosystem. Please PR as the community grows. Feel free to convert this to a more structured table.

• gpt4all [MD5 Signature]
  • gpt4all-ggml-converted [MD5 Signature]
• gpt4all-unfiltered [MD5 Signature]
• ggml-vicuna-7b-4bit
• vicuna-13b-GPTQ-4bit-128g
• LLaMa-Storytelling-4Bit
• Alpaca Native 4bit

• (Done) Train a GPT4All model based on GPTJ to alleviate llama distribution issues.
• (Done) Create improved CPU and GPU interfaces for this model.
• (Done) Integrate llama.cpp bindings
• (Done) Create a good conversational chat interface for the model.
• (Done) Allow users to opt in and submit their chats for subsequent training runs

• (NOT STARTED) Integrate GPT4All with Atlas to allow for document retrieval.
  • BLOCKED by GPT4All based on GPTJ
• (Done) Integrate GPT4All with Langchain.
• (IN PROGRESS) Build easy custom training scripts to allow users to fine tune models.

• (NOT STARTED) Allow anyone to curate training data for subsequent GPT4All releases using Atlas.
• (IN PROGRESS) Democratize AI.

Trained Model Weights:

• gpt4all-lora (four full epochs of training): https://huggingface.co/nomic-ai/gpt4all-lora
• gpt4all-lora-epoch-2 (three full epochs of training) https://huggingface.co/nomic-ai/gpt4all-lora-epoch-2
• gpt4all-j (one full epoch of training) (https://huggingface.co/nomic-ai/gpt4all-j)
• gpt4all-j-lora (one full epoch of training) (https://huggingface.co/nomic-ai/gpt4all-j-lora)

Raw Data:

• Training Data Without P3
  • Explorer: https://atlas.nomic.ai/map/gpt4all_data_clean_without_p3
• Full Dataset with P3
  • Explorer: https://atlas.nomic.ai/map/gpt4all_data_clean
• GPT4All-J Dataset
  • Explorer Indexed on Prompts: https://atlas.nomic.ai/map/gpt4all-j-prompts-curated
  • Exporer Indexed on Responses: https://atlas.nomic.ai/map/gpt4all-j-response-curated

We are not distributing a LLaMa 7B checkpoint.

You can reproduce our trained model by doing the following:

Clone the repo

git clone --recurse-submodules https://github.com/nomic-ai/gpt4all.git
git submodule update --init

Setup the environment

python -m pip install -r requirements.txt

cd ../peft
pip install -e .

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

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

Join the Discord and ask for help in #gpt4all-help

Leg Raises

1. Stand with your feet shoulder-width apart and your knees slightly bent.
2. Raise your arms straight out in front of you.
3. Slowly bend your knees and raise your heels off the ground.
4. Hold this position for a count of 3, then slowly lower your heels back to the ground.
5. Repeat the exercise 10 times.
6. Do not let your knees go beyond your toes.
7. Keep your back straight and your abdominal muscles engaged.
8. Do not let your hips sag or your back arch.
9. Keep your arms straight and your elbows locked.
10. Do not let your head drop or your shoulders rise.
11. Breathe normally throughout the exercise.
12. Rest for 30 seconds before repeating the exercise.
13. Gradually increase the number of repetitions as you become stronger.

The CSS code associated with the color provided is: #FF6347

1. The title of your movie plot should be "The Last Stand".
2. The summary of your movie plot should be: A group of mercenaries are hired to protect a small town from a dangerous criminal who is on the run. However, the town's sheriff is not willing to let them take over and decides to take matters into his own hands.

In Python, you can reverse a list or tuple by using the reversed() function on it. Here's an example of how to use this method with strings:

my_string = "Hello World" # Define your original string here
reversed_str = my_string[::-1]   # Use a slice and negative index for reverse order
print(reversed_str)     # Output will be reversed version of the inputted string.

1. Labrador Retriever
2. Golden Retriever
3. Beagle
4. Poodle
5. German Shepherd
6. Bulldog
7. Rottweiler
8. Boxer
9. Shih Tzu
10. Pug

The fall of Julius Caesar into a caesar salad
In iambic pentameter,
The mighty Roman emperor
Fell into a salad of lettuce and croutons,
His empire crumbling around him,
As he was devoured by the hungry diners.
The once mighty emperor
Was now just a salad topping,
His legacy reduced to a mere garnish,
As he was consumed by the masses.
The fall of Julius Caesar into a caesar salad,
In iambic pentameter,
Was a tragic end to a great man,
Who had once ruled the world with an iron fist.
But now, he was just a salad topping,
His legacy reduced to a mere garnish,
As he was consumed by the masses.

Sports, athletics, games

If you utilize this repository, 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}},
}