fin?

app.py

@@ -127,43 +127,40 @@ def run_example(files, text_input=None):
     if not files:
         return "Please upload at least one image for analysis."
     
-    # Add loading message
-    yield "Processing your images... Please wait."
-    
     temp_paths = []
     processed_images = []
     
     try:
         clear_gpu_memory()
         
-        # Validate all files first
-        for file in files:
-            try:
-                validate_file(file)
-            except ValueError as e:
-                return str(e)
-        
         # Process files
         for file in files:
             try:
-                images = process_uploaded_file(file)
-                if images:
+                # For HuggingFace Spaces, we need to handle the file path directly
+                file_path = file.name
+                file_extension = Path(file_path).suffix.lower()
+                
+                if file_extension == '.pdf':
+                    # Convert PDF to images
+                    images = pdf2image.convert_from_path(file_path)
                     processed_images.extend(images)
+                else:
+                    # Handle regular image files
+                    img = Image.open(file_path)
+                    processed_images.append(img)
+                    
             except Exception as e:
                 return f"Error processing file {file.name}: {str(e)}"
         
         if not processed_images:
             return "No valid images were processed. Please check your files."
             
-        # Process images with progress updates
-        yield "Images loaded successfully. Starting analysis..."
-        
         # Save processed images temporarily
         image_paths = []
         for idx, img in enumerate(processed_images):
             try:
                 timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-                filename = f"temp_image_{timestamp}_{idx}.png"
+                filename = f"/tmp/temp_image_{timestamp}_{idx}.png"
                 # Resize image to reduce memory usage
                 img = img.resize((512, 512), Image.Resampling.LANCZOS)
                 img.save(filename)
@@ -174,7 +171,7 @@ def run_example(files, text_input=None):
                 return f"Error saving processed image: {str(e)}"
         
         try:
-            # Process all images in a single batch with reduced memory usage
+            # Process images with the model
             messages = [
                 {
                     "role": "user",
@@ -188,13 +185,12 @@ def run_example(files, text_input=None):
                 }
             ]
             
-            # Process images with the model
             text = processor.apply_chat_template(
                 messages, tokenize=False, add_generation_prompt=True
             )
             image_inputs, video_inputs = process_vision_info(messages)
             
-            with torch.cuda.amp.autocast():  # Use automatic mixed precision
+            with torch.cuda.amp.autocast():
                 inputs = processor(
                     text=[text],
                     images=image_inputs,
@@ -204,12 +200,11 @@ def run_example(files, text_input=None):
                 )
                 inputs = inputs.to("cuda")
                 
-                # Generate response with reduced memory usage
                 generated_ids = model.generate(
                     **inputs,
-                    max_new_tokens=512,  # Reduced from 1024
-                    do_sample=False,     # Deterministic generation
-                    num_beams=1,         # No beam search
+                    max_new_tokens=512,
+                    do_sample=False,
+                    num_beams=1,
                     pad_token_id=processor.tokenizer.pad_token_id,
                     eos_token_id=processor.tokenizer.eos_token_id,
                 )
@@ -223,7 +218,6 @@ def run_example(files, text_input=None):
                     clean_up_tokenization_spaces=False
                 )
                 
-                # Clean up CUDA memory
                 del inputs, generated_ids, generated_ids_trimmed
                 clear_gpu_memory()
                 
@@ -236,7 +230,6 @@ def run_example(files, text_input=None):
         return f"Error processing images: {str(e)}"
     
     finally:
-        # Clean up temporary files and GPU memory
         cleanup_temp_files(temp_paths)
         clear_gpu_memory()
 
@@ -362,13 +355,5 @@ with gr.Blocks(css=css, theme=gr.themes.Soft()) as demo:
             concurrency_limit=10  # Higher limit for clear operation as it's lightweight
         )
 
-# Launch with production configurations
-demo.launch(
-    debug=False,
-    show_error=True,
-    server_name="0.0.0.0",
-    server_port=7860,
-    share=False,
-    max_threads=40,  # Set maximum number of threads
-    enable_queue=True
-)
+# Simplified launch for HuggingFace Spaces
+demo.launch()