Create app.py

app.py

@@ -0,0 +1,87 @@
+import gradio as gr
+import librosa
+import numpy as np
+from transformers import Wav2Vec2ForClassification, Wav2Vec2Processor
+import os
+
+
+model_id = "AescF/hubert-base-ls960-finetuned-common_language"
+processor = Wav2Vec2Processor.from_pretrained(model_id)
+model = Wav2Vec2ForClassification.from_pretrained(model_id)
+language_classes = {
+    0: "Arabic",
+    1: "Basque",
+    2: "Breton",
+    3: "Catalan",
+    4: "Chinese_China",
+    5: "Chinese_Hongkong",
+    6: "Chinese_Taiwan",
+    7: "Chuvash",
+    8: "Czech",
+    9: "Dhivehi",
+    10: "Dutch",
+    11: "English",
+    12: "Esperanto",
+    13: "Estonian",
+    14: "French",
+    15: "Frisian",
+    16: "Georgian",
+    17: "German",
+    18: "Greek",
+    19: "Hakha_Chin",
+    20: "Indonesian",
+    21: "Interlingua",
+    22: "Italian",
+    23: "Japanese",
+    24: "Kabyle",
+    25: "Kinyarwanda",
+    26: "Kyrgyz",
+    27: "Latvian",
+    28: "Maltese",
+    29: "Mongolian",
+    30: "Persian",
+    31: "Polish",
+    32: "Portuguese",
+    33: "Romanian",
+    34: "Romansh_Sursilvan",
+    35: "Russian",
+    36: "Sakha",
+    37: "Slovenian",
+    38: "Spanish",
+    39: "Swedish",
+    40: "Tamil",
+    41: "Tatar",
+    42: "Turkish",
+    43: "Ukranian",
+    44: "Welsh"
+}
+
+
+def predict_language(audio):
+    # Read audio file
+    audio_input, sr = librosa.load(audio, sr=16000)
+    
+    # Convert to suitable format
+    input_values = processor(audio_input, return_tensors="pt", padding=True).input_values
+    
+    # Make prediction
+    with torch.no_grad():
+        logits = model(input_values).logits
+    
+    # Compute probabilities
+    probabilities = torch.softmax(logits, dim=1)
+    
+    # Retrieve label
+    predicted_language_idx = torch.argmax(probabilities[0]).item()
+    
+    return {language_classes[predicted_language_idx]: float(probabilities[0][predicted_language_idx])}
+
+iface = gr.Interface(
+    predict_language, 
+    inputs=gr.inputs.Audio(type="filepath", label="Upload Language Audio file"),
+    outputs=gr.outputs.Label(),
+    title="Language Classifier",
+    live=True
+)
+script_dir = os.path.abspath(os.path.join(os.path.abspath(''), os.pardir))
+iface.launch()