Update app.py

app.py

@@ -1,30 +1,50 @@
-from transformers import pipeline,AutoConfig,AutoModelForAudioClassification,AutoFeatureExtractor
+import gradio as gr
+import torchaudio
+from transformers import pipeline
+from datasets import load_dataset, Audio
 
-model_id = "Ahmed107/whisper-tiny-finetuned-eos"
+# Load your model
+classifier = pipeline("audio-classification", model="Ahmed107/whisper-tiny-finetuned-eos")
 
-# define mappings as dictionaries
-id2label = {"0": "NOT_EOS", "1": "EOS"}
-label2id = {"NOT_EOS": "0", "EOS": "1"}
+# Function to resample audio to 16kHz
+def resample_audio(audio_file, target_sampling_rate=16000):
+    waveform, original_sample_rate = torchaudio.load(audio_file)
+    if original_sample_rate != target_sampling_rate:
+        resampler = torchaudio.transforms.Resample(orig_freq=original_sample_rate, new_freq=target_sampling_rate)
+        waveform = resampler(waveform)
+    return waveform.squeeze().numpy(), target_sampling_rate
 
-# define config
-config = AutoConfig.from_pretrained(model_id, label2id=label2id, id2label=id2label)
-model = AutoModelForAudioClassification.from_pretrained(model_id, config = config)
-feature_extractor = AutoFeatureExtractor.from_pretrained(
-    model_id,
-)
-pipe = pipeline("audio-classification", model=model,feature_extractor=feature_extractor)
+# Define the prediction function
+def classify_audio(audio_file):
+    # Resample the audio to 16kHz
+    resampled_audio, _ = resample_audio(audio_file)
+    
+    # Classify the audio
+    prediction = classifier(resampled_audio)
+    
+    # Return predictions as a dictionary
+    return {entry['label']: entry['score'] for entry in prediction}
 
-def classify_audio(filepath):
-    preds = pipe(filepath)
-    print(preds)
-    outputs = {}
-    for p in preds:
-        outputs[p["label"]] = p["score"]
-    return outputs
+# Define Gradio interface
+def demo():
+    with gr.Blocks(theme=gr.themes.Soft()) as demo:
+        gr.Markdown("## Eos")
+        
+        # Input Audio
+        with gr.Row():
+            audio_input = gr.Audio(type="filepath", label="Input Audio")
+        
+        # Output Labels
+        with gr.Row():
+            label_output = gr.Label(label="Prediction")
+        
+        # Predict Button
+        classify_btn = gr.Button("Classify")
 
-import gradio as gr
+        # Define the interaction
+        classify_btn.click(fn=classify_audio, inputs=audio_input, outputs=label_output)
+        
+    return demo
 
-demo = gr.Interface(
-    fn=classify_audio, inputs=gr.Audio(type="filepath"), outputs=gr.Label()
-)
-demo.launch(debug=True)
+# Launch the demo
+demo().launch()