funasr/bin/compute_audio_cmvn.py
@@ -59,6 +59,7 @@ dataset_conf = kwargs.get("dataset_conf") dataset_conf["batch_type"] = "example" dataset_conf["batch_size"] = 1 dataset_conf["num_workers"] = os.cpu_count() or 32 batch_sampler_train = batch_sampler_class(dataset_train, is_training=False, **dataset_conf) dataloader_train = torch.utils.data.DataLoader(dataset_train, collate_fn=dataset_train.collator, **batch_sampler_train)
