minor fixes

app.py

@@ -2,8 +2,7 @@ import gradio as gr
 from PIL import Image
 import torch
 import soundfile as sf
-from transformers import AutoModelForCausalLM, AutoProcessor, GenerationConfig
-from urllib.request import urlopen
+from transformers import AutoModelForCausalLM, AutoProcessor
 import spaces
 
 # Define model path
@@ -28,7 +27,7 @@ prompt_suffix = '<|end|>'
 @spaces.GPU
 def process_input(input_type, file, question):
     if not file or not question:
-        return "Please upload a file and provide a question."
+        return "Please upload a file and provide a question.", None
 
     # Prepare the prompt
     if input_type == "Image":
@@ -36,13 +35,15 @@ def process_input(input_type, file, question):
         # Open image from uploaded file
         image = Image.open(file)
         inputs = processor(text=prompt, images=image, return_tensors='pt').to(model.device)
+        media_output = image  # Return the image for display
     elif input_type == "Audio":
         prompt = f'{user_prompt}<|audio_1|>{question}{prompt_suffix}{assistant_prompt}'
         # Read audio from uploaded file
         audio, samplerate = sf.read(file)
         inputs = processor(text=prompt, audios=[(audio, samplerate)], return_tensors='pt').to(model.device)
+        media_output = (samplerate, audio)  # Return audio in format (samplerate, data) for Gradio
     else:
-        return "Invalid input type selected."
+        return "Invalid input type selected.", None
 
     # Generate response
     with torch.no_grad():
@@ -56,7 +57,7 @@ def process_input(input_type, file, question):
         generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False
     )[0]
 
-    return response
+    return response, media_output
 
 # Gradio interface
 with gr.Blocks(
@@ -71,7 +72,7 @@ with gr.Blocks(
         """
         # Phi-4 Multimodal Demo
         Upload an **image** or **audio** file, ask a question, and get a response from the model!  
-        Built with the `microsoft/Phi-4-multimodal-instruct` model by xAI.
+        Built with the `microsoft/Phi-4-multimodal-instruct` model by Microsoft.
         """
     )
     
@@ -94,12 +95,37 @@ with gr.Blocks(
             submit_btn = gr.Button("Submit", variant="primary")
         
         with gr.Column(scale=2):
-            output_text = gr.Textbox(
-                label="Model Response",
-                placeholder="Response will appear here...",
-                lines=10,
-                interactive=False,
-            )
+            with gr.Tab("Preview"):
+                media_output = gr.Image(label="Uploaded Image", visible=True)  # Default to image
+                gr.Audio(label="Uploaded Audio", visible=False)  # Hidden by default
+            with gr.Tab("Response"):
+                output_text = gr.Textbox(
+                    label="Model Response",
+                    placeholder="Response will appear here...",
+                    lines=10,
+                    interactive=False,
+                )
+
+    # Dynamically update media visibility based on input type
+    def update_media_visibility(input_type):
+        if input_type == "Image":
+            return gr.update(visible=True), gr.update(visible=False)
+        elif input_type == "Audio":
+            return gr.update(visible=False), gr.update(visible=True)
+        return gr.update(visible=False), gr.update(visible=False)
+
+    input_type.change(
+        fn=update_media_visibility,
+        inputs=input_type,
+        outputs=[media_output, demo.blocks["Audio"]]
+    )
+
+    # Connect the submit button
+    submit_btn.click(
+        fn=process_input,
+        inputs=[input_type, file_input, question_input],
+        outputs=[output_text, media_output],
+    )
 
     # Example section
     with gr.Accordion("Examples", open=False):
@@ -110,17 +136,10 @@ with gr.Blocks(
                 ["Audio", "https://upload.wikimedia.org/wikipedia/commons/b/b0/Barbara_Sahakian_BBC_Radio4_The_Life_Scientific_29_May_2012_b01j5j24.flac", "Transcribe the audio to text."],
             ],
             inputs=[input_type, file_input, question_input],
-            outputs=output_text,
+            outputs=[output_text, media_output],
             fn=process_input,
             cache_examples=False,
         )
 
-    # Connect the submit button
-    submit_btn.click(
-        fn=process_input,
-        inputs=[input_type, file_input, question_input],
-        outputs=output_text,
-    )
-
 # Launch the demo
 demo.launch()