Visualization

This page describes the visualization components of the package.

Dashboard

The dashboard displays multiple plots that update live during training.

Currently implemented plots:

• Training loss vs Iteration
• Parameter distance (L2 distance from initial weights)
• Parameter update size (step-to-step distance)
• Gradient norm
• Gradient norm test (measuring signal-to-noise)
• Gradient distribution (1D histogram)

The dashboard automatically links the X-axis (Iteration) for all relevant plots, ensuring synchronized visualization of the training progress.

Makie integration

The package uses Makie.jl and observable variables to enable live updates of plots as training progresses.