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
bugfix
游雁c7bfd1fdc597
diff --git a/funasr/models/sense_voice/whisper_lib/decoding.py b/funasr/models/sense_voice/whisper_lib/decoding.pyindex 609d6a607..a468efaa9 100644--- a/funasr/models/sense_voice/whisper_lib/decoding.py+++ b/funasr/models/sense_voice/whisper_lib/decoding.py@@ -63,8 +63,8 @@ def detect_language(else:x = x.to(mel.device)# FIX(funasr): sense vocie- # logits = model.logits(x[:, :-1], mel)[:, -1]- logits = model.logits(x[:, :], mel)[:, -1]+ logits = model.logits(x[:, :-1], mel)[:, -1]+ # logits = model.logits(x[:, :], mel)[:, -1]# collect detected languages; suppress all non-language tokensmask = torch.ones(logits.shape[-1], dtype=torch.bool)