Datasets:
Tasks:
Automatic Speech Recognition
Formats:
parquet
Languages:
English
Size:
1M - 10M
ArXiv:
License:
Fix streaming mode bug
#4
by
sanchit-gandhi
HF staff
- opened
- peoples_speech.py +1 -1
peoples_speech.py
CHANGED
@@ -157,7 +157,7 @@ class PeoplesSpeech(datasets.GeneratorBasedBuilder):
|
|
157 |
# In non-streaming mode, we extract the archives to have the data locally:
|
158 |
local_extracted_archive_paths = dl_manager.extract(audio_archive_paths) \
|
159 |
if not dl_manager.is_streaming else \
|
160 |
-
{split: [None] * len(audio_archive_paths) for split in splits_to_configs}
|
161 |
|
162 |
manifest_urls = {
|
163 |
split: _MANIFEST_URL.format(split=split, config=config) for split, config in splits_to_configs.items()
|
|
|
157 |
# In non-streaming mode, we extract the archives to have the data locally:
|
158 |
local_extracted_archive_paths = dl_manager.extract(audio_archive_paths) \
|
159 |
if not dl_manager.is_streaming else \
|
160 |
+
{split: [None] * len(audio_archive_paths[split]) for split in splits_to_configs}
|
161 |
|
162 |
manifest_urls = {
|
163 |
split: _MANIFEST_URL.format(split=split, config=config) for split, config in splits_to_configs.items()
|