Datasets:

lelapa

/

Inkuba-instruct

like 7

Follow

Lelapa AI 45

Hey Lelapa, thanks for the great resource.

It seems I am no longer able to download this directly anymore using datasets.load_dataset because of some misformatted metadata. My understanding is that the huggingface API expects some sort of 'tags' key in the metadata ?

I used to be able to access it (downloaded it fine in August 2024)

Steps to reproduce:

• Using datasets 2.16.01
• HF says I have been granted access to this dataset
• Running inkuba = datasets.load_dataset("lelapa/Inkuba-instruct", split="swahili_train", token="<MY_HF_TOKEN>")

Error:
"""

KeyError Traceback (most recent call last)
Cell In[15], line 1
----> 1 inkuba = datasets.load_dataset("lelapa/Inkuba-instruct", split="swahili_train", token="")

File ~/miniconda3/lib/python3.11/site-packages/datasets/load.py:2523, in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, verification_mode, ignore_verifications, keep_in_memory, save_infos, revision, token, use_auth_token, task, streaming, num_proc, storage_options, trust_remote_code, **config_kwargs)
2518 verification_mode = VerificationMode(
2519 (verification_mode or VerificationMode.BASIC_CHECKS) if not save_infos else VerificationMode.ALL_CHECKS
2520 )
2522 # Create a dataset builder
-> 2523 builder_instance = load_dataset_builder(
2524 path=path,
2525 name=name,
2526 data_dir=data_dir,
2527 data_files=data_files,
2528 cache_dir=cache_dir,
2529 features=features,
2530 download_config=download_config,
2531 download_mode=download_mode,
2532 revision=revision,
2533 token=token,
2534 storage_options=storage_options,
2535 trust_remote_code=trust_remote_code,
2536 _require_default_config_name=name is None,
2537 **config_kwargs,
2538 )
2540 # Return iterable dataset in case of streaming
2541 if streaming:

File ~/miniconda3/lib/python3.11/site-packages/datasets/load.py:2195, in load_dataset_builder(path, name, data_dir, data_files, cache_dir, features, download_config, download_mode, revision, token, use_auth_token, storage_options, trust_remote_code, _require_default_config_name, **config_kwargs)
2193 download_config = download_config.copy() if download_config else DownloadConfig()
2194 download_config.storage_options.update(storage_options)
-> 2195 dataset_module = dataset_module_factory(
2196 path,
2197 revision=revision,
2198 download_config=download_config,
2199 download_mode=download_mode,
2200 data_dir=data_dir,
2201 data_files=data_files,
2202 cache_dir=cache_dir,
2203 trust_remote_code=trust_remote_code,
2204 _require_default_config_name=_require_default_config_name,
2205 _require_custom_configs=bool(config_kwargs),
2206 )
2207 # Get dataset builder class from the processing script
2208 builder_kwargs = dataset_module.builder_kwargs

File ~/miniconda3/lib/python3.11/site-packages/datasets/load.py:1846, in dataset_module_factory(path, revision, download_config, download_mode, dynamic_modules_path, data_dir, data_files, cache_dir, trust_remote_code, _require_default_config_name, _require_custom_configs, **download_kwargs)
1841 if isinstance(e1, FileNotFoundError):
1842 raise FileNotFoundError(
1843 f"Couldn't find a dataset script at {relative_to_absolute_path(combined_path)} or any data file in the same directory. "
1844 f"Couldn't find '{path}' on the Hugging Face Hub either: {type(e1).name}: {e1}"
1845 ) from None
-> 1846 raise e1 from None
1847 else:
1848 raise FileNotFoundError(
1849 f"Couldn't find a dataset script at {relative_to_absolute_path(combined_path)} or any data file in the same directory."
1850 )

File ~/miniconda3/lib/python3.11/site-packages/datasets/load.py:1791, in dataset_module_factory(path, revision, download_config, download_mode, dynamic_modules_path, data_dir, data_files, cache_dir, trust_remote_code, _require_default_config_name, _require_custom_configs, **download_kwargs)
1787 raise DatasetNotFoundError(
1788 msg + ". If the repo is private or gated, make sure to log in with huggingface-cli login."
1789 )
1790 else:
-> 1791 raise e
1792 if filename in [sibling.rfilename for sibling in dataset_info.siblings]: # contains a dataset script
1793 fs = HfFileSystem(endpoint=config.HF_ENDPOINT, token=download_config.token)

File ~/miniconda3/lib/python3.11/site-packages/datasets/load.py:1765, in dataset_module_factory(path, revision, download_config, download_mode, dynamic_modules_path, data_dir, data_files, cache_dir, trust_remote_code, _require_default_config_name, _require_custom_configs, **download_kwargs)
1763 hf_api = HfApi(config.HF_ENDPOINT)
1764 try:
-> 1765 dataset_info = hf_api.dataset_info(
1766 repo_id=path,
1767 revision=revision,
1768 token=download_config.token,
1769 timeout=100.0,
1770 )
1771 except Exception as e: # noqa catch any exception of hf_hub and consider that the dataset doesn't exist
1772 if isinstance(
1773 e,
1774 (
(...)
1778 ),
1779 ):

File ~/miniconda3/lib/python3.11/site-packages/huggingface_hub/utils/_validators.py:118, in validate_hf_hub_args.._inner_fn(*args, **kwargs)
115 if check_use_auth_token:
116 kwargs = smoothly_deprecate_use_auth_token(fn_name=fn.name, has_token=has_token, kwargs=kwargs)
--> 118 return fn(*args, **kwargs)

File ~/miniconda3/lib/python3.11/site-packages/huggingface_hub/hf_api.py:2150, in HfApi.dataset_info(self, repo_id, revision, timeout, files_metadata, token)
2148 hf_raise_for_status(r)
2149 data = r.json()
-> 2150 return DatasetInfo(**data)

File ~/miniconda3/lib/python3.11/site-packages/huggingface_hub/hf_api.py:770, in DatasetInfo.init(self, **kwargs)
768 self.likes = kwargs.pop("likes")
769 self.paperswithcode_id = kwargs.pop("paperswithcode_id", None)
--> 770 self.tags = kwargs.pop("tags")
771 card_data = kwargs.pop("cardData", None) or kwargs.pop("card_data", None)
772 self.card_data = (
773 DatasetCardData(**card_data, ignore_metadata_errors=True) if isinstance(card_data, dict) else card_data
774 )

KeyError: 'tags'
"""