Update app.py

app.py

@@ -5,26 +5,28 @@ from deepseek_vl.models import VLChatProcessor, MultiModalityCausalLM
 from deepseek_vl.utils.io import load_pil_images
 from io import BytesIO
 from PIL import Image
+import spaces  # Import spaces for ZeroGPU support
 
 # Load the model and processor
 model_path = "deepseek-ai/deepseek-vl-1.3b-chat"
 vl_chat_processor = VLChatProcessor.from_pretrained(model_path)
 tokenizer = vl_chat_processor.tokenizer
 
-# Define the function for image description (CPU version)
-def describe_image(image, user_question="Describe this image in great detail."):
+# Define the function for image description with ZeroGPU support
+@spaces.GPU  # Ensures GPU allocation for this function
+def describe_image(image, user_question="Solve the problem in the image"):
     try:
-        # Convert the PIL Image to a BytesIO object
+        # Convert the PIL Image to a BytesIO object for compatibility
         image_byte_arr = BytesIO()
-        image.save(image_byte_arr, format="PNG")
-        image_byte_arr.seek(0)
+        image.save(image_byte_arr, format="PNG")  # Save image in PNG format
+        image_byte_arr.seek(0)  # Move pointer to the start
 
-        # Define the conversation
+        # Define the conversation, using the user's question
         conversation = [
             {
                 "role": "User",
                 "content": f"<image_placeholder>{user_question}",
-                "images": [image_byte_arr]
+                "images": [image_byte_arr]  # Pass the image byte array instead of an object
             },
             {
                 "role": "Assistant",
@@ -32,35 +34,27 @@ def describe_image(image, user_question="Describe this image in great detail."):
             }
         ]
 
-        # Convert byte array back to PIL image
-        pil_images = [Image.open(BytesIO(image_byte_arr.read()))]
-        image_byte_arr.seek(0)
+        # Convert image byte array back to a PIL image for processing
+        pil_images = [Image.open(BytesIO(image_byte_arr.read()))]  # Convert byte back to PIL Image
+        image_byte_arr.seek(0)  # Reset the byte stream again for reuse
 
-        # Prepare inputs
+        # Load images and prepare the inputs
         prepare_inputs = vl_chat_processor(
             conversations=conversation,
             images=pil_images,
             force_batchify=True
-        )
-
-        # Convert all tensors in prepare_inputs to float32 for CPU compatibility
-        for key in prepare_inputs:
-            if isinstance(prepare_inputs[key], torch.Tensor):
-                prepare_inputs[key] = prepare_inputs[key].to(dtype=torch.float32)
+        ).to('cuda')
 
-        # Load model with CPU and float32 weights
-        vl_gpt = AutoModelForCausalLM.from_pretrained(
-            model_path, 
-            trust_remote_code=True
-        ).float().eval()  # Convert all weights to float32
+        # Load and prepare the model
+        vl_gpt = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True).to(torch.bfloat16).cuda().eval()
 
-        # Generate embeddings with CPU
+        # Generate embeddings from the image input
         inputs_embeds = vl_gpt.prepare_inputs_embeds(**prepare_inputs)
 
-        # Generate response with CPU
+        # Generate the model's response
         outputs = vl_gpt.language_model.generate(
             inputs_embeds=inputs_embeds,
-            attention_mask=prepare_inputs["attention_mask"],
+            attention_mask=prepare_inputs.attention_mask,
             pad_token_id=tokenizer.eos_token_id,
             bos_token_id=tokenizer.bos_token_id,
             eos_token_id=tokenizer.eos_token_id,
@@ -69,36 +63,38 @@ def describe_image(image, user_question="Describe this image in great detail."):
             use_cache=True
         )
 
-        # Decode the response
-        answer = tokenizer.decode(outputs[0].tolist(), skip_special_tokens=True)
+        # Decode the generated tokens into text
+        answer = tokenizer.decode(outputs[0].cpu().tolist(), skip_special_tokens=True)
         return answer
 
     except Exception as e:
+        # Provide detailed error information
         return f"Error: {str(e)}"
 
 # Gradio interface
 def gradio_app():
     with gr.Blocks() as demo:
-        gr.Markdown("# Image Description with DeepSeek VL 1.3b 🐬 (CPU Version)")
+        gr.Markdown("# Image Description with DeepSeek VL 1.3b 🐬\n### Upload an image and ask a question about it.")
 
         with gr.Row():
             image_input = gr.Image(type="pil", label="Upload an Image")
             question_input = gr.Textbox(
                 label="Question (optional)", 
-                placeholder="Ask a question about the image",
+                placeholder="Ask a question about the image (e.g., 'What is happening in this image?')",
                 lines=2
             )
 
         output_text = gr.Textbox(label="Image Description", interactive=False)
+
         submit_btn = gr.Button("Generate Description")
 
         submit_btn.click(
             fn=describe_image,
-            inputs=[image_input, question_input],
+            inputs=[image_input, question_input],  # Pass both image and question as inputs
             outputs=output_text
         )
 
     demo.launch()
 
-# Launch the app
+# Launch the Gradio app
 gradio_app()