Problem Loading the Dataset
#2
by
MauroVazquez
- opened
I'm encountering two separate issues when trying to load the dataset projecte-aina/corts_valencianes_asr_a from Hugging Face using the datasets library.
Attempt 1:
corts_val = load_dataset(
"projecte-aina/corts_valencianes_asr_a",
None,
split="clean_train_short",
token=HF_TOKEN,
)
as soon as it finishes downloading the files, then it starts to print:
Extraction of ../d/08de2aad9b79cfcbbe1d83c925199c08.wav is blocked (illegal path)
Extraction of ../d/08d2ee0dc5546d8648016fdbf9ae905.wav is blocked (illegal path)
Extraction of ../d/08da668f1a0627951ed1711cbe26c36a.wav is blocked (illegal path)
Extraction of ../d/08d003e7881da129f1b7a93ae7634a3b.wav is blocked (illegal path)
....
I looked into the hub folder inside huggingface_cache and it appears to have the content correctly laid out. It weighs 80 GB.
Attempt 2:
Just in case, I deleted the second argument (name) and retried loading:
corts_val = load_dataset(
"projecte-aina/corts_valencianes_asr_a",
split="clean_train_short", #[:20%] Not available on streaming dataset
token=HF_TOKEN,
)
and it raised the following error:
---------------------------------------------------------------------------
IsADirectoryError Traceback (most recent call last)
Cell In[6], line 1
----> 1 corts_val = load_dataset(
2 "projecte-aina/corts_valencianes_asr_a",
3 split="clean_train_short",
4 token=HF_TOKEN,
5 )
File ~/.local/lib/python3.10/site-packages/datasets/load.py:2083, in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, verification_mode, keep_in_memory, save_infos, revision, token, streaming, num_proc, storage_options, trust_remote_code, **config_kwargs)
2080 return builder_instance.as_streaming_dataset(split=split)
2082 # Download and prepare data
-> 2083 builder_instance.download_and_prepare(
2084 download_config=download_config,
2085 download_mode=download_mode,
2086 verification_mode=verification_mode,
2087 num_proc=num_proc,
2088 storage_options=storage_options,
2089 )
2091 # Build dataset for splits
2092 keep_in_memory = (
2093 keep_in_memory if keep_in_memory is not None else is_small_dataset(builder_instance.info.dataset_size)
2094 )
File ~/.local/lib/python3.10/site-packages/datasets/builder.py:925, in DatasetBuilder.download_and_prepare(self, output_dir, download_config, download_mode, verification_mode, dl_manager, base_path, file_format, max_shard_size, num_proc, storage_options, **download_and_prepare_kwargs)
923 if num_proc is not None:
924 prepare_split_kwargs["num_proc"] = num_proc
--> 925 self._download_and_prepare(
926 dl_manager=dl_manager,
927 verification_mode=verification_mode,
928 **prepare_split_kwargs,
929 **download_and_prepare_kwargs,
930 )
931 # Sync info
932 self.info.dataset_size = sum(split.num_bytes for split in self.info.splits.values())
File ~/.local/lib/python3.10/site-packages/datasets/builder.py:1649, in GeneratorBasedBuilder._download_and_prepare(self, dl_manager, verification_mode, **prepare_splits_kwargs)
1648 def _download_and_prepare(self, dl_manager, verification_mode, **prepare_splits_kwargs):
-> 1649 super()._download_and_prepare(
1650 dl_manager,
1651 verification_mode,
1652 check_duplicate_keys=verification_mode == VerificationMode.BASIC_CHECKS
1653 or verification_mode == VerificationMode.ALL_CHECKS,
1654 **prepare_splits_kwargs,
1655 )
File ~/.local/lib/python3.10/site-packages/datasets/builder.py:979, in DatasetBuilder._download_and_prepare(self, dl_manager, verification_mode, **prepare_split_kwargs)
977 split_dict = SplitDict(dataset_name=self.dataset_name)
978 split_generators_kwargs = self._make_split_generators_kwargs(prepare_split_kwargs)
--> 979 split_generators = self._split_generators(dl_manager, **split_generators_kwargs)
981 # Checksums verification
982 if verification_mode == VerificationMode.ALL_CHECKS and dl_manager.record_checksums:
File /opt/huggingface_cache/modules/datasets_modules/datasets/projecte-aina--corts_valencianes_asr_a/88e5386f4b8a43e5aeac7fb2008ab02bb8468c55c7299b58eb5f1d54c378c9c3/corts_valencianes_asr_a.py:183, in CortsValencianesASR._split_generators(self, dl_manager)
179 audio_paths = dl_manager.download(hash_tar_files)
181 splits=["clean_train_short","clean_test_short","clean_dev_short","other_train_short","other_test_short","other_dev_short","clean_train_long","clean_test_long","clean_dev_long","other_train_long","other_test_long","other_dev_long"]
182 local_extracted_audio_paths = (
--> 183 dl_manager.extract(audio_paths) if not dl_manager.is_streaming else
184 {
185 split:[None] * len(audio_paths[split]) for split in splits
186 }
187 )
189 return [
190 datasets.SplitGenerator(
191 name="clean_train_short",
(...)
