lanpa/tensorboardX

Features

• ML Metric Logging - Writes scalars, images, audio, and graphs to event files for visualizing model performance.
• Framework-Agnostic Logging - Provides a standardized set of writers that decouple data recording from specific deep learning engines.
• Machine Learning Experiment Trackers - Provides a system for monitoring metrics and visual data to track machine learning experiments.
• TensorBoard Event Generators - Generates event files specifically compatible with the TensorBoard visualization dashboard.
• Visualization File Serialization - Serializes training metrics into binary formats compatible with visualization dashboards.
• Local Event Generation - Generates visualization log files locally on the training machine rather than streaming them to a remote server.
• ML Visualization Libraries - Generates visual representations of model behavior using graphs and images from training data.
• Dashboard Integrations - Connects PyTorch training loops to visualization interfaces for real-time monitoring of loss curves and scalars.
• Training Log Analysis - Enables analysis of model training progress through the creation of event files for centralized dashboards.
• Protobuf Serialization - Uses protocol buffers to ensure cross-platform compatibility between the logging library and the visualization viewer.
• Visualization and Analysis - Utility for writing experiment events for visualization.

Star history