0.1.0-alpha

Minimal working baseline

• Decompose DT workflow into its fundamental steps: pre-processing and AI blocks
• New features for AI module: itwinai v0.0
• New features for AI module: itwinai v0.1
• Support YAML configuration files
• MNIST use case: image classification
• PyTorch Lightning support

AI module versions

itwinai v0.0 - Minimal working baseline (MWB)

• Input dataset path
• Pytorch Lightning (PL) Model name: already available inside itwinai module. Just need to be imported
• Simple numeric hyperparams: learning rate, batch size
• MNIST use case
• No logs are saved

itwinai v0.1 - Consolidated AI training

• Output logs path/URI
• Mlflow: logs saved to local filesystem
  • Log metrics (metrics logger)
  • Save model parameters
• Hyperparams:
  • Optimizer name (standard only, to be imported form torch)
  • Scheduler name (standard only, to be imported form torch)
  • PL callbacks