285 ),
286 ]
File ~/.local/lib/python3.10/site-packages/datasets/download/download_manager.py:299, in DownloadManager.extract(self, path_or_paths)
297 download_config.extract_compressed_file = True
298 extract_func = partial(self._download_single, download_config=download_config)
--> 299 extracted_paths = map_nested(
300 extract_func,
301 path_or_paths,
302 num_proc=download_config.num_proc,
303 desc="Extracting data files",
304 )
305 path_or_paths = NestedDataStructure(path_or_paths)
306 extracted_paths = NestedDataStructure(extracted_paths)
File ~/.local/lib/python3.10/site-packages/datasets/utils/py_utils.py:496, in map_nested(function, data_struct, dict_only, map_list, map_tuple, map_numpy, num_proc, parallel_min_length, batched, batch_size, types, disable_tqdm, desc)
494 num_proc = 1
495 if any(isinstance(v, types) and len(v) > len(iterable) for v in iterable):
--> 496 mapped = [
497 map_nested(
498 function=function,
499 data_struct=obj,
500 num_proc=num_proc,
501 parallel_min_length=parallel_min_length,
502 batched=batched,
503 batch_size=batch_size,
504 types=types,
505 )
506 for obj in iterable
507 ]
508 elif num_proc != -1 and num_proc <= 1 or len(iterable) < parallel_min_length:
509 if batched:
File ~/.local/lib/python3.10/site-packages/datasets/utils/py_utils.py:497, in <listcomp>(.0)
494 num_proc = 1
495 if any(isinstance(v, types) and len(v) > len(iterable) for v in iterable):
496 mapped = [
--> 497 map_nested(
498 function=function,
499 data_struct=obj,
500 num_proc=num_proc,
501 parallel_min_length=parallel_min_length,
502 batched=batched,
503 batch_size=batch_size,
504 types=types,
505 )
506 for obj in iterable
507 ]
508 elif num_proc != -1 and num_proc <= 1 or len(iterable) < parallel_min_length:
509 if batched:
File ~/.local/lib/python3.10/site-packages/datasets/utils/py_utils.py:513, in map_nested(function, data_struct, dict_only, map_list, map_tuple, map_numpy, num_proc, parallel_min_length, batched, batch_size, types, disable_tqdm, desc)
511 batch_size = max(len(iterable) // num_proc + int(len(iterable) % num_proc > 0), 1)
512 iterable = list(iter_batched(iterable, batch_size))
--> 513 mapped = [
514 _single_map_nested((function, obj, batched, batch_size, types, None, True, None))
515 for obj in hf_tqdm(iterable, disable=disable_tqdm, desc=desc)
516 ]
517 if batched:
518 mapped = [mapped_item for mapped_batch in mapped for mapped_item in mapped_batch]
File ~/.local/lib/python3.10/site-packages/datasets/utils/py_utils.py:514, in <listcomp>(.0)
511 batch_size = max(len(iterable) // num_proc + int(len(iterable) % num_proc > 0), 1)
512 iterable = list(iter_batched(iterable, batch_size))
513 mapped = [
--> 514 _single_map_nested((function, obj, batched, batch_size, types, None, True, None))
515 for obj in hf_tqdm(iterable, disable=disable_tqdm, desc=desc)
516 ]
517 if batched:
518 mapped = [mapped_item for mapped_batch in mapped for mapped_item in mapped_batch]
File ~/.local/lib/python3.10/site-packages/datasets/utils/py_utils.py:375, in _single_map_nested(args)
373 return function([data_struct])[0]
374 else:
--> 375 return function(data_struct)
376 if (
377 batched
378 and not isinstance(data_struct, dict)
379 and isinstance(data_struct, types)
380 and all(not isinstance(v, (dict, types)) for v in data_struct)
381 ):
382 return [mapped_item for batch in iter_batched(data_struct, batch_size) for mapped_item in function(batch)]
File ~/.local/lib/python3.10/site-packages/datasets/download/download_manager.py:229, in DownloadManager._download_single(self, url_or_filename, download_config)
226 if is_relative_path(url_or_filename):
227 # append the relative path to the base_path
228 url_or_filename = url_or_path_join(self._base_path, url_or_filename)
