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 pull request #1825 from modelscope/dev_libt
Shi Xian6c467e6f0abf
diff --cc funasr/models/seaco_paraformer/export_meta.pyindex a246a293c,94018a594..d03ebd8ee--- a/funasr/models/seaco_paraformer/export_meta.py+++ b/funasr/models/seaco_paraformer/export_meta.py@@@ -194,10 -192,5 +196,7 @@@ def export_backbone_dynamic_axes(self)"bias_embed": {0: "batch_size", 1: "num_hotwords"},"logits": {0: "batch_size", 1: "logits_length"},"pre_acoustic_embeds": {1: "feats_length1"},+ "us_alphas": {0: "batch_size", 1: "alphas_length"},+ "us_cif_peak": {0: "batch_size", 1: "alphas_length"},}-- def export_backbone_name(self):- return "model.onnx"diff --cc runtime/python/onnxruntime/funasr_onnx/paraformer_bin.pyindex 2cd43a85f,1c58d9382..a60b6d64d--- a/runtime/python/onnxruntime/funasr_onnx/paraformer_bin.py+++ b/runtime/python/onnxruntime/funasr_onnx/paraformer_bin.py@@@ -326,12 -326,8 +326,11 @@@ class ContextualParaformer(Paraformer)def __call__(self, wav_content: Union[str, np.ndarray, List[str]], hotwords: str, **kwargs) -> List:+ # def __call__(+ # self, waveform_list:list, hotwords: str, **kwargs+ # ) -> List:# make hotword listhotwords, hotwords_length = self.proc_hotword(hotwords)- # import pdb; pdb.set_trace()[bias_embed] = self.eb_infer(hotwords, hotwords_length)# index from bias_embedbias_embed = bias_embed.transpose(1, 0, 2)