v0.22.3

Full Changelog: v0.22.2...v0.22.3

v0.22.2

Full Changelog: v0.22.1...v0.22.2

0.22.1

What's Changed

⚙️ ACTGAN improvements

• Provide error message for bad batch_size and pac values by @kboyd
• Update cbn_sample_size to 250_000 by @mvansegbroeck

🧹 Housekeeping

• Introduce new standardized error categories by @pimlock

Full Changelog: v0.22.0...v0.22.1

0.22.0

What's Changed

🔼 Dependency updates:

• update TF 2.8 refs to 2.11 (#158) @johntmyers 5760bcb

⚙️ ACTGAN improvements:

• Refactor ACTGAN in preparation of anyway training @kboyd 0ad0296

🧹 Miscellaneous housekeeping:

• Use loky that is vendored with joblib. @pimlock 067a324
• Bump sphinx to fix RTD issue @matthewgrossman 7441d70
• Unify sphinx-rtd-theme with gretel-client @matthewgrossman d193d37

0.21.0

What's Changed

🔼 Dependency updates:

• Require Python >= 3.9
• Update TensorFlow to 2.11.1
• Require Torch >= 1.13 (2.0 is recommended)
• Constraint other packages (sdv, pandas, numpy)

⚙️ ACTGAN improvements:

• Sampling improvements:
  • Use eval mode for inference during sampling in ACTGAN.
  • Vectorize conditional row selection ACTGAN sampling.
  • Fix sampling of conditional vectors in ACTGAN.
• Memory improvements.
• Factor out original condvec sampler for ACTGAN.
• Add cbn_sample_size param for ACTGAN model.

⚙️ DGAN improvements:

• Raise clearly worded exception if generation is called before training DGAN.
• Resolve apply_along_axis bug in DGAN.
• Skip key error during attempt cast to int.

⚙️ LSTM improvements:

• Reload LSTM model before generation.
• Change memory allocation.
• Reset generation callback state each time new generation starts.
• Update dp_microbatches default value to 1.

🐞 Bug fixes:

• Fix RDT FloatFormatter bug

🧹 Miscellaneous housekeeping and plumbing updates:

• Add a readthedocs config file to configure the Python version (#154)
• Fixed typos by @mahmadza in #148

New Contributors

• @mahmadza made their first contribution in #148

Full Changelog: v0.20.0...v0.21.0

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.