appready

app.py

@@ -1,6 +1,3 @@
-import subprocess
-subprocess.run('pip install bitsandbytes', shell=True)
-
 import gradio as gr
 from PIL import Image 
 from transformers import AutoModelForCausalLM 
@@ -15,29 +12,19 @@ model = AutoModelForCausalLM.from_pretrained(
     model_id, 
     device_map="cpu", 
     trust_remote_code=True, 
-    torch_dtype=torch.float16,  # Reduced precision
-    load_in_8bit=True,  # 8-bit quantization
+    torch_dtype=torch.float32,
     _attn_implementation="eager"
 )
 processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
 
-PLACEHOLDER = """
-<div style="padding: 30px; text-align: center;">
-   <h1>Phi3 Vision Model</h1>
-   <p>Upload an image and ask a question</p>
-</div>
-"""
-
 @spaces.CPU
 def bot_streaming(message, history):
     try:
-        # Image extraction
         image = (message["files"][-1]["path"] if isinstance(message["files"][-1], dict) else message["files"][-1]) if message["files"] else None
         
         if not image:
             raise ValueError("No image uploaded")
 
-        # Conversation preparation
         conversation = []
         for user, assistant in history:
             conversation.extend([
@@ -47,17 +34,15 @@ def bot_streaming(message, history):
         
         conversation.append({"role": "user", "content": f"<|image_1|>\n{message['text']}"})
         
-        # Prompt and image processing
         prompt = processor.tokenizer.apply_chat_template(conversation, tokenize=False, add_generation_prompt=True)
         image = Image.open(image)
         inputs = processor(prompt, image, return_tensors="pt")
 
-        # Streaming generation with reduced tokens
         streamer = TextIteratorStreamer(processor, skip_special_tokens=True, skip_prompt=True)
         generation_kwargs = dict(
             inputs, 
             streamer=streamer, 
-            max_new_tokens=256,  # Reduced token generation
+            max_new_tokens=256,
             do_sample=False, 
             temperature=0.1, 
             eos_token_id=processor.tokenizer.eos_token_id
@@ -74,20 +59,16 @@ def bot_streaming(message, history):
     except Exception as e:
         yield f"Error: {str(e)}"
 
-# Gradio Interface Configuration
-chatbot = gr.Chatbot(scale=1, placeholder=PLACEHOLDER)
-chat_input = gr.MultimodalTextbox(interactive=True, file_types=["image"], placeholder="Upload image and ask a question")
-
 demo = gr.Blocks()
 with demo:
     gr.ChatInterface(
         fn=bot_streaming,
         title="Phi3 Vision 128K",
         description="Multimodal AI Vision Model",
-        multimodal=True,
-        textbox=chat_input,
-        chatbot=chatbot
+        examples=[
+            {"text": "Describe this image", "files": ["./example.jpg"]},
+        ]
     )
 
-demo.queue(max_size=10)  # Limit queue size
-demo.launch(debug=False, show_error=True)
+demo.queue()
+demo.launch(debug=True)