feat: Train › Results view — training run history and metrics #50

New issue

Closed

opened 2026-05-01 12:13:13 -07:00 by pyr0ball · 1 comment

pyr0ball commented

2026-05-01 12:13:13 -07:00

Owner

Context

New view showing historical training runs and their metrics, replacing the inline results in the old BenchmarkView finetune section.

Work

TrainResultsView.vue at /train/results
List of completed runs (from job queue + training_info.json files)
Per-run: model name, base model, val_macro_F1, val_accuracy, sample count, training time, timestamp
Run comparison table (side-by-side metrics across runs)
"Register in Fleet" action — adds completed model to cf-orch node catalog
Inline per-run stats: label distribution bar chart, F1 per class

Acceptance

All past training runs visible (reads existing training_info.json files)
"Register in Fleet" triggers cf-orch catalog registration
Metrics display correctly for both classifier and llm-sft run types

## Context New view showing historical training runs and their metrics, replacing the inline results in the old BenchmarkView finetune section. ## Work - `TrainResultsView.vue` at `/train/results` - List of completed runs (from job queue + `training_info.json` files) - Per-run: model name, base model, val_macro_F1, val_accuracy, sample count, training time, timestamp - Run comparison table (side-by-side metrics across runs) - "Register in Fleet" action — adds completed model to cf-orch node catalog - Inline per-run stats: label distribution bar chart, F1 per class ## Acceptance - All past training runs visible (reads existing training_info.json files) - "Register in Fleet" triggers cf-orch catalog registration - Metrics display correctly for both classifier and llm-sft run types