v0.1.1 - 2025-03-23

🚀 Features

• (callbacks) Added training callbacks.
• (watch) Added method to watch trained agents (notebooks).
• (record) Added RecordVideos callback with minor optimizations
• (train) Enhanced agent training loop and metric storage.
• (metrics) Moved TrainMetrics intoTrainState for callbacks.
• (gradients) Added method to compute gradients of NCP networks.
• (analytics) Added CometAnalytics callback for cloud tracking.
• (version) Made velora compatible with Python 3.11.

🐛 Bug Fixes

• (buffer) Reorganized buffer items to fix circular imports.
• (config) Fixed config copy bug.
• (imports) Fixed import circulation errors.
• (buffer) Fixed actions for multi-dimensional spaces.
• (ddpg) Fixed DDPG learning bug.
• (wiring) Fixed number of neuron connections bug.
• (metrics) Fixed metrics std() call bug when window_size=1.

🚜 Refactor

• (callbacks) Moved callbacks location for clarity.
• (save) Enhanced saved method to include config.json.
• (buffer) Added warm method to ReplayBuffer for simplicity.
• (config) Simplified config creation in LiquidDDPG.
• (watch) Removed watch_notebook method - redundant.
• (callbacks) Simplified callback handling in train method.
• (old) Removed old redundant code.
• (config) Simplified agent config creation.
• (callback) Refactored SaveCheckpoints to simplify init.
• (metrics) Simplified and enhanced metric implementation.
• (metrics) Added attributes for stored metrics.
• (callbacks) Modified RecordVideo env wrapper logic.
• (metrics) Minor updates + custom gradient removal.
• (callbacks) Simplified RecordVideos callback parameters.
• (metrics) Updated metric tracking to use SQLite database.
• (record) Added root_path to record_last_episode method.
• (checkpoints) Updated checkpoint target condition to simplify.
• (lnn) Updated LNN architecture for accuracy.
• (lnn) QoL additions and improvements.
• (ncp) Simplified and improved architecture.
• (tweaks) Minor changes to DDPG and buffer methods.
• (ncp) Added NCPModule base class to simplify Actor-Critics.
• (train) Improved console training output.
• (sparse) Removed SparseParameter (deprecated).
• (output) Added running step count to training output.
• (save) Updated saving and loading to safetensors.

📚 Documentation

• (callbacks) Added and updated documentation for better flow.
• (config) Added config API reference and updated save docs.
• (install) Updated install docs for GPU and CPU versions.
• (callbacks) Added docs for RecordVideos & updated existing ones.
• (metrics) Added docs for training metrics.
• (roadmap) Added content to roadmap.
• (additions) Updated and added docs for Comet & new metric system.
• (docstrings) Updated codebase docstrings for clarity.
• (lnn) Updated customization docs for clarity.
• (updates) Updated Buffer and Metrics docs with new changes.
• (save): Added API references for saving and loading models.

⚡ Performance

• (buffer) Reduced total training time by 63%.
• (metrics) Reduced total training time by a further 40%.
• (ncp) Reduced training time with NCP compiling.
• (compile) Simplified compilation to agent level.

🎨 Styling

• (output) Updated format for training output.
• (train) Improved training output details.

🧪 Testing

• (buffer) Fixed failed tests.
• (callbacks) Added unit tests for new feature coverage.
• (save) Added and fixed broken tests for saving and loading.
• (cov) Added tests for coverage.
• (fix) Updated test container packages & removed redundant tests.
• (fix) Fixed broken tests for new training logic.
• (fix) Fixed broken test cases for TrainState.metrics refactor.
• (fix) Added tests and fixed broken ones.
• (fix) Fixed broken DDPG test.
• (fix) Fixed unit tests.
• (stopping) Fixed EarlyStoppingtests.
• (fix) Fixed broken tests with recent code changes.
• (all) Updated and added tests cases for coverage.
• (fix) Removed pre-release code to fix test cases.
• (fix) Updated tests to accommodate recent changes.

⚙️ Miscellaneous Tasks

• (build) Added matplotlib package as main dependency.
• (build) Removed redundant packages and updated existing ones.
• (packages) Refactored packages to fix torch install bug.
• (torch) Moved torch dependencies to group.