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(spk): 修复 speaker embedding 集群后的重新排序问题
msgka8591060d388
diff --git a/funasr/auto/auto_model.py b/funasr/auto/auto_model.pyindex 96c642e52..f5cbe01ff 100644--- a/funasr/auto/auto_model.py+++ b/funasr/auto/auto_model.py@@ -549,8 +549,41 @@ class AutoModel:# speaker embedding cluster after resortedif self.spk_model is not None and kwargs.get("return_spk_res", True):- if raw_text is None:- logging.error("Missing punc_model, which is required by spk_model.")+ # 1. 先检查时间戳+ has_timestamp = (+ hasattr(self.model, "internal_punc") or+ self.punc_model is not None or+ "timestamp" in result+ )++ if not has_timestamp:+ logging.error("Need timestamp support...")+ return results_ret_list++ # 2. 初始化 punc_res+ punc_res = None++ # 3. 根据不同情况设置 punc_res+ if hasattr(self.model, "internal_punc"):+ punc_res = [{+ "text": result["text"],+ "punc_array": result.get("punc_array", []),+ "timestamp": result.get("timestamp", [])+ }]+ elif self.punc_model is not None:+ punc_res = self.inference(+ result["text"],+ model=self.punc_model,+ kwargs=self.punc_kwargs,+ **cfg+ )+ else:+ # 如果只有时间戳,创建一个基本的 punc_res+ punc_res = [{+ "text": result["text"],+ "punc_array": [], # 空的标点数组+ "timestamp": result["timestamp"]+ }]all_segments = sorted(all_segments, key=lambda x: x[0])spk_embedding = result["spk_embedding"]labels = self.cb_model(