Update app.py

app.py

@@ -1,50 +1,57 @@
 import gradio as gr
 import torchaudio
 from transformers import pipeline
-from datasets import load_dataset, Audio
 
 # Load your model
 classifier = pipeline("audio-classification", model="Ahmed107/Hamsa-Conversational-v1.0-mulaw-eos-v3-mulaw")
 
-# Function to resample audio to 16kHz
+# Function to resample audio to 16kHz and convert to mono if needed
 def resample_audio(audio_file, target_sampling_rate=16000):
     waveform, original_sample_rate = torchaudio.load(audio_file)
+    
+    # Resample if necessary
     if original_sample_rate != target_sampling_rate:
         resampler = torchaudio.transforms.Resample(orig_freq=original_sample_rate, new_freq=target_sampling_rate)
         waveform = resampler(waveform)
+    
+    # Convert stereo to mono by averaging channels (if needed)
+    if waveform.shape[0] > 1:
+        waveform = waveform.mean(dim=0, keepdim=True)
+    
     return waveform.squeeze().numpy(), target_sampling_rate
 
 # Define the prediction function
 def classify_audio(audio_file):
-    # Resample the audio to 16kHz
-    resampled_audio, _ = resample_audio(audio_file)
+    # Resample the audio to 16kHz and handle channels
+    resampled_audio, sampling_rate = resample_audio(audio_file)
     
-    # Classify the audio
-    prediction = classifier(resampled_audio)
+    # Pass both the array and sampling_rate to the classifier
+    input_audio = {"array": resampled_audio, "sampling_rate": sampling_rate}
+    prediction = classifier(input_audio)
     
-    # Return predictions as a dictionary
+    # Return predictions as a dictionary with labels and their scores
     return {entry['label']: entry['score'] for entry in prediction}
 
 # Define Gradio interface
 def demo():
     with gr.Blocks(theme=gr.themes.Soft()) as demo:
-        gr.Markdown("## Eos")
+        gr.Markdown("## Eos Audio Classification")
         
-        # Input Audio
+        # Input Audio component
         with gr.Row():
             audio_input = gr.Audio(type="filepath", label="Input Audio")
         
-        # Output Labels
+        # Output Labels component
         with gr.Row():
             label_output = gr.Label(label="Prediction")
         
         # Predict Button
         classify_btn = gr.Button("Classify")
-
-        # Define the interaction
+        
+        # Set the button click action
         classify_btn.click(fn=classify_audio, inputs=audio_input, outputs=label_output)
         
     return demo
 
-# Launch the demo
-demo().launch()
+# Launch the Gradio demo
+demo().launch()