BojAI allows you to save and load complete training sessions.
You can use these features from both the CLI and UI, making it easy to pause, resume, or revisit past experiments.

Saving a Session

Example of Saved JSON

{
  "model_name": "YourModel",
  "hyperparameters": {
    "batch_size": 32,
    "learning_rate": 0.00001
  },
  "loggs": {
    "0": {"train": 0.87, "valid": 0.83, "loss": 0.6},
    "1": {"train": 0.89, "valid": 0.85, "loss": 0.4}
  },
  "timestamp": "2025-04-19T18:00:00",
  "model_weights": "your_session_model.bin"
}

Loading a Session

To resume from a saved session you can load a session, this will:

• Load the model weights from {session_name}_model.bin
• Restore your hyperparameters
• Restore the logger’s training and evaluation logs
• Print the logs to confirm successful loading

If either file is missing, a message will be shown and loading is aborted.

Files Saved per Session

These files are located inside the train_sessions/ folder inside the applets folder.

Use Cases

• Resume training from where you left off
• Visualize training/evaluation without retraining
• Track experiments with timestamped logs
• Share sessions across machines

Let me know if you’d like session versioning, auto-saving, or cloud integration added next!