mrjdomingus / gpt4all-j

Python bindings for the C++ port of GPT4All-J model.

GPT4All-J

Python bindings for the C++ port of GPT4All-J model.

Installation

pip install gpt4all-j

Download the model from here.

Usage

from gpt4allj import Model

model = Model('/path/to/ggml-gpt4all-j.bin')

print(model.generate('AI is going to'))

Run in Google Colab

If you are getting illegal instruction error, try using instructions='avx' or instructions='basic':

model = Model('/path/to/ggml-gpt4all-j.bin', instructions='avx')

If it is running slow, try building the C++ library from source. Learn more

Parameters

model.generate(prompt,
               seed=-1,
               n_threads=-1,
               n_predict=200,
               top_k=40,
               top_p=0.9,
               temp=0.9,
               n_batch=8,
               repeat_penalty=1.0,
               repeat_last_n=64,
               callback=None)

callback

If a callback function is passed to model.generate(), it will be called once per each generated token. To stop generating more tokens, return False inside the callback function.

def callback(token):
    print(token)

model.generate('AI is going to', callback=callback)

C++ Library

To build the C++ library from source, please see gptj.cpp. Once you have built the shared libraries, you can use them as:

from gpt4allj import Model, load_library

lib = load_library('/path/to/libgptj.so', '/path/to/libggml.so')

model = Model('/path/to/ggml-gpt4all-j.bin', lib=lib)

License

MIT