Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.
Python asraudiochineseemotion-recognitionfunasrmcp-servermultilingual-asropenai-compatible-apiparaformerpunctuationpytorchreal-time-asrspeaker-diarizationspeech-recognitionspeech-to-textstreaming-asrtranscriptionvllmvoice-activity-detectionwhisper-alternative
test
语帆12d8bd77a668
diff --git a/funasr/models/lcbnet/model.py b/funasr/models/lcbnet/model.pyindex 6ee5342d4..82a1b787d 100644--- a/funasr/models/lcbnet/model.py+++ b/funasr/models/lcbnet/model.py@@ -111,8 +111,8 @@ class LCBNet(nn.Module):)self.blank_id = blank_id- self.sos = sos if sos is not None else vocab_size - 1- self.eos = eos if eos is not None else vocab_size - 1+ self.sos = vocab_size - 1+ self.eos = vocab_size - 1self.vocab_size = vocab_sizeself.ignore_id = ignore_idself.ctc_weight = ctc_weight@@ -375,14 +375,14 @@ class LCBNet(nn.Module):scorers["ngram"] = ngramweights = dict(- decoder=1.0 - kwargs.get("decoding_ctc_weight", 0.5),- ctc=kwargs.get("decoding_ctc_weight", 0.5),+ decoder=1.0 - kwargs.get("decoding_ctc_weight", 0.3),+ ctc=kwargs.get("decoding_ctc_weight", 0.3),lm=kwargs.get("lm_weight", 0.0),ngram=kwargs.get("ngram_weight", 0.0),length_bonus=kwargs.get("penalty", 0.0),)beam_search = BeamSearch(- beam_size=kwargs.get("beam_size", 10),+ beam_size=kwargs.get("beam_size", 20),weights=weights,scorers=scorers,sos=self.sos,