雾聪
2023-08-10 ffb05b9ae7eccc47416e9e7fae9dea54d400a245
funasr/datasets/small_datasets/sequence_iter_factory.py
@@ -57,7 +57,7 @@
            data_path_and_name_and_type,
            preprocess=preprocess_fn,
            dest_sample_rate=dest_sample_rate,
            speed_perturb=args.speed_perturb if mode=="train" else None,
            speed_perturb=args.speed_perturb if mode == "train" else None,
        )
        # sampler
@@ -85,7 +85,7 @@
            args.max_update = len(bs_list) * args.max_epoch
            logging.info("Max update: {}".format(args.max_update))
        if args.distributed and mode=="train":
        if args.distributed and mode == "train":
            world_size = torch.distributed.get_world_size()
            rank = torch.distributed.get_rank()
            for batch in batches: