ITrainingDashboard

Interfaces · AiDotNet.TrainingMonitoring.Dashboard

Interface for training dashboards that visualize metrics and training progress.

Implements: IDisposable

Known implementations: ConsoleDashboard, HtmlDashboard, LiveDashboard

How It Works

For Beginners: A training dashboard provides visual feedback about your model’s training progress. Think of it like TensorBoard - it shows you graphs of loss, accuracy, and other metrics over time.

This library provides multiple dashboard implementations:

  • HtmlDashboard: Generates interactive HTML reports
  • LiveDashboard: Provides real-time updates via embedded web server
  • ConsoleDashboard: Text-based visualization for terminal environments

Example usage:

Properties

Property Summary
bool IsRunning { get; } Gets whether the dashboard is running.
string LogDirectory { get; } Gets the log directory.
string Name { get; } Gets the dashboard name.

Methods

Method Summary
Dictionary<string, List<HistogramDataPoint>> GetHistogramData() Gets all logged histogram series.
Dictionary<string, List<ScalarDataPoint>> GetScalarData() Gets all logged scalar series.
string GenerateReport(string outputPath) Generates an HTML report of all logged data.
void Clear() Clears all logged data.
void ExportTensorBoardFormat(string outputDirectory) Exports data in TensorBoard-compatible format.
void Flush() Flushes any buffered data to storage.
void LogConfusionMatrix(string name, long step, int[] matrix, string[] labels, DateTime? wallTime) Logs a confusion matrix.
void LogHistogram(string name, long step, double[] values, DateTime? wallTime) Logs a histogram (distribution) of values.
void LogHyperparameters(Dictionary<string, object> hyperparams, Dictionary<string, double> metrics) Logs hyperparameters.
void LogImage(string name, long step, byte[] imageData, int width, int height, DateTime? wallTime) Logs an image.
void LogModelGraph(string modelDescription) Logs the model graph/architecture.
void LogPRCurve(string name, long step, double[] predictions, int[] labels, DateTime? wallTime) Logs a precision-recall curve.
void LogROCCurve(string name, long step, double[] predictions, int[] labels, DateTime? wallTime) Logs an ROC curve.
void LogScalar(string name, long step, double value, DateTime? wallTime) Logs a scalar metric.
void LogScalars(Dictionary<string, double> scalars, long step, DateTime? wallTime) Logs multiple scalars at once.
void LogText(string name, long step, string text, DateTime? wallTime) Logs text.
void Start() Starts the dashboard.
void Stop() Stops the dashboard.