feat: training improvements — W&B logging, Optuna HPO, discriminative LR#16
Merged
ziv-lazarov-nagish merged 1 commit intomainfrom Apr 14, 2026
Merged
Conversation
AmitMY
requested changes
Apr 13, 2026
| "wandb", | ||
| "optuna", | ||
| "optuna-integration[pytorch_lightning]", | ||
| "pyyaml", |
Contributor
There was a problem hiding this comment.
not needed for normal install for inference? only for dev, right?
ziv-lazarov-nagish
force-pushed
the
feat/training-improvements
branch
2 times, most recently
from
03c0394 to
7576ce0
Compare
AmitMY
approved these changes
Apr 13, 2026
Enable W&B logging by default with --wandb_entity/--wandb_project. Add --lr_scale_backbone for discriminative fine-tuning (lower LR for backbone, full LR for classification heads). Add Optuna HPO via --optuna <yaml> with dedicated hpo.py module — YAML search space split into architecture/training sections, monitor_metric configurable, W&B callback per trial, pruning for unpromising trials. Refactor training loop into train(overrides, monitor_metric) function.
ziv-lazarov-nagish
force-pushed
the
feat/training-improvements
branch
from
7576ce0 to
cebfb68
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Depends on #15 (
refactor/data-layer).--wandb_entity/--wandb_projectargs and full hyperparameter tracking--lr_scale_backbonefor discriminative fine-tuning (lower LR for CNN + transformer backbone, full LR for classification heads)--optuna <yaml>with a dedicatedhpo.pymodulearchitectureandtrainingsections — architecture params auto-skipped when fine-tuningmonitor_metricconfigurable in the YAMLas_multirun) for per-trial logging, trials named<run_name>-t0,<run_name>-t1, etc. (--run_namerequired for Optuna)train(overrides, monitor_metric)functionwandbfrom dev to main deps, addoptuna,optuna-integration,pyyamlTest plan
ruff check .passes🤖 Generated with Claude Code