fix: add custom load dataset function for MLSUM tasks (#405)

* fix: add custom load dataset function

* fix: fix linter

* docs: added points

---------

Co-authored-by: Imene Kerboua <imene.kerboua@esker.com>
Co-authored-by: Kenneth Enevoldsen <kennethcenevoldsen@gmail.com>

docs/mmteb/points.md

@@ -2,15 +2,15 @@
 
 | GitHub            | Total points | New dataset | New task | Dataset annotations | (Bug)fixes | Running Models | Review PR |  Paper Writing | Ideation | Coordination |
 |-------------------| ------------ |-------------| -------- | ------------------- | ---------- | -------------- |-----------| -------------- | -------- | ------------- |
-| KennethEnevoldsen |              | 54          |          |                   8 |         18 |                | 32        |                |          |             5 |
+| KennethEnevoldsen |              | 54          |          |                   8 |         18 |                | 33        |                |          |             5 |
 | x-tabdeveloping   |              | 48          |          |                     |            |                |           |                |          |               |
-| imenelydiaker     |              | 88          |          |                     |            |                | 15        |                |          |               |
+| imenelydiaker     |              | 88          |          |                     |          2 |                | 15        |                |          |               |
 | wissam-sib        |              | 88          |          |                     |            |                | 1         |                |          |               |
 | GabrielSequeira   |              | 88          |          |                     |            |                |           |                |          |               |
 | schmarion         |              | 88          |          |                     |            |                |           |                |          |               |
 | MathieuCiancone   |              | 88          |          |                     |            |                |           |                |          |               |
 | Sakshamrzt        |              | 10          |          |                     |            |                | 2         |                |          |               |
-| MartinBernstorff  |              | 2           |          |                     | 7          |                | 3         |                |          |               |
+| MartinBernstorff  |              | 2           |          |                     | 7          |                | 4         |                |          |               |
 | guenthermi        |              | 12          |          |                     |            |                |           |                |          |               |
 | Muennighoff       |              |             |          |                     |            |                | 8         |                |          |               |
 | rasdani           |              | 4           |          |                     |            |                |           |                |          |               |

mteb/tasks/Clustering/fra/MLSUMClusteringP2P.py

@@ -38,6 +38,21 @@ class MLSUMClusteringP2P(AbsTaskClustering):
         avg_character_length=None,
     )
 
+    def load_data(self, **kwargs):
+        """
+        Load dataset from HuggingFace hub and convert it to the standard format.
+        """
+        if self.data_loaded:
+            return
+        self.dataset = datasets.load_dataset(
+            self.metadata.dataset["path"],
+            self.metadata.dataset["name"],
+            split=self.metadata.eval_splits[0],
+            revision=self.metadata.dataset["revision"],
+        )
+        self.dataset_transform()
+        self.data_loaded = True
+
     def create_description(self, example):
         example["text"] = example["title"] + " " + example["text"]
         return example

mteb/tasks/Clustering/fra/MLSUMClusteringS2S.py

@@ -38,6 +38,21 @@ class MLSUMClusteringS2S(AbsTaskClustering):
         avg_character_length=None,
     )
 
+    def load_data(self, **kwargs):
+        """
+        Load dataset from HuggingFace hub and convert it to the standard format.
+        """
+        if self.data_loaded:
+            return
+        self.dataset = datasets.load_dataset(
+            self.metadata.dataset["path"],
+            self.metadata.dataset["name"],
+            split=self.metadata.eval_splits[0],
+            revision=self.metadata.dataset["revision"],
+        )
+        self.dataset_transform()
+        self.data_loaded = True
+
     def dataset_transform(self):
         """
         Convert to standard format