From dc4db379ea7f2992501651f31e81edf41b7746aa Mon Sep 17 00:00:00 2001 From: Bailey Hirota Date: Thu, 17 Jul 2025 02:01:03 +0900 Subject: [PATCH] PR review suggestions implemented --- egs/multi_ja_en/ASR/local/utils/asr_datamodule.py | 3 ++- egs/multi_ja_en/ASR/local/utils/update_cutset_paths.py | 2 +- 2 files changed, 3 insertions(+), 2 deletions(-) diff --git a/egs/multi_ja_en/ASR/local/utils/asr_datamodule.py b/egs/multi_ja_en/ASR/local/utils/asr_datamodule.py index d82375dd7..417eb3325 100644 --- a/egs/multi_ja_en/ASR/local/utils/asr_datamodule.py +++ b/egs/multi_ja_en/ASR/local/utils/asr_datamodule.py @@ -302,13 +302,14 @@ class MultiDatasetAsrDataModule: logging.info("Loading sampler state dict") train_sampler.load_state_dict(sampler_state_dict) - seed = torch.randint(0, 100000, ()).item() + seed = 42 worker_init_fn = _SeedWorkers(seed) train_dl = DataLoader( train, sampler=train_sampler, batch_size=None, + pin_memory=True, num_workers=self.args.num_workers, persistent_workers=True, worker_init_fn=worker_init_fn, diff --git a/egs/multi_ja_en/ASR/local/utils/update_cutset_paths.py b/egs/multi_ja_en/ASR/local/utils/update_cutset_paths.py index ce8d2805a..af0da4364 100644 --- a/egs/multi_ja_en/ASR/local/utils/update_cutset_paths.py +++ b/egs/multi_ja_en/ASR/local/utils/update_cutset_paths.py @@ -46,7 +46,7 @@ def update_paths(cuts: CutSet, dataset_name: str, old_feature_prefix: str) -> Cu logger.info( f"Updating cut {cut.id}: {original_storage_path} → {new_storage_path}" ) - cut.features.storage_path = str(new_storage_path) + new_storage_path.as_posix() updated_cuts.append(cut) else: logger.warning(f"Skipping update for cut {cut.id}: has no features.")