actions-ml-cicd
actions-ml-cicd copied to clipboard
Update train.py
- Increased the size of embeddings from 50 to 85
- Hypothesis: this makes the models better
/run-full-test
ML Workflow For SHA 52872721ab6090ea05aa32ca44d11e935d5cfa4f has been instantiated.
The following Docker images were built and tagged with the SHA:
Check run Argo-Workflow created, with status pending completion of the workflow.
Model Evaluation Results
| Category | Run ID | SHA | Train Loss | Val Loss | Acc | Val Acc | Runtime |
|---|---|---|---|---|---|---|---|
| candidate | ddscgocn | 52872721ab6090ea05aa32ca44d11e935d5cfa4f | 0.366 | 0.534 | 0.862 | 0.796 | 542.478 |
| baseline | d2mg9r7l | 0fbe4ae56ac355e3b4018d22b95fadfd5fd33349 | 0.392 | 0.527 | 0.851 | 0.798 | 577.173 |
Results queried from W&B project github/predict-issue-labels
/deploy ddscgocn