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
fix cif_predictor.py bug in mac GPU (#2727)
ylyingaefcf86b2c62
diff --git a/funasr/models/bicif_paraformer/cif_predictor.py b/funasr/models/bicif_paraformer/cif_predictor.pyindex ca98cdc2b..77d4e9659 100644--- a/funasr/models/bicif_paraformer/cif_predictor.py+++ b/funasr/models/bicif_paraformer/cif_predictor.py@@ -62,7 +62,7 @@ def cif(hidden, alphas, threshold):max_label_len = len_labels.max()for b in range(batch_size):fire = fires[b, :]- l = torch.index_select(frames[b, :, :], 0, torch.nonzero(fire >= threshold).squeeze())+ l = torch.index_select(frames[b, :, :], 0, torch.nonzero(fire >= threshold).squeeze(-1))pad_l = torch.zeros([max_label_len - l.size(0), hidden_size], device=hidden.device)list_ls.append(torch.cat([l, pad_l], 0))return torch.stack(list_ls, 0), fires@@ -541,3 +541,4 @@ def cif_wo_hidden_export(alphas, threshold: float):fires = torch.stack(list_fires, 1)return fires+