voili
Browse files
run_speech_recognition_ctc.py
CHANGED
@@ -643,7 +643,7 @@ def main():
|
|
643 |
|
644 |
pred.label_ids[pred.label_ids == -100] = tokenizer.pad_token_id
|
645 |
|
646 |
-
pred_str = tokenizer.batch_decode(pred_ids)
|
647 |
# we do not want to group tokens when computing the metrics
|
648 |
label_str = tokenizer.batch_decode(pred.label_ids, group_tokens=False)
|
649 |
|
|
|
643 |
|
644 |
pred.label_ids[pred.label_ids == -100] = tokenizer.pad_token_id
|
645 |
|
646 |
+
pred_str = tokenizer.batch_decode(pred_ids, skip_special_tokens=True)
|
647 |
# we do not want to group tokens when computing the metrics
|
648 |
label_str = tokenizer.batch_decode(pred.label_ids, group_tokens=False)
|
649 |
|