Update app.py

app.py

@@ -94,7 +94,37 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
     # Filter the dataset to include ONLY the target task ID
     # This uses the 'filter' method available on Hugging Face datasets.
     #subset = dataset.filter(lambda example: example['task_id'] in target_task_ids)
-    subset = dataset[20:51]
+    specific_target_ids = [
+    'e1fc63a2-da7a-432f-be78-7c4a95598703',
+    'a1e91b78-d3d8-4675-bb8d-62741b4b68a6',
+    '4fc2f1ae-8625-45b5-ab34-ad4433bc21f8',
+    '8e867cd7-cff9-4e6c-867a-ff5ddc2550be',
+    'ec09fa32-d03f-4bf8-84b0-1f16922c3ae4',
+    '2d83110e-a098-4ebb-9987-066c06fa42d0',
+    '5cfb274c-0207-4aa7-9575-6ac0bd95d9b2',
+    '27d5d136-8563-469e-92bf-fd103c28b57c',
+    'dc28cf18-6431-458b-83ef-64b3ce566c10',
+    '42576abe-0deb-4869-8c63-225c2d75a95a'
+    ]
+    # --- END SPECIFIC TARGET IDS ---
+    
+    # 1. Get the list of Task IDs from the slice (indices 20 to 50)
+    # We must fetch the task_id column data specifically.
+    sliced_ids = dataset.select(range(20, 51))['task_id']
+    
+    # 2. Combine the sliced IDs with the specific IDs into a single set for uniqueness
+    # This ensures we don't accidentally duplicate tasks if some specific IDs are in the slice range.
+    all_unique_target_ids = set(sliced_ids)
+    all_unique_target_ids.update(specific_target_ids)
+    all_unique_target_ids_list = list(all_unique_target_ids)
+    
+    print(f"Total unique tasks to run: {len(all_unique_target_ids_list)}")
+    
+    # 3. Filter the original dataset using the complete list of unique IDs
+    # This replaces the need for complex concatenation.
+    subset = dataset.filter(lambda example: example['task_id'] in all_unique_target_ids_list)
+    subset = subset.to_list() 
+
 
     results_log = []
     answers_payload = []