Fix memory leak with CTC training script on Chinese languages (#30358)

* Fix memory leak with CTC training script on Chinese languages

* Fix lint

examples/pytorch/speech-recognition/run_speech_recognition_ctc.py

@@ -28,7 +28,6 @@ from typing import Dict, List, Optional, Union
 
 import datasets
 import evaluate
-import numpy as np
 import torch
 from datasets import DatasetDict, load_dataset
 
@@ -712,10 +711,14 @@ def main():
         logger.info(f"Data preprocessing finished. Files cached at {vectorized_datasets.cache_files}")
         return
 
-    def compute_metrics(pred):
+    # For languages like Chinese with large vocabulary size, we need to discard logits
-        pred_logits = pred.predictions
+    # and only keep the argmax, otherwise we run out of memory during evaluation.
-        pred_ids = np.argmax(pred_logits, axis=-1)
+    def preprocess_logits_for_metrics(logits, labels):
+        pred_ids = torch.argmax(logits, dim=-1)
+        return pred_ids, labels
 
+    def compute_metrics(pred):
+        pred_ids = pred.predictions[0]
         pred.label_ids[pred.label_ids == -100] = tokenizer.pad_token_id
 
         pred_str = tokenizer.batch_decode(pred_ids)
@@ -762,6 +765,7 @@ def main():
         train_dataset=vectorized_datasets["train"] if training_args.do_train else None,
         eval_dataset=vectorized_datasets["eval"] if training_args.do_eval else None,
         tokenizer=processor,
+        preprocess_logits_for_metrics=preprocess_logits_for_metrics,
     )
 
     # 8. Finally, we can start training