- Fix float32 serialization issue in unified logger metrics flushing
- Add jax.numpy import for proper type handling
- Add wandb/ directory to .gitignore to exclude temporary tracking files
- Tested wandb integration: metrics, artifacts, and local backup working correctly
- Add Quick Start guide with installation and configuration steps
- Document unified logging framework features
- Explain configuration management for multiple researchers
- Add project structure overview showing experiment_logger
- Update training examples with new patterns
- Add logging best practices to CONTRIBUTING.md
- Add default_ppo.yaml template for training configurations
- Create configs/README.md with usage documentation
- Enable per-researcher configuration without code changes
- Support YAML config files with CLI parameter overrides
- Document how to set personal WandB credentials safely
This allows researchers to maintain personal configs without
committing credentials to the repository.
- Replace direct WandB calls with experiment_logger.UnifiedLogger
- Replace all print() calls with proper logging framework
- Add automatic checkpoint saving every N iterations
- Configure root logger with proper format and level
- Maintain backward compatibility with TensorBoard writer
- Save final model with metadata using unified logger
Create reusable experiment logging package with:
- UnifiedLogger for multi-backend logging (WandB, disk, stdout)
- Automatic checkpoint and model saving with metadata
- WandB artifact upload support
- Graceful degradation when WandB unavailable
- Comprehensive API documentation
This is a standalone, project-agnostic library that can be reused
across different ML projects.
