Update tasks/audio.py

tasks/audio.py

@@ -4,6 +4,10 @@ from datasets import load_dataset
 from sklearn.metrics import accuracy_score
 import random
 import os
+import librosa
+import joblib
+import numpy as np
+import lightgbm
 
 from .utils.evaluation import AudioEvaluationRequest
 from .utils.emissions import tracker, clean_emissions_data, get_space_info
@@ -43,6 +47,17 @@ async def evaluate_audio(request: AudioEvaluationRequest):
     # Split dataset
     train_test = dataset["train"]
     test_dataset = dataset["test"]
+
+    def preprocess_data(row, sr):
+        new_row = librosa.resample(row['audio']['array'], orig_sr=row['audio']['sampling_rate'], target_sr=sr)
+        new_row = np.pad(new_row, (0, 3 * sr - len(new_row)), 'constant')
+        new_row = librosa.feature.mfcc(y=new_row, sr=sr, n_mfcc=10)
+        return new_row.flatten()
+
+    test_list_mfcc = np.vstack([preprocess_data(row, 12000) for row in test_dataset])
+
+    model_filename = "lightgbm_10_mfcc.pkl"
+    clf = joblib.load(model_filename)
     
     # Start tracking emissions
     tracker.start()
@@ -55,7 +70,7 @@ async def evaluate_audio(request: AudioEvaluationRequest):
     
     # Make random predictions (placeholder for actual model inference)
     true_labels = test_dataset["label"]
-    predictions = [random.randint(0, 1) for _ in range(len(true_labels))]
+    predictions = clf.predict(test_list_mfcc)
     
     #--------------------------------------------------------------------------------------------
     # YOUR MODEL INFERENCE STOPS HERE