Spaces:
Sleeping
Sleeping
Commit History
11 days ago
-
Report test-set results on an 80/20 split and auto-advance the workflow
- Split samples 80/20 into training and an untouched, stratified test set, as in the BioLM-NET paper. Early stopping uses 1/8 of the training part (10% of all samples), so neither fitting nor checkpoint selection sees the test set. All reported metrics, the confusion matrix and the predictions table come from the test set. - Label results as Test accuracy / Test macro F1 / Test macro AUC. Artifacts saved before this change keep their Validation labels. - Move to the next page automatically after Build (to Train), Train (to Export) and Predict (to Results); add Continue to Predict on Export. Failed steps stay on their page so the error remains visible. - Make scTrioseq2 the default example dataset. - Results page readability: white cards, body font and full-strength text in tables, larger labels, numbers rounded to 3 decimals on screen (downloads keep full precision).1e59649
-
Redesign workbench dashboard shell (#1)
- Redesign workbench dashboard shell (72c0d0ba7a7859042da2633a5c048ec50f59cd87) - Fix dataset loading and add intro cards (121eaa8832a7dfe6e03d01b13909bd13d51934fe) Co-authored-by: Faruk Yildirim <F4RUKYLDRM@users.noreply.huggingface.co>2e6a2a4