XVIS

XVIS / Data loop

Training runs

Which run is winning, and when did the diverging one leave the pack?

Input
Four runs: loss, val mAP, learning-rate schedule
Output
Smoothed curves, best run, divergence flag
Demos
1 · all run in the browser
Loading engine
Static charts
  • baseline
  • +tail-data
  • lr 3e-3
  • +aug

Built with Run comparison + divergence

Loss / mAP / LR for four runs with smoothing and run toggles; a divergence detector flags the run whose loss leaves the median envelope for 10 steps.