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52fee96d71ba
diff --git a/funasr/auto/auto_model.py b/funasr/auto/auto_model.pyindex 5cb2e6e48..ba7dcabaa 100644--- a/funasr/auto/auto_model.py+++ b/funasr/auto/auto_model.py@@ -141,7 +141,7 @@ class AutoModel:kwargs = download_model(**kwargs)set_all_random_seed(kwargs.get("seed", 0))- pdb.set_trace()+device = kwargs.get("device", "cuda")if not torch.cuda.is_available() or kwargs.get("ngpu", 1) == 0:device = "cpu"diff --git a/funasr/models/lcbnet/model.py b/funasr/models/lcbnet/model.pyindex 45b1ee5d1..f8bbf7af1 100644--- a/funasr/models/lcbnet/model.py+++ b/funasr/models/lcbnet/model.py@@ -274,15 +274,12 @@ class LCBNet(nn.Module):ind: int"""with autocast(False):- pdb.set_trace()# Data augmentationif self.specaug is not None and self.training:speech, speech_lengths = self.specaug(speech, speech_lengths)- pdb.set_trace()# Normalization for feature: e.g. Global-CMVN, Utterance-CMVNif self.normalize is not None:speech, speech_lengths = self.normalize(speech, speech_lengths)- pdb.set_trace()# Forward encoder# feats: (Batch, Length, Dim)# -> encoder_out: (Batch, Length2, Dim2)