0.20.0

• 🚀 New model available! ACTGAN (Anyway CTGAN) is an alternate implementation of the SDV CTGAN model that provides additional functionality to improve memory usage, autodetection and transformation of columns, and more.
• 💨 Performance improvements to DGAN and SQS
• 🔧 Bug fixes, including improving handling of inf, NaN, and regex metachars, and more
• 🖊️ Dependency updates
  • Pin tensorflow_estimator at 2.8
  • Update sentencepiece to 0.1.97
  • Add category-encoders 2.2.2

Full Changelog: v0.19.0...v0.20.0

0.19.0

📖 Apply Gretel's Source Available License

DGAN updates!

🐛 + ⚙️: Bug fixes and optimizations for the DGAN timeseries synthetic model!

Full Changelog: v0.18.0...v0.18.1

v0.18.0

📜 GPU helper script updates + README

v0.18.0 RC1

RC for v0.18.0

⚙️ New model available! DoppelGANger for time series 📈 synthetic data generation!
⚙️ Updates to header clustering logic, added options for isolation batches based on average record length and field complexity
🐞 Bug fixes during record generation and for epsilon calculation when using DP
🔼 Updates to recommended core dependencies, specifically Python 3.9 and TensorFlow 2.8.

New stats and header clustering modules

• 🚀 Adds a new gretel_synthetics.utils module containing stats and header clustering helpers. These modules should be installed via the new utils extra

pip install gretel-synthetics[utils]

• 🧹 Miscellaneous housekeeping and plumbing updates.

v0.16.12

⚙️ Slight performance improvements for TensorFlow and memory footprint reductions

Early Stopping Updates

Enable early stopping for:

• Specified min_delta for the metric used between epochs
• Enable early stopping after a maximum run time

Beta2 Release

🚀 New release to support Gretel Beta2!

Bug fix

🐛 Fixed DataFrame buffering where fields may be missing from a generated record