improve: update InfoLM class to dynamically set higher_is_better

CHANGELOG.md

@@ -29,6 +29,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Tracker higher is better integration ([#2649](https://github.com/Lightning-AI/torchmetrics/pull/2649))
 
 
+- update `InfoLM` class to dynamically set `higher_is_better` ([#2674](https://github.com/Lightning-AI/torchmetrics/pull/2674))
+
+
 ### Removed
 
 -

src/torchmetrics/text/infolm.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import os
-from typing import Any, Dict, List, Optional, Sequence, Tuple, Union
+from typing import Any, ClassVar, Dict, List, Optional, Sequence, Tuple, Union
 
 import torch
 from torch import Tensor
@@ -111,12 +111,25 @@ class InfoLM(Metric):
     """
 
     is_differentiable = False
-    higher_is_better = True
     preds_input_ids: List[Tensor]
     preds_attention_mask: List[Tensor]
     target_input_ids: List[Tensor]
     target_attention_mask: List[Tensor]
 
+    _information_measure_higher_is_better: ClassVar = {
+        # following values are <0
+        "kl_divergence": True,
+        "alpha_divergence": True,
+        # following values are >0
+        "beta_divergence": False,
+        "ab_divergence": False,
+        "renyi_divergence": False,
+        "l1_distance": False,
+        "l2_distance": False,
+        "l_infinity_distance": False,
+        "fisher_rao_distance": False,
+    }
+
     def __init__(
         self,
         model_name_or_path: Union[str, os.PathLike] = "bert-base-uncased",
@@ -156,6 +169,15 @@ def __init__(
         self.add_state("target_input_ids", [], dist_reduce_fx="cat")
         self.add_state("target_attention_mask", [], dist_reduce_fx="cat")
 
+    @property
+    def higher_is_better(self) -> bool:  # type: ignore[override]
+        """Returns a bool indicating whether a higher value of the information measure is better.
+
+        Done this way as depends on if the information measure is positive or negative.
+
+        """
+        return self._information_measure_higher_is_better[self.information_measure]
+
     def update(self, preds: Union[str, Sequence[str]], target: Union[str, Sequence[str]]) -> None:
         """Update state with predictions and targets."""
         preds_input_ids, preds_attention_mask, target_input_ids, target_attention_mask = _infolm_update(

tests/unittests/text/test_infolm.py

@@ -182,3 +182,18 @@ def test_infolm_differentiability(self, preds, targets, information_measure, idf
             metric_functional=infolm,
             metric_args=metric_args,
         )
+
+    @skip_on_connection_issues()
+    def test_infolm_higher_is_better_property(self, preds, targets, information_measure, idf, alpha, beta):
+        """Test the `higher_is_better` property of the metric."""
+        metric_args = {
+            "model_name_or_path": MODEL_NAME,
+            "information_measure": information_measure,
+            "idf": idf,
+            "alpha": alpha,
+            "beta": beta,
+            "max_length": MAX_LENGTH,
+        }
+
+        metric = InfoLM(**metric_args)
+        assert metric.higher_is_better == metric._information_measure_higher_is_better[information_measure]