def from_protobuf(
cls,
message,
):
return ModelMetrics(
confusion_matrix=ConfusionMatrix.from_protobuf(
message.scoreMatrix),
regression_metrics=RegressionMetrics.from_protobuf(
message.regressionMetrics),
model_type=message.modelType,
)
def test_confusion_matrix_to_protobuf():
targets_1 = ["cat", "dog", "pig"]
predictions_1 = ["cat", "dog", "dog"]
scores_1 = [0.1, 0.2, 0.4]
labels_1 = ["cat", "dog", "pig"]
conf_M_1 = ConfusionMatrix(labels_1)
conf_M_1.add(predictions_1, targets_1, scores_1)
message = conf_M_1.to_protobuf()
expected_1 = [[1, 0, 0], [0, 1, 1], [0, 0, 0]]
new_conf = ConfusionMatrix.from_protobuf(message)
for idx, value in enumerate(new_conf.labels):
assert value == conf_M_1.labels[idx]
for idx, value in enumerate(new_conf.labels):
for jdx, value_2 in enumerate(new_conf.labels):
assert new_conf.confusion_matrix[
idx, jdx].floats.count == expected_1[idx][jdx]
def test_parse_empty_protobuf_should_return_none():
empty_message = ScoreMatrixMessage()
assert ConfusionMatrix.from_protobuf(empty_message) is None
def from_protobuf(
cls,
message,
):
return ModelMetrics(confusion_matrix=ConfusionMatrix.from_protobuf(
message.scoreMatrix))