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
update sensevoice onnx
shixiana62cd7a3fdae
diff --git a/runtime/python/onnxruntime/funasr_onnx/sensevoice_bin.py b/runtime/python/onnxruntime/funasr_onnx/sensevoice_bin.pyindex b3d8dcf6e..bf325c84d 100644--- a/runtime/python/onnxruntime/funasr_onnx/sensevoice_bin.py+++ b/runtime/python/onnxruntime/funasr_onnx/sensevoice_bin.py@@ -3,8 +3,6 @@# Copyright FunASR (https://github.com/FunAudioLLM/SenseVoice). All Rights Reserved.# MIT License (https://opensource.org/licenses/MIT)--import torchimport os.pathimport librosaimport numpy as np@@ -181,12 +179,12 @@ class SenseVoiceSmall:)for b in range(feats.shape[0]):# back to torch.Tensor- if isinstance(ctc_logits, np.ndarray):- ctc_logits = torch.from_numpy(ctc_logits).float()+ # if isinstance(ctc_logits, np.ndarray):+ # ctc_logits = torch.from_numpy(ctc_logits).float()# support batch_size=1 only currentlyx = ctc_logits[b, : encoder_out_lens[b].item(), :]yseq = x.argmax(dim=-1)- yseq = torch.unique_consecutive(yseq, dim=-1)+ yseq = np.unique(yseq)mask = yseq != self.blank_idtoken_int = yseq[mask].tolist()