mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-09 01:52:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

attempt to fix musan paths

Browse Source
This commit is contained in:
Bailey Hirota 2025-07-14 18:33:23 +09:00
parent ea88c55794
commit 259fafab55
1 changed files with 6 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
egs/multi_ja_en/ASR/local/utils/update_cutset_paths.py
View File

@ -37,7 +37,6 @@ def update_paths(cuts: CutSet, dataset_name: str, old_feature_prefix: str = "dat
original_storage_path.parts[0] == old_feature_prefix.split(os.sep)[0] and \ original_storage_path.parts[0] == old_feature_prefix.split(os.sep)[0] and \
original_storage_path.parts[1] == old_feature_prefix.split(os.sep)[1] and \ original_storage_path.parts[1] == old_feature_prefix.split(os.sep)[1] and \
not original_storage_path.parts[2].startswith(dataset_name): # Assumes dataset_name does not start with feats_ not original_storage_path.parts[2].startswith(dataset_name): # Assumes dataset_name does not start with feats_
# This gives us 'feats_train/feats-12.lca' # This gives us 'feats_train/feats-12.lca'
# It's important to be robust to potentially different original prefixes # It's important to be robust to potentially different original prefixes
# So we take the part of the path *after* the `old_feature_prefix` # So we take the part of the path *after* the `old_feature_prefix`
@ -51,12 +50,14 @@ def update_paths(cuts: CutSet, dataset_name: str, old_feature_prefix: str = "dat
continue continue
# Construct the new path: data/manifests/<dataset_name>/feats_train/feats-12.lca # Construct the new path: data/manifests/<dataset_name>/feats_train/feats-12.lca
new_storage_path = Path(old_feature_prefix) / dataset_name / relative_path_from_old_prefix new_storage_path = Path("data/manifests") / dataset_name / relative_path_from_old_prefix
# cut = cut.with_features(cut.features.with_path(str(new_storage_path))) cut = cut.with_features(cut.features.with_path(str(new_storage_path)))
cut.features.storage_path = str(new_storage_path) # cut.features.storage_path = str(new_storage_path)
updated_cuts.append(cut) updated_cuts.append(cut)
else: else:
updated_cuts.append(cut) # No features, or not a path we need to modify updated_cuts.append(cut) # No features, or not a path we need to modify
logger.warning(f"Skipping update for: {original_storage_path}")
return CutSet.from_cuts(updated_cuts) return CutSet.from_cuts(updated_cuts)
if __name__ == "__main__": if __name__ == "__main__":
@ -67,7 +68,6 @@ if __name__ == "__main__":
dataset_manifest_prefixes = { dataset_manifest_prefixes = {
"reazonspeech": "reazonspeech_cuts", "reazonspeech": "reazonspeech_cuts",
"mls_english": "mls_eng_cuts", "mls_english": "mls_eng_cuts",
"musan": "musan_cuts",
} }
splits = ["train", "dev", "test"] splits = ["train", "dev", "test"]
@ -86,7 +86,7 @@ if __name__ == "__main__":
updated_musan_cuts = update_paths( updated_musan_cuts = update_paths(
musan_cuts, musan_cuts,
"musan", "musan",
old_feature_prefix=original_feature_base_path old_feature_prefix="data/fbank"
) )
# Make sure we're overwriting the correct path even if it's a symlink # Make sure we're overwriting the correct path even if it's a symlink
if musan_manifest_path.is_symlink() or musan_manifest_path.exists(): if musan_manifest_path.is_symlink() or musan_manifest_path.exists():