removed j

app.py

@@ -7,7 +7,7 @@ from transformers import AutoProcessor, Qwen2_5_VLForConditionalGeneration
 # Model configuration
 MODEL_ID = "numind/NuMarkdown-8B-reasoning"
 
-# Load processor
+# Load processor and model
 processor = AutoProcessor.from_pretrained(
     MODEL_ID,
     trust_remote_code=True,
@@ -15,7 +15,6 @@ processor = AutoProcessor.from_pretrained(
     max_pixels=5000*28*28   
 )
 
-# Load model
 model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     MODEL_ID,
     torch_dtype=torch.bfloat16,
@@ -25,21 +24,19 @@ model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
 )
 
 @spaces.GPU
-def process_image(image):
-    """Process an image using the NuMarkdown-8B-reasoning model."""
+def analyze_image(image):
+    """Analyze an image using the NuMarkdown-8B-reasoning model."""
     if image is None:
         return "Please upload an image.", ""
     
     try:
-        # Convert image to RGB if needed
+        # Convert image to RGB
         img = image.convert("RGB")
         
-        # Prepare messages for the model
+        # Prepare messages
         messages = [{
             "role": "user",
-            "content": [
-                {"type": "image"},
-            ],
+            "content": [{"type": "image"}],
         }]
         
         # Apply chat template
@@ -56,7 +53,7 @@ def process_image(image):
             return_tensors="pt"
         ).to(model.device)
         
-        # Generate output
+        # Generate
         with torch.no_grad():
             model_output = model.generate(
                 **model_input, 
@@ -68,95 +65,101 @@ def process_image(image):
         result = processor.decode(model_output[0])
         
         # Extract reasoning and answer
-        try:
-            reasoning = result.split("<think>")[1].split("</think>")[0]
-        except IndexError:
-            reasoning = "No reasoning found in output."
+        reasoning = "No reasoning found."
+        answer = "No answer found."
         
         try:
-            answer = result.split("<answer>")[1].split("</answer>")[0]
-        except IndexError:
-            answer = "No answer found in output."
+            if "<think>" in result and "</think>" in result:
+                reasoning = result.split("<think>")[1].split("</think>")[0].strip()
+        except:
+            pass
+            
+        try:
+            if "<answer>" in result and "</answer>" in result:
+                answer = result.split("<answer>")[1].split("</answer>")[0].strip()
+        except:
+            pass
         
-        return reasoning.strip(), answer.strip()
+        # If no structured output, return the raw result
+        if reasoning == "No reasoning found." and answer == "No answer found.":
+            return result[:2000] + "..." if len(result) > 2000 else result, result
+        
+        return reasoning, answer
         
     except Exception as e:
-        error_msg = f"Error processing image: {str(e)}"
+        error_msg = f"Error: {str(e)}"
         return error_msg, error_msg
 
-def create_gradio_interface():
-    """Create the Gradio interface"""
+# Create custom CSS
+css = """
+.gradio-container {
+    max-width: 1200px !important;
+}
+.output-text {
+    height: 400px !important;
+}
+"""
+
+# Create the interface using gr.Interface (simpler, more stable)
+with gr.Blocks(css=css, title="NuMarkdown-8B Reasoning Demo") as demo:
     
-    with gr.Blocks(title="NuMarkdown-8B Reasoning Demo") as demo:
-        
-        gr.HTML("""
-        <div style="text-align: center">
-            <h1>🤖 NuMarkdown-8B Reasoning Demo</h1>
-            <p>Upload an image and let the model analyze it with detailed reasoning.</p>
-        </div>
-        """)
-        
-        with gr.Row():
-            with gr.Column():
-                gr.Markdown("### 📸 Upload Your Image")
-                image_input = gr.Image(
-                    type="pil", 
-                    label="Input Image",
-                    height=400
-                )
-                process_btn = gr.Button(
-                    "🔍 Analyze Image", 
-                    variant="primary",
-                    size="lg"
-                )
+    gr.HTML("""
+    <div style="text-align: center; margin-bottom: 20px;">
+        <h1>🤖 NuMarkdown-8B Reasoning Demo</h1>
+        <p style="color: #666;">Upload an image and see the model's detailed reasoning process and final answer.</p>
+    </div>
+    """)
+    
+    with gr.Row():
+        with gr.Column(scale=1):
+            image_input = gr.Image(
+                label="📸 Upload Your Image", 
+                type="pil",
+                height=500
+            )
+            analyze_btn = gr.Button(
+                "🔍 Analyze Image", 
+                variant="primary", 
+                size="lg"
+            )
             
-            with gr.Column():
-                gr.Markdown("### 🧠 Model Reasoning")
-                reasoning_output = gr.Textbox(
-                    label="Thinking Process",
-                    lines=12,
-                    max_lines=15,
-                    placeholder="The model's reasoning will appear here...",
-                    show_copy_button=True
-                )
-                
-                gr.Markdown("### 💡 Final Answer")
-                answer_output = gr.Textbox(
-                    label="Answer",
-                    lines=8,
-                    max_lines=12,
-                    placeholder="The model's answer will appear here...",
-                    show_copy_button=True
-                )
-        
-        # Event handlers
-        process_btn.click(
-            fn=process_image,
-            inputs=[image_input],
-            outputs=[reasoning_output, answer_output]
-        )
-        
-        image_input.upload(
-            fn=process_image,
-            inputs=[image_input],
-            outputs=[reasoning_output, answer_output]
-        )
-        
-        gr.HTML("""
-        <div style="text-align: center; margin-top: 20px; color: #666;">
-            <p><strong>Model:</strong> numind/NuMarkdown-8B-reasoning</p>
-            <p>This demo runs on HuggingFace Zero GPU Spaces for fast inference.</p>
-        </div>
-        """)
+        with gr.Column(scale=1):
+            reasoning_output = gr.Textbox(
+                label="🧠 Model Reasoning",
+                placeholder="The model's step-by-step thinking will appear here...",
+                lines=15,
+                max_lines=20,
+                elem_classes=["output-text"]
+            )
+            
+            answer_output = gr.Textbox(
+                label="💡 Final Answer", 
+                placeholder="The model's final conclusion will appear here...",
+                lines=10,
+                max_lines=15,
+                elem_classes=["output-text"]
+            )
+    
+    gr.HTML("""
+    <div style="text-align: center; margin-top: 20px; padding: 15px; background-color: #f8f9fa; border-radius: 8px;">
+        <p><strong>Model:</strong> numind/NuMarkdown-8B-reasoning</p>
+        <p><strong>Features:</strong> Vision-Language Model with detailed reasoning capabilities</p>
+        <p style="color: #666; font-size: 0.9em;">Powered by HuggingFace Zero GPU Spaces</p>
+    </div>
+    """)
+    
+    # Event handlers
+    analyze_btn.click(
+        fn=analyze_image,
+        inputs=image_input,
+        outputs=[reasoning_output, answer_output]
+    )
     
-    return demo
+    image_input.upload(
+        fn=analyze_image,
+        inputs=image_input,
+        outputs=[reasoning_output, answer_output]
+    )
 
 if __name__ == "__main__":
-    demo = create_gradio_interface()
-    demo.queue(max_size=10).launch(
-        server_name="0.0.0.0",
-        server_port=7860,
-        share=False,
-        debug=True,
-        show_error=True
-    )
+    demo.launch(share=True)