Update app.py with improved Gradio interface and system prompt for AI assistant

app.py

@@ -1,4 +1,5 @@
 import gradio as gr
+import torch
 from transformers import AutoModel, AutoTokenizer
 
 # Load the model
@@ -9,10 +10,13 @@ tokenizer = AutoTokenizer.from_pretrained("openbmb/MiniCPM-V-2", trust_remote_co
 
 model.eval()
 
-# Define the Gradio components
+
+# Image and text inputs for the interface
 image = gr.Image(type="pil", label="Image")
-question = gr.Textbox(value="Using the standard 9x9 sudoku format, solve the sudoku puzzle in the image correctly.", label="Question")
-answer = gr.Textbox(label="Answer", show_label=True, show_copy_button=True)
+question = gr.Textbox(label="Question")
+
+# Output for the interface
+answer = gr.Textbox(label="Predicted answer", show_label=True, show_copy_button=True)
 
 title = "Sudoku Solver by FG"
 description = "Sudoku Solver using MiniCPM-V-2 model by FG. Upload an image of a sudoku puzzle and ask a question to solve it."
@@ -24,7 +28,10 @@ def solve_sudoku(image, question):
         image=image,
         msgs=msgs,
         tokenizer=tokenizer,
-        system_prompt="You are an expert in solving sudoku puzzles. Please solve the sudoku puzzle in the image correctly.",
+        sampling=True,
+        temperature=0.7,
+        stream=True,
+        system_prompt="You are an AI assistant specialized in visual content analysis. Given an image and a related question, analyze the image thoroughly and provide a precise and informative answer based on the visible content. Ensure your response is clear, accurate, and directly addresses the question.",
     )
     return "".join(res)