autogluon.multimodal 1.1.2b20241112__tar.gz → 1.1.2b20241114__tar.gz

{autogluon.multimodal-1.1.2b20241112 → autogluon.multimodal-1.1.2b20241114}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.multimodal
-Version: 1.1.2b20241112
+Version: 1.1.2b20241114
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community

{autogluon.multimodal-1.1.2b20241112 → autogluon.multimodal-1.1.2b20241114}/src/autogluon/multimodal/utils/metric.py

@@ -8,7 +8,7 @@ import evaluate
 import numpy as np
 from sklearn.metrics import f1_score
 
-from autogluon.core.metrics import Scorer, get_metric
+from autogluon.core.metrics import Scorer, compute_metric, get_metric
 
 from ..constants import (
     ACCURACY,
@@ -231,15 +231,22 @@ def compute_score(
         return metric.compute(references=metric_data[Y_TRUE], predictions=metric_data[Y_PRED])
 
     metric = get_metric(metric)
-    if metric.name in [ROC_AUC, AVERAGE_PRECISION]:
-        return metric._sign * metric(metric_data[Y_TRUE], metric_data[Y_PRED_PROB][:, pos_label])
-    elif metric.name in [F1]:  # only for binary classification
-        return f1_score(metric_data[Y_TRUE], metric_data[Y_PRED], pos_label=pos_label)
+
+    y = metric_data[Y_TRUE]
+    if metric.needs_proba or metric.needs_threshold:
+        return metric.convert_score_to_original(
+            compute_metric(y=y, y_pred_proba=metric_data[Y_PRED_PROB][:, pos_label], metric=metric, weights=None)
+        )
     else:
-        try:
-            return metric._sign * metric(metric_data[Y_TRUE], metric_data[Y_PRED], y_prob=metric_data[Y_PRED_PROB])
-        except:
-            return metric._sign * metric(metric_data[Y_TRUE], metric_data[Y_PRED])
+        y_pred = metric_data[Y_PRED]
+
+        # TODO: This is a hack. Doesn't support `f1_macro`, `f1_micro`, `f1_weighted`, or custom `f1` metrics with different names.
+        # TODO: Longterm the solution should be to have the input data to this function use the internal representation without the original class names. This way `pos_label` would not need to be specified.
+        if metric.name == F1:  # only for binary classification
+            y = (metric_data[Y_TRUE] == pos_label).astype(int)
+            y_pred = (metric_data[Y_PRED] == pos_label).astype(int)
+
+        return metric.convert_score_to_original(compute_metric(y=y, y_pred=y_pred, metric=metric, weights=None))
 
 
 class RankingMetrics:

{autogluon.multimodal-1.1.2b20241112 → autogluon.multimodal-1.1.2b20241114}/src/autogluon/multimodal/version.py

@@ -1,3 +1,3 @@
 """This is the autogluon version file."""
-__version__ = '1.1.2b20241112'
+__version__ = '1.1.2b20241114'
 __lite__ = False

{autogluon.multimodal-1.1.2b20241112 → autogluon.multimodal-1.1.2b20241114}/src/autogluon.multimodal.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.multimodal
-Version: 1.1.2b20241112
+Version: 1.1.2b20241114
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community

{autogluon.multimodal-1.1.2b20241112 → autogluon.multimodal-1.1.2b20241114}/src/autogluon.multimodal.egg-info/requires.txt

@@ -20,9 +20,9 @@ text-unidecode<1.4,>=1.3
 torchmetrics<1.3.0,>=1.2.0
 nptyping<2.5.0,>=1.4.4
 omegaconf<2.3.0,>=2.1.1
-autogluon.core[raytune]==1.1.2b20241112
-autogluon.features==1.1.2b20241112
-autogluon.common==1.1.2b20241112
+autogluon.core[raytune]==1.1.2b20241114
+autogluon.features==1.1.2b20241114
+autogluon.common==1.1.2b20241114
 pytorch-metric-learning<2.4,>=1.3.0
 nlpaug<1.2.0,>=1.1.10
 nltk<3.9,>=3.4.5