update app.py

app.py

@@ -18,33 +18,30 @@ def evaluate_model(model_id, num_questions):
     model = SentenceTransformer(model_id, device=device)
     matryoshka_dimensions = [768, 512, 256, 128, 64]
 
-    # Prepare datasets
+    # Prepare datasets (only load the necessary split and limit to num_questions)
     datasets_info = [
         {
             "name": "Financial",
             "dataset_id": "Omartificial-Intelligence-Space/Arabic-finanical-rag-embedding-dataset",
-            "split": "train",
-            "size": 7000,
+            "split": "train",  # Only train split
             "columns": ("question", "context"),
             "sample_size": num_questions
         },
         {
             "name": "MLQA",
             "dataset_id": "google/xtreme",
-            "subset": "MLQA.ar.ar",
+            "subset": "MLQA.ar.ar",  # Validation split only
             "split": "validation",
-            "size": 500,
             "columns": ("question", "context"),
             "sample_size": num_questions
         },
         {
             "name": "ARCD",
             "dataset_id": "hsseinmz/arcd",
-            "split": "train",
-            "size": None,
+            "split": "train",  # Only train split
             "columns": ("question", "context"),
             "sample_size": num_questions,
-            "last_rows": True  # Take the last n rows
+            "last_rows": True  # Take the last num_questions rows
         }
     ]
     
@@ -58,13 +55,13 @@ def evaluate_model(model_id, num_questions):
         else:
             dataset = load_dataset(dataset_info["dataset_id"], split=dataset_info["split"])
 
-        # Take the last n rows if specified
+        # Limit the number of samples to num_questions (500 max)
         if dataset_info.get("last_rows"):
-            dataset = dataset.select(range(len(dataset) - dataset_info["sample_size"], len(dataset)))
+            dataset = dataset.select(range(len(dataset) - dataset_info["sample_size"], len(dataset)))  # Take last n rows
         else:
-            dataset = dataset.select(range(min(dataset_info["sample_size"], len(dataset))))
+            dataset = dataset.select(range(min(dataset_info["sample_size"], len(dataset))))  # Take first n rows
         
-        # Rename columns
+        # Rename columns to 'anchor' and 'positive'
         dataset = dataset.rename_column(dataset_info["columns"][0], "anchor")
         dataset = dataset.rename_column(dataset_info["columns"][1], "positive")