Training Observability
Select an experiment from the left to inspect its training run: loss function, dataset, checkpoint location, notes, per-batch metric charts, and the latest evaluation values.
·
Latest evaluation
from the most recent logged batchRun metadata
provider-neutral identity and provenanceRun details
editable · changes save on blur/Enter
Run ID
Created
✓ saved
Notes
free-form text for anything about this runPerformance vs step
x-axis: training step · hover for valuesNo metrics logged yet.