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
Dev gzf new (#1557)
zhifu gao08a318b72f1a
diff --git a/funasr/datasets/audio_datasets/espnet_samplers.py b/funasr/datasets/audio_datasets/espnet_samplers.pyindex 1524a6a34..c6ed919cd 100644--- a/funasr/datasets/audio_datasets/espnet_samplers.py+++ b/funasr/datasets/audio_datasets/espnet_samplers.py@@ -56,8 +56,8 @@ class EspnetStyleBatchSampler(DistributedSampler):self.shuffle = shuffle and is_trainingself.drop_last = drop_last- # self.total_size = len(self.dataset)- # self.num_samples = int(math.ceil(self.total_size / self.num_replicas))+ self.total_size = len(self.dataset)+ self.num_samples = int(math.ceil(self.total_size / self.num_replicas))self.epoch = 0self.sort_size = sort_size * num_replicasself.max_token_length = kwargs.get("max_token_length", 2048)diff --git a/funasr/datasets/audio_datasets/samplers.py b/funasr/datasets/audio_datasets/samplers.pyindex b4fb84605..108e68a48 100644--- a/funasr/datasets/audio_datasets/samplers.py+++ b/funasr/datasets/audio_datasets/samplers.py@@ -323,7 +323,7 @@ class CustomDistributedBufferDynamicBatchSampler(DistributedSampler):self.shuffle = shuffle and is_trainingself.drop_last = drop_last- # self.total_size = len(self.dataset)+ self.total_size = len(self.dataset)self.num_samples = int(math.ceil(self.total_size / self.num_replicas))self.epoch = 0self.sort_size = sort_size * num_replicas@@ -350,7 +350,7 @@ class CustomDistributedBufferDynamicBatchSampler(DistributedSampler):max_len_in_batch = 0for idx in buffer:original_sample_length = self.dataset.get_source_len(idx)- if original_sample_length > self.max_sample_length:+ if original_sample_length > self.max_token_length:continuesample_length = 1 if self.batch_type == "example" else original_sample_lengthpotential_batch_length = max(max_len_in_batch, sample_length) * (len(batch) + 1)