Update app.py

app.py

@@ -1,55 +1,12 @@
 import gradio as gr
-import torch
-import torchaudio
-import json
-import os
+from transformers import pipeline
 
-# Import your model architecture
-from model import AudioLanguageClassifier, AudioLanguageClassifierConfig, AudioFeatureExtractor
-
-MODEL_DIR = "."
-
-# Load config and mappings
-with open(os.path.join(MODEL_DIR, "config.json")) as f:
-    config_dict = json.load(f)
-with open(os.path.join(MODEL_DIR, "language_mappings.json")) as f:
-    mappings = json.load(f)
-id_to_language = {int(k): v for k, v in mappings["id_to_language"].items()}
-
-config = AudioLanguageClassifierConfig(**config_dict)
-model = AudioLanguageClassifier(config)
-model.load_state_dict(torch.load(os.path.join(MODEL_DIR, "model.pt"), map_location="cpu"))
-model.eval()
-
-feature_extractor = AudioFeatureExtractor(config)
-max_length = 256  # Or whatever you used in training
+classifier = pipeline("audio-classification", model="hriteshMaikap/languageClassifier")
 
 def predict_language(audio):
-    waveform, sample_rate = torchaudio.load(audio)
-    # Resample and mono
-    if sample_rate != 16000:
-        waveform = torchaudio.transforms.Resample(sample_rate, 16000)(waveform)
-    if waveform.shape[0] > 1:
-        waveform = torch.mean(waveform, dim=0, keepdim=True)
-    features = feature_extractor(waveform)
-    input_values = features["input_values"]
-    _, height, width = input_values.shape
-    # Pad/truncate
-    if width < max_length:
-        padding = torch.zeros(1, height, max_length - width)
-        input_values = torch.cat([input_values, padding], dim=2)
-    elif width > max_length:
-        input_values = input_values[:, :, :max_length]
-    with torch.no_grad():
-        outputs = model(input_values=input_values)
-        logits = outputs["logits"]
-        probs = torch.nn.functional.softmax(logits, dim=1)[0]
-        top_probs, top_ids = torch.topk(probs, 3)
-        results = []
-        for prob, pred_id in zip(top_probs, top_ids):
-            lang = id_to_language[pred_id.item()]
-            results.append(f"{lang}: {prob.item():.2f}")
-    return "\n".join(results)
+    out = classifier(audio)
+    # out is a list of dicts: [{'label': 'Hindi', 'score': 0.98}, ...]
+    return "\n".join([f"{res['label']}: {res['score']:.2f}" for res in out])
 
 demo = gr.Interface(
     fn=predict_language,