Save logits with model predictions

[mihow]

No description provided.

[mihow]

Take it away @rhine3!

[mihow]

Code review

Found 1 issue:

1. save_results methods expect 3-tuples but some post_process_batch implementations still return 2-tuples, causing ValueError at runtime. The PR updates Resnet50Classifier.post_process_batch() to return (label, score, logits) 3-tuples and updates both BinaryClassifier.save_results() and SpeciesClassifier.save_results() to unpack 3 values. However, EfficientNetClassifier.post_process_batch() and Resnet50Classifier_Turing.post_process_batch() still return 2-tuples (label, score). Concrete classes that inherit from these will crash:
   • MothNonMothClassifier2022(EfficientNetClassifier, BinaryClassifier) — MRO resolves post_process_batch to EfficientNetClassifier (2-tuple), but save_results to BinaryClassifier (expects 3-tuple)
   • TuringCostaRicaSpeciesClassifier, TuringAnguillaSpeciesClassifier, TuringUKSpeciesClassifier — all inherit post_process_batch from Resnet50Classifier_Turing (2-tuple), but save_results from SpeciesClassifier (expects 3-tuple)

EfficientNetClassifier.post_process_batch returns 2-tuples:

ami-data-companion/trapdata/ml/models/classification.py

Lines 79 to 83 in 66ec695

	scores = predictions.max(axis=1).astype(float)
	result = list(zip(labels, scores))
	logger.debug(f"Post-processing result batch: {result}")
	return result

Resnet50Classifier_Turing.post_process_batch returns 2-tuples:

ami-data-companion/trapdata/ml/models/classification.py

Lines 142 to 146 in 66ec695

	scores = predictions.max(axis=1).astype(float)
	result = list(zip(labels, scores))
	logger.debug(f"Post-processing result batch: {result}")
	return result

BinaryClassifier.save_results unpacks 3-tuples:

ami-data-companion/trapdata/ml/models/classification.py

Lines 266 to 270 in 66ec695

	"model_name": self.name,
	}
	for label, score, _logits in batch_output
	]
	save_classified_objects(self.db_path, object_ids, classified_objects_data)

SpeciesClassifier.save_results unpacks 3-tuples:

ami-data-companion/trapdata/ml/models/classification.py

Lines 327 to 331 in 66ec695

	"in_queue": True,
	}
	for label, top_score, logits in batch_output
	]
	save_classified_objects(self.db_path, object_ids, classified_objects_data)

🤖 Generated with Claude Code

_{- If this code review was useful, please react with 👍. Otherwise, react with 👎.}