Update app.py

app.py

@@ -9,7 +9,6 @@ import torchaudio
 import time
 from transformers import WhisperForAudioClassification, AutoFeatureExtractor
 
-
 # Set page title and favicon
 st.set_page_config(page_title="Audio Visualization", page_icon="🎧")
 
@@ -20,6 +19,7 @@ audio_file = st.file_uploader("Upload Audio file for Assessment", type=["wav", "
 model = WhisperForAudioClassification.from_pretrained("Huma10/Whisper_Stuttered_Speech")
 feature_extractor = AutoFeatureExtractor.from_pretrained("Huma10/Whisper_Stuttered_Speech")
 total_inference_time = 0  # Initialize the total inference time
+
 # Check if an audio file is uploaded
 if audio_file is not None:
     st.audio(audio_file, format="audio/wav")
@@ -39,8 +39,12 @@ if audio_file is not None:
     for clip in audio_clips:
         inputs = feature_extractor(clip.squeeze().numpy(), return_tensors="pt")
         input_features = inputs.input_features
-
         
+        # Pad input features to length 3000
+        if input_features.shape[-1] < 3000:
+            pad_length = 3000 - input_features.shape[-1]
+            input_features = torch.nn.functional.pad(input_features, (0, pad_length))
+
         # Measure inference time
         start_time = time.time()
         # Perform inference
@@ -57,10 +61,11 @@ if audio_file is not None:
         predicted_labels_list.extend(predicted_labels)
     
     st.markdown(f"Total inference time: **{total_inference_time:.4f}** seconds")
+    
     def calculate_percentages(predicted_labels):
-    # Count each type of disfluency
-     disfluency_count = pd.Series(predicted_labels).value_counts(normalize=True)
-     return disfluency_count * 100  # Convert fractions to percentages
+        # Count each type of disfluency
+        disfluency_count = pd.Series(predicted_labels).value_counts(normalize=True)
+        return disfluency_count * 100  # Convert fractions to percentages
 
     def plot_disfluency_percentages(percentages):
         fig, ax = plt.subplots()
@@ -71,7 +76,7 @@ if audio_file is not None:
         plt.xticks(rotation=45)
         return fig
 
-# Streamlit application
+    # Streamlit application
     def main():
         st.title("Speech Profile")
         st.write("This app analyzes the percentage of different types of disfluencies in stuttered speech.")
@@ -83,12 +88,8 @@ if audio_file is not None:
         fig = plot_disfluency_percentages(percentages)
         st.pyplot(fig)
 
-
     main()
 
-    success_check=st.success(' Assessment Completed Successfully!', icon="✅")
+    success_check = st.success('Assessment Completed Successfully!', icon="✅")
     time.sleep(5)
-    success_check=st.empty()
-        
-    
- 
+    success_check = st.empty()