trainer

Gretel Trainer

This module is designed to provide a simple interface to help users successfully train synthetic models on complex datasets with high row and column counts, and offers features such as Cloud SaaS based training and multi-GPU based parallelization. Get started for free with an API key from Gretel.ai.

Current functionality and features:

• Synthetic data generators for text, tabular, and time-series data with the following features:
  • Balance datasets or boost a minority class using Conditional Data Generation.
  • Automated data validation.
  • Synthetic data quality reports.
  • Privacy filters and optional differential privacy support.
• Multiple model types supported:
  • Gretel-LSTM model type supports text, tabular, time-series, and conditional data generation.
  • Gretel-ACTGAN model type supports tabular and conditional data generation.
  • Gretel-GPT natural language synthesis based on an open-source implementation of GPT-3 (coming soon).
  • Gretel-DGAN multi-variate time series based on DoppelGANger (coming soon).

Try it out now!

If you want to quickly get started synthesizing data with Gretel.ai, simply click the button below and follow the examples. See additional Python3 and Jupyter Notebook examples in the ./notebooks folder.

Join the Synthetic Data Community Discord

If you want to be part of the Synthetic Data Community to receive announcements of the latest releases, ask questions, suggest new features or participate in the development meetings, please join the Synthetic Data Community Server!

Install

Using pip:

pip install -U gretel-trainer

Quickstart

1. Add your Gretel API key via the Gretel CLI.

Use the Gretel client to store your API key to disk. This step is optional, the trainer will prompt for an API key in the next step.

gretel configure

2. Train or fine-tune a model using the Gretel API

from gretel_trainer import trainer

dataset = "https://gretel-public-website.s3-us-west-2.amazonaws.com/datasets/USAdultIncome5k.csv"

model = trainer.Trainer()
model.train(dataset)

3. Generate synthetic data!

df = model.generate()

Development

Setup environment and install dependencies.

python -m venv venv
source venv/bin/activate
pip install -r requirements-dev.txt
pip install -e .

• Run tests via make test
• Run type-checking (limited coverage) via make type

TODOs / Roadmap

• Enable conditional generation via SDK interface (supported in Notebooks currently).