Skip to content

Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.

fix sense_voice_datasets

北念a98550fdf51a
diff --git a/funasr/datasets/sense_voice_datasets/datasets.py b/funasr/datasets/sense_voice_datasets/datasets.py
index 6b57a9f30..20232cc92 100644
--- a/funasr/datasets/sense_voice_datasets/datasets.py
+++ b/funasr/datasets/sense_voice_datasets/datasets.py
@@ -328,7 +328,7 @@ class SenseVoiceCTCDataset(torch.utils.data.Dataset):
emo_target = item["emo_target"]
event_target = item["event_target"]
text_language = item.get("text_language", "<|zh|>")
- punc_itn_bottom = item.get("with_or_wo_itn", "<|SPECIAL_TOKEN_13|>")
+ punc_itn_bottom = item.get("with_or_wo_itn", "<|woitn|>")
target_ids = self.tokenizer.encode(asr_target, allowed_special="all")
target_ids_len = len(target_ids) # [text]