Dashboard
Experiment intelligence
Select a configuration or run
Choose a model-selection configuration or a final run to compare its loss and score histories across epochs or recorded training steps.
Metric source
Plotter → metrics_data.torch
Run
Selection
—
Checkpoint graph
Model graph
Checkpoint graph
Model graph
Live comparison
Model selection analysis
Build a mixed collection of training trends and checkpoint-aware metric comparisons.
Waiting for data
Axes (drag ranges to select curves)
Observed search space
Configuration space
Reconstruct the values actually materialized for every configuration field in one outer fold.
Config-like reconstruction
Values tried