Update space

app.py

@@ -14,6 +14,7 @@ from email.mime.text import MIMEText
 from huggingface_hub import HfApi
 from tempfile import NamedTemporaryFile
 import shutil
+import tempfile
 
 from stark_qa import load_qa
 from stark_qa.evaluator import Evaluator
@@ -484,6 +485,7 @@ def process_submission(
     code_repo, csv_file, model_description, hardware, paper_link
 ):
     """Process and validate submission"""
+    temp_files = []
     try:
         # Input validation
         if not all([method_name, team_name, dataset, split, contact_email, code_repo, csv_file]):
@@ -513,42 +515,45 @@ def process_submission(
         }
         
         # Save and process files
-        api = HfApi()
         REPO_ID = "snap-stanford/stark-leaderboard"  # Replace with your space name
         HF_TOKEN = os.getenv("HF_TOKEN")
 
         timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
         folder_name = f"{sanitize_name(method_name)}_{sanitize_name(team_name)}"
         
-        # Handle CSV file upload
-        csv_filename = f"predictions_{timestamp}.csv"
-        csv_path_in_repo = f"submissions/{folder_name}/{csv_filename}"
-        
-        # Create a temporary file to store the CSV content
-        with NamedTemporaryFile(mode='wb', delete=False) as tmp_file:
-            # Handle different types of file objects that Gradio might provide
+        temp_csv_path = None
+        if isinstance(csv_file, str):
+            # If it's already a file path, use it directly
+            temp_csv_path = csv_file
+        else:
+            # Create a temporary file with a .csv extension
+            temp_fd, temp_csv_path = tempfile.mkstemp(suffix='.csv')
+            temp_files.append(temp_csv_path)
+            os.close(temp_fd)
+            
+            # Write the content to the temporary file
             if hasattr(csv_file, 'name'):
-                # If it's a path string or has a name attribute
-                with open(csv_file if isinstance(csv_file, str) else csv_file.name, 'rb') as source:
-                    shutil.copyfileobj(source, tmp_file)
+                # If it's a file object with a name attribute
+                shutil.copy2(csv_file.name, temp_csv_path)
             else:
                 # If it's a file-like object
-                csv_file.seek(0)
-                shutil.copyfileobj(csv_file, tmp_file)
-        
-        try:
-            hub_storage.save_to_hub(
-                file_content=tmp_file.name,
-                path_in_repo=csv_path_in_repo,
-                commit_message=f"Add submission: {method_name} by {team_name}"
-            )
-        finally:
-            os.unlink(tmp_file.name)
+                with open(temp_csv_path, 'wb') as temp_file:
+                    if hasattr(csv_file, 'seek'):
+                        csv_file.seek(0)
+                    if hasattr(csv_file, 'read'):
+                        shutil.copyfileobj(csv_file, temp_file)
+                    else:
+                        temp_file.write(csv_file)
+
+        # Verify the CSV file exists and is readable
+        if not os.path.exists(temp_csv_path):
+            raise FileNotFoundError(f"Failed to create temporary CSV file at {temp_csv_path}")
+
         
         # Process evaluation
         # Use the temporary file path for evaluation
         results = compute_metrics(
-            tmp_file.name,  # Use the temporary file path
+            csv_path=temp_csv_path,  # Use the temporary file path
             dataset=dataset.lower(),
             split=split,
             num_workers=4
@@ -557,6 +562,18 @@ def process_submission(
         if isinstance(results, str):
             send_error_notification(meta_data, results)
             return f"Evaluation error: {results}"
+
+        csv_filename = f"predictions_{timestamp}.csv"
+        csv_path_in_repo = f"submissions/{folder_name}/{csv_filename}"
+
+        try:
+            hub_storage.save_to_hub(
+                file_content=temp_csv_path,
+                path_in_repo=csv_path_in_repo,
+                commit_message=f"Add submission: {method_name} by {team_name}"
+            )
+        except Exception as e:
+            raise RuntimeError(f"Failed to save CSV to HuggingFace Hub: {str(e)}")
         
         # Process results (multiply by 100)
         processed_results = {
@@ -574,19 +591,23 @@ def process_submission(
             "submission_date": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
             "csv_path": csv_path_in_repo
         }
+        metadata_fd, temp_metadata_path = tempfile.mkstemp(suffix='.json')
+        temp_files.append(temp_metadata_path)
+        os.close(metadata_fd)
+        
+        with open(temp_metadata_path, 'w') as f:
+            json.dump(submission_data, f, indent=4)
         
         metadata_path = f"submissions/{folder_name}/metadata_{timestamp}.json"
-        with NamedTemporaryFile(mode='w', delete=False) as tmp_file:
-            json.dump(submission_data, tmp_file, indent=4)
             
         try:
             hub_storage.save_to_hub(
-                file_content=tmp_file.name,
+                file_content=temp_metadata_path,
                 path_in_repo=metadata_path,
                 commit_message=f"Add metadata: {method_name} by {team_name}"
             )
-        finally:
-            os.unlink(tmp_file.name)
+        except Exception as e:
+            raise RuntimeError(f"Failed to save metadata to HuggingFace Hub: {str(e)}")
         
         # Send confirmation email and update leaderboard
         send_submission_confirmation(meta_data, processed_results)
@@ -613,6 +634,14 @@ def process_submission(
         # meta_data will always be defined here since we create it at the beginning
         send_error_notification(meta_data, error_message)
         return error_message
+    finally:
+        # Clean up temporary files
+        for temp_file in temp_files:
+            try:
+                if os.path.exists(temp_file):
+                    os.unlink(temp_file)
+            except Exception as e:
+                print(f"Warning: Failed to delete temporary file {temp_file}: {str(e)}")
     
 def filter_by_model_type(df, selected_types):
     if not selected_types: