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
wenetspeech
游雁b37e7b15e02f
diff --git a/examples/aishell/paraformer/utils/text2token.py b/examples/aishell/paraformer/utils/text2token.pyindex 263cb9d1a..a89679f53 100755--- a/examples/aishell/paraformer/utils/text2token.py+++ b/examples/aishell/paraformer/utils/text2token.py@@ -60,6 +60,12 @@ def get_parser():read from SI1279.PHN file -> "sil b r ih sil k s aa r er n aa lsil t er n ih sil t ih v sil" """,)+ parser.add_argument(+ "--text_format",+ default="text",+ type=str,+ help="text, jsonl",+ )return parser@@ -82,6 +88,9 @@ def main():line = f.readline()n = args.ncharwhile line:+ if args.text_format == "jsonl":+ data = json.loads(line.strip())+ line = data["target"]x = line.split()print(" ".join(x[: args.skip_ncols]), end=" ")a = " ".join(x[args.skip_ncols :])diff --git a/funasr/bin/compute_audio_cmvn.py b/funasr/bin/compute_audio_cmvn.pyindex 91bb2ac84..697da94bb 100644--- a/funasr/bin/compute_audio_cmvn.py+++ b/funasr/bin/compute_audio_cmvn.py@@ -52,7 +52,7 @@ def main(**kwargs):frontend=frontend,tokenizer=None,is_training=False,- **kwargs.get("dataset_conf")+ **kwargs.get("dataset_conf"),)# dataloader@@ -68,11 +68,13 @@ def main(**kwargs):dataset_train, collate_fn=dataset_train.collator, **batch_sampler_train)- iter_stop = int(kwargs.get("scale", 1.0) * len(dataloader_train))-total_frames = 0for batch_idx, batch in enumerate(dataloader_train):- if batch_idx >= iter_stop:+ iter_stop = int(kwargs.get("scale", -1.0) * len(dataloader_train))+ log_step = iter_stop // 100+ if batch_idx % log_step == 0:+ print(f"prcessed: {batch_idx}/{iter_stop}")+ if batch_idx >= iter_stop and iter_stop > 0.0:breakfbank = batch["speech"].numpy()[0, :, :]