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
1. bug fix:list(mean)和list(var),由于mean和var是numpy,导致写入到文件的格式错误,参考上面的话,大概率是list(mean.tolist()),其实外层list没有必要 (#2437)
天地6e69d784e481
diff --git a/funasr/bin/compute_audio_cmvn.py b/funasr/bin/compute_audio_cmvn.pyindex 79c94c6e6..880a24745 100644--- a/funasr/bin/compute_audio_cmvn.py+++ b/funasr/bin/compute_audio_cmvn.py@@ -88,8 +88,8 @@ def main(**kwargs):total_frames += fbank.shape[0]cmvn_info = {- "mean_stats": list(mean_stats.tolist()),- "var_stats": list(var_stats.tolist()),+ "mean_stats": mean_stats.tolist(),+ "var_stats": var_stats.tolist(),"total_frames": total_frames,}cmvn_file = kwargs.get("cmvn_file", "cmvn.json")@@ -118,11 +118,9 @@ def main(**kwargs):+ str(dims)+ "\n")- mean_str = str(list(mean)).replace(",", "").replace("[", "[ ").replace("]", " ]")- fout.write("<LearnRateCoef> 0 " + mean_str + "\n")+ fout.write("<LearnRateCoef> 0 [ " + " ".join([str(item) for item in mean]) + " ]\n")fout.write("<Rescale> " + str(dims) + " " + str(dims) + "\n")- var_str = str(list(var)).replace(",", "").replace("[", "[ ").replace("]", " ]")- fout.write("<LearnRateCoef> 0 " + var_str + "\n")+ fout.write("<LearnRateCoef> 0 [ " + " ".join([str(item) for item in var]) + " ]\n")fout.write("</Nnet>" + "\n")