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
Merge branch 'dev_gzf_deepspeed' into main
zhifu gao2175736ab0e2
diff --cc funasr/models/llm_asr/adaptor.pyindex 9b79ed28a,c93988328..93534fef0--- a/funasr/models/llm_asr/adaptor.py+++ b/funasr/models/llm_asr/adaptor.py@@@ -123,6 -125,7 +125,8 @@@ class Transformer(nn.Module)olens = Noneolens = (ilens - 1) // self.k + 1masks = (~make_pad_mask(olens)[:, None, :]).to(x.device)- for layer, block in enumerate(self.blocks):- x, masks = block(x, masks)+++ if self.blocks is not None:+ for layer, block in enumerate(self.blocks):+ x, masks = block(x, masks)return x, olens