--> 229 out = cached_path(url_or_filename, download_config=download_config)
230 out = tracked_str(out)
231 out.set_origin(url_or_filename)
File ~/.local/lib/python3.10/site-packages/datasets/utils/file_utils.py:252, in cached_path(url_or_filename, download_config, **download_kwargs)
249 return output_path
251 # Eager extraction
--> 252 output_path = ExtractManager(cache_dir=download_config.cache_dir).extract(
253 output_path, force_extract=download_config.force_extract
254 )
255 return relative_to_absolute_path(output_path)
File ~/.local/lib/python3.10/site-packages/datasets/utils/extract.py:43, in ExtractManager.extract(self, input_path, force_extract)
42 def extract(self, input_path: str, force_extract: bool = False) -> str:
---> 43 extractor_format = self.extractor.infer_extractor_format(input_path)
44 if not extractor_format:
45 return input_path
File ~/.local/lib/python3.10/site-packages/datasets/utils/extract.py:316, in Extractor.infer_extractor_format(cls, path)
314 magic_number = cls._read_magic_number(path, magic_number_max_length)
315 for extractor_format, extractor in cls.extractors.items():
--> 316 if extractor.is_extractable(path, magic_number=magic_number):
317 return extractor_format
File ~/.local/lib/python3.10/site-packages/datasets/utils/extract.py:84, in TarExtractor.is_extractable(cls, path, **kwargs)
82 @classmethod
83 def is_extractable(cls, path: Union[Path, str], **kwargs) -> bool:
---> 84 return tarfile.is_tarfile(path)
File /home/.conda/envs/whisperx/lib/python3.10/tarfile.py:2816, in is_tarfile(name)
2814 t = open(fileobj=name)
2815 else:
-> 2816 t = open(name)
2817 t.close()
2818 return True
File /home/.conda/envs/whisperx/lib/python3.10/tarfile.py:1833, in TarFile.open(cls, name, mode, fileobj, bufsize, **kwargs)
1831 saved_pos = fileobj.tell()
1832 try:
-> 1833 return func(name, "r", fileobj, **kwargs)
1834 except (ReadError, CompressionError) as e:
1835 error_msgs.append(f'- method {comptype}: {e!r}')
File /home/.conda/envs/whisperx/lib/python3.10/tarfile.py:1899, in TarFile.gzopen(cls, name, mode, fileobj, compresslevel, **kwargs)
1896 raise CompressionError("gzip module is not available") from None
1898 try:
-> 1899 fileobj = GzipFile(name, mode + "b", compresslevel, fileobj)
1900 except OSError as e:
1901 if fileobj is not None and mode == 'r':
File /home/.conda/envs/whisperx/lib/python3.10/gzip.py:174, in GzipFile.__init__(self, filename, mode, compresslevel, fileobj, mtime)
172 mode += 'b'
173 if fileobj is None:
--> 174 fileobj = self.myfileobj = builtins.open(filename, mode or 'rb')
175 if filename is None:
176 filename = getattr(fileobj, 'name', '')
IsADirectoryError: [Errno 21] Is a directory: '/opt/huggingface_cache/hub/datasets--projecte-aina--corts_valencianes_asr_a/snapshots/7e8f3355bec3ed1bd926ba26fe97e4bf134c4ed9'
I do not understand how this changed the error, given that the "name" parameter has "None" already as its default value.
I'd appreciate any insights into what's causing these errors or how I might work around them. If there's something I'm missing in the expected loading of the dataset, or if a specific version of the datasets library might solve it.
Hi Mauro,
Thanks for reporting this. We identified that some .tar.gz files in the dataset contained unsafe paths (like ../), which caused the extraction errors you saw (Extraction of ... is blocked (illegal path)). This was triggering the problem when loading the dataset.
We have now sanitized all the tar files by cleaning their internal paths to remove these components. After this fix, the extraction should work correctly without those errors.
Please try loading the dataset again with the updated files and instructions from readme, and let us know if you encounter any further issues.