Optimal settings for training a synthetic data generation model on the UCI heart disease dataset from Kaggle

gretel_synthetics_config-uci.py

from pathlib import Path

from gretel_synthetics.config import LocalConfig

# Create a config that we can use for both training and generating, with CPU-friendly settings
# The default values for ``max_chars`` and ``epochs`` are better suited for GPUs
config = LocalConfig(
    max_lines=0, # read all lines (zero)
    epochs=15, # 15-30 epochs for production
    vocab_size=20000, # tokenizer model vocabulary size
    character_coverage=1.0, # tokenizer model character coverage percent
    gen_chars=0, # the maximum number of characters possible per-generated line of text
    gen_lines=10000, # the number of generated text lines
    rnn_units=256, # dimensionality of LSTM output space
    batch_size=64, # batch size
    buffer_size=1000, # buffer size to shuffle the dataset
    dropout_rate=0.2, # fraction of the inputs to drop
    dp=False, # let's use differential privacy
    dp_learning_rate=0.015, # learning rate
    dp_noise_multiplier=1.1, # control how much noise is added to gradients
    dp_l2_norm_clip=1.0, # bound optimizer's sensitivity to individual training points
    dp_microbatches=256, # split batches into minibatches for parallelism
    checkpoint_dir=(Path.cwd() / 'checkpoints').as_posix(),
    save_all_checkpoints=False,
    input_data_path=annotated_file # filepath or S3
)