Update app.py

app.py

@@ -1,42 +1,43 @@
 import gradio as gr
-from transformers import AutoModel, AutoTokenizer
-from PIL import Image
 import torch
+from PIL import Image
+from transformers import AutoModel, AutoTokenizer
 
-# Load the model and tokenizer
-model_path = "./ggml-model-IQ3_M.gguf"  # Adjust the path to your .gguf model
-model = AutoModel.from_pretrained(model_path, torch_dtype=torch.bfloat16)
-tokenizer = AutoTokenizer.from_pretrained(model_path)
+# Load the model and tokenizer from the local path
+model = AutoModel.from_pretrained('minicpm/models', trust_remote_code=True)
+tokenizer = AutoTokenizer.from_pretrained('minicpm/models', trust_remote_code=True)
 
-# Function to generate responses
-def generate_response(question, image=None):
-    # Tokenize the question input
-    inputs = tokenizer(question, return_tensors="pt")
+# Set the model to evaluation mode
+model.eval()
 
-    # Process the image if provided
-    if image is not None:
-        image = Image.open(image).convert("RGB")
-        # You may need to preprocess the image here (e.g., resizing)
+def predict(image, question):
+    # Preprocess the image
+    image = image.convert('RGB')
 
-    # Generate output from the model
-    with torch.no_grad():
-        output = model(**inputs)  # Adjust based on your model's input requirements
+    # Create the message list
+    msgs = [{'role': 'user', 'content': question}]
 
-    # Extract the response
-    response = output  # Modify this to match your model's output structure
-    return response  # Convert the response to a string if necessary
+    # Generate a response
+    res = model.chat(
+        image=image,
+        msgs=msgs,
+        tokenizer=tokenizer,
+        sampling=True,
+        temperature=0.1
+    )
+    return res
 
 # Create the Gradio interface
-interface = gr.Interface(
-    fn=generate_response,
+iface = gr.Interface(
+    fn=predict,
     inputs=[
-        gr.Textbox(label="Enter your question"),
-        gr.Image(label="Upload an image", type="filepath")
+        gr.inputs.Image(type="pil", label="Upload an Image"),
+        gr.inputs.Textbox(label="Ask a Question")
     ],
     outputs="text",
-    title="MiniCPM Q&A Model",
-    description="Ask questions about the uploaded image."
+    title="Image Question Answering",
+    description="Upload an image and ask a question about it."
 )
 
-# Launch the interface
-interface.launch()
+# Launch the app
+iface.launch()