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
ds
游雁a19ee46d2c9d
diff --git a/funasr/train_utils/average_nbest_models.py b/funasr/train_utils/average_nbest_models.pyindex 0f0880425..b10c23145 100644--- a/funasr/train_utils/average_nbest_models.py+++ b/funasr/train_utils/average_nbest_models.py@@ -62,7 +62,8 @@ def average_checkpoints(output_dir: str, last_n: int = 5, **kwargs):# Check if we have any state_dicts to averageif len(state_dicts) < 1:- raise RuntimeError("No checkpoints found for averaging.")+ print("No checkpoints found for averaging.")+ return# Average or sum weightsavg_state_dict = OrderedDict()diff --git a/funasr/train_utils/trainer_ds.py b/funasr/train_utils/trainer_ds.pyindex 8a52746d0..fead9ca08 100644--- a/funasr/train_utils/trainer_ds.py+++ b/funasr/train_utils/trainer_ds.py@@ -168,8 +168,7 @@ class Trainer:"""step_in_epoch = None if step is None else step_in_epochif self.use_deepspeed:- with torch.no_grad():- model.save_checkpoint(save_dir=model_dir, tag=tag, client_state=info_dict)+logging.info(f"Save checkpoint: {epoch}, rank: {self.local_rank}\n")# self.step_or_epoch += 1state = {@@ -273,8 +272,7 @@ class Trainer:elif self.use_fsdp:pass- step_in_epoch = None if step is None else step_in_epoch- if self.rank == 0:+ elif self.rank == 0:logging.info(f"Save checkpoint: {epoch}, rank: {self.local_rank}\n")# self.step_or_epoch += 1state = {@@ -385,7 +383,7 @@ class Trainer:if self.use_deepspeed:ckpt = os.path.join(self.output_dir, "model.pt")- if os.path.isfile(ckpt):+ if os.path.exists(ckpt):_, checkpoint = model_engine.load_checkpoint(self.output_dir, "model.pt")self.saved_ckpts = checkpoint["saved_ckpts"]@@ -712,7 +710,7 @@ class Trainer:"data_split_num": kwargs.get("data_split_num", 1),"log_step": batch_idx + kwargs.get("start_step", 0),"batch_total": batch_idx,- "step_in_epoch": step_in_epoch,+ "step_in_epoch": batch_idx,"lr": 0.0,}