Update tasks/audio.py

tasks/audio.py

@@ -56,12 +56,8 @@ async def evaluate_audio(request: AudioEvaluationRequest):
     feature_extractor = AutoFeatureExtractor.from_pretrained("facebook/wav2vec2-base")
     
     # Prétraitement en streaming avec fonction explicite
-    test_dataset = test_dataset.map(
-        preprocess_function, 
-        fn_kwargs={"feature_extractor": feature_extractor}, 
-        remove_columns=["audio"]
-    )
-    logging.info(test_dataset)
+    test_dataset = test_dataset.map(preprocess_function, remove_columns="audio", batched=True)
+    
     del feature_extractor  # Libération explicite
     gc.collect()
     
@@ -72,6 +68,7 @@ async def evaluate_audio(request: AudioEvaluationRequest):
     logging.info("Début des prédictions par batch")
     
     for data in test_dataset:
+        logging.info(data)
         with torch.no_grad():
             result = classifier(np.asarray(data["input_values"]))