Random forest #67
Random forest #67
Conversation
There was a problem hiding this comment.
Nice PR @KorenMary! I added a few comments here and there to small changes but it's looking good!
I encountered an error when running by clicking "Generate Labels": sklearn.exceptions.NotFittedError: This RandomForestClassifier instance is not fitted yet. Call 'fit' with appropriate arguments before using this estimator. Please make sure this is addressed!
src/server/test/test_integration.py
Outdated
| if "classifier" in attrs: | ||
| assert(model.classifier.loss<0.4) | ||
| attrs_classifier = model.classifier.__dict__.keys() | ||
| if "loss" in attrs_classifier: | ||
| assert(model.classifier.loss<0.4) | ||
| if "metric" in attrs_classifier: | ||
| assert(model.classifier.metric>0.4) | ||
| if "metric" in attrs: | ||
| assert(model.metric>0.1) | ||
| if "loss" in attrs: |
There was a problem hiding this comment.
This confuses me and isn't very clean. Currently we only have loss and metric for CustomCellposeModel and CellClassifierFCNN. Suggestion to have cleaner approach:
- Always only check model.loss and model.metric (i.e. remove if "classifier" in attrs check).
- In models.py add loss and metric for all classes. For those who have a segmentor and a classifier part: model.loss=(classifier.loss + segmentor.loss)/2 and same for metric. For CustomCellposeModel add self.loss_fn to the init of to the class and uncomment line where we compute it.
src/server/dcp_server/models.py
Outdated
| masks_classes, | ||
| masks_instances, | ||
| noise_intensity = self.train_config["classifier"]["train_data"]["noise_intensity"], | ||
| max_patch_size = self.train_config["classifier"]["train_data"]["patch_size"]) | ||
|
|
||
| if self.classifier_class == "RandomForest": | ||
| patches = create_dataset_for_rf(patches, patch_masks) |
There was a problem hiding this comment.
can we rename patches to something more explicit? here this is now features not patches so it is confusing.
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## main #67 +/- ##
==========================================
+ Coverage 85.61% 86.37% +0.75%
==========================================
Files 20 21 +1
Lines 1314 1431 +117
==========================================
+ Hits 1125 1236 +111
- Misses 189 195 +6 ☔ View full report in Codecov by Sentry. |
… integration test.
No description provided.