Update app.py

app.py

@@ -28,6 +28,9 @@ user_prompt = '<|user|>\n'
 assistant_prompt = '<|assistant|>\n'
 prompt_suffix = "<|end|>\n"
 
+title_html = """
+<h2> This space uses the model/microsoft/Phi-3.5-vision-instruct </h2>
+"""
 
 def call_model(raw_image = None, text_input = None):
     prompt = f"{user_prompt}<|image_1|>\n{text_input}{prompt_suffix}{assistant_prompt}"
@@ -63,10 +66,39 @@ def process(raw_image, prompt):
     return memory_usage, model_response, execution_time_min
 
 
-iface = gr.Interface(process,
-                     inputs = [gr.Image(type = 'pil'), gr.Textbox(label = "What do you want to ask?")],
-                     outputs = [gr.Textbox(label = "Memory usage"), gr.Textbox(label = "Model response"),
-                                gr.Textbox(label = "Execution time (min)")])
+with gr.Blocks() as demo:
+    gr.HTML(title_html)
+    gr.Markdown("""
+        NOTES :
+        - The performance of this model is low since it runs on a CPU and a free space, it takes 1min minimum !.
+        - If the input text in not specified the model will describe the image, that will take more time
+    """)
+    with gr.Row():
+        with gr.Column():
+            _raw_image = gr.Image(type = 'pil')
+            user_input = gr.Textbox(label = "What do you want to ask?")
+            submit_btn = gr.Button(value = "Submit")
+        with gr.Column():
+            memory = gr.Textbox(label = "Memory usage")
+            results = gr.Textbox(label = "Model response")
+            exec_time = gr.Textbox(label = "Execution time (min)")
+
+    submit_btn.click(
+        process, inputs = [_raw_image, user_input], outputs = [memory, results, exec_time]
+    )
+
+    gr.Examples(
+        examples=[
+            ["assets/img.jpg", 'after you can split horizontally  the image into 6 rows,  extract all text into JSON format. ignore "Au-dessous de Normal" and "Au-dessus de Normal"'],
+            ["assets/cats.jpg", 'how many cats are here? and what are they doing ?'],
+            ["assets/demo.jpg", 'is it night time ?'],
+        ],
+        inputs=[_raw_image, user_input],
+        outputs=[memory, results, exec_time],
+        fn=process,
+        label="Examples",
+    )
+
 
 if __name__ == '__main__':
-    iface.launch()
+    demo.launch()