model-one 0.1.14

beyondml

Model One python wrapper

Model One is a few-shot AutoML system.

Model One can do almost anything that requires understanding or generating natural language. Model One is able to solve tasks in 12 languages: English, Spanish, Portuguese, Russian, Turkish, French, German, Italian, Arabic, Polish, Dutch, and Hebrew.

This package provides a simple wrapper for using our api.

Using model-one package allows you to train and apply models.

Documentation

You can find the documentation at our ModelOne API docs site.

Just try

We have fine-tuned several models. You can use the notebook to try them out. You can get the token to fine tune your own model.

Getting started

Firstly fill out the form to get a key to access the API. We will send you the key within a day.

Installation

To install the package just use pip install -U model-one.

Usage

import model_one
import pandas as pd

model_one.set_api_key('YOUR_API_KEY')

# load datasets
tdf = pd.read_csv('train.csv')
vdf = pd.read_csv('test.csv')

# Call one method. It will do everything for you:
# create a model, save it to the file, upload datasets and put the model in the queue for training.
model = model_one.train_generator(
    'filename.json',
    tdf['inputs'], tdf['outputs'],
    vdf['inputs'], vdf['outputs']
)

# wait...
# a few hours
# while our GPUs train your model
model.wait_for_training_finish()

print(model.status)
print(model.is_ready)

# inference!
the_answer = model.generate('The Answer to the Ultimate Question of Life, the Universe, and Everything')
print(the_answer)