html_url
stringlengths 48
51
| title
stringlengths 1
290
| comments
sequencelengths 0
30
| body
stringlengths 0
228k
⌀ | number
int64 2
7.08k
|
|---|---|---|---|---|
https://github.com/huggingface/datasets/issues/4363 | The dataset preview is not available for this split. | [
"Hi! A dataset has to be streamable to work with the viewer. I did a quick test, and yours is, so this might be a bug in the viewer. cc @severo \r\n",
"Looking at it. The message is now:\r\n\r\n```\r\nMessage: cannot cache function '__shear_dense': no locator available for file '/src/services/worker/.venv/lib/python3.9/site-packages/librosa/util/utils.py'\r\n```\r\n\r\nso possibly it's related to the libraries versions?\r\n",
"Maybe this SO thread can help: https://stackoverflow.com/questions/59290386/runtimeerror-at-cannot-cache-function-shear-dense-no-locator-available-fo",
"Same error for https://huggingface.co/datasets/LIUM/tedlium/viewer/release1/test. cc @sanchit-gandhi . I'm on it",
"Fixed in the datasets viewer, by setting the `NUMBA_CACHE_DIR` env var to a writable directory.",
"https://huggingface.co/datasets/Roh/ryanspeech/viewer/male/train\r\n\r\n<img width=\"1538\" alt=\"Capture d’écran 2022-06-08 à 11 30 08\" src=\"https://user-images.githubusercontent.com/1676121/172583285-4cd49a0f-5715-423b-95dd-5f6ace3b2416.png\">\r\n",
"https://huggingface.co/datasets/LIUM/tedlium/viewer/\r\n\r\n<img width=\"1538\" alt=\"Capture d’écran 2022-06-08 à 14 31 52\" src=\"https://user-images.githubusercontent.com/1676121/172616897-fbcb7df7-0308-4d09-a17d-48826bc91374.png\">\r\n"
] | I have uploaded the corpus developed by our lab in the speech domain to huggingface [datasets](https://huggingface.co/datasets/Roh/ryanspeech). You can read about the companion paper accepted in interspeech 2021 [here](https://arxiv.org/abs/2106.08468). The dataset works fine but I can't make the dataset preview work. It gives me the following error that I don't understand. Can you help me to begin debugging it?
```
Status code: 400
Exception: AttributeError
Message: 'NoneType' object has no attribute 'split'
``` | 4,363 |
https://github.com/huggingface/datasets/issues/4361 | `udhr` doesn't load, dataset checksum mismatch | [] | ## Describe the bug
Loading `udhr` fails due to a checksum mismatch for some source files. Looks like both of the source files on unicode.org have changed:
size + checksum in datasets repo:
```
(hfdev) leon@blade:~/datasets/datasets/udhr$ jq .default.download_checksums < dataset_infos.json
{
"https://unicode.org/udhr/assemblies/udhr_xml.zip": {
"num_bytes": 2273633,
"checksum": "0565fa62c2ff155b84123198bcc967edd8c5eb9679eadc01e6fb44a5cf730fee"
},
"https://unicode.org/udhr/assemblies/udhr_txt.zip": {
"num_bytes": 2107471,
"checksum": "087b474a070dd4096ae3028f9ee0b30dcdcb030cc85a1ca02e143be46327e5e5"
}
}
```
size + checksum regenerated from current source files:
```
(hfdev) leon@blade:~/datasets/datasets/udhr$ rm dataset_infos.json
(hfdev) leon@blade:~/datasets/datasets/udhr$ datasets-cli test --save_infos udhr.py
Using custom data configuration default
Testing builder 'default' (1/1)
Downloading and preparing dataset udhn/default (download: 4.18 MiB, generated: 6.15 MiB, post-processed: Unknown size, total: 10.33 MiB) to /home/leon/.cache/huggingface/datasets/udhn/default/0.0.0/ad74b91fa2b3c386e5751b0c52bdfda76d334f76731142fd432d4acc2e2fde66...
Dataset udhn downloaded and prepared to /home/leon/.cache/huggingface/datasets/udhn/default/0.0.0/ad74b91fa2b3c386e5751b0c52bdfda76d334f76731142fd432d4acc2e2fde66. Subsequent calls will reuse this data.
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 686.69it/s]
Dataset Infos file saved at dataset_infos.json
Test successful.
(hfdev) leon@blade:~/datasets/datasets/udhr$ jq .default.download_checksums < dataset_infos.json
{
"https://unicode.org/udhr/assemblies/udhr_xml.zip": {
"num_bytes": 2389690,
"checksum": "a3350912790196c6e1b26bfd1c8a50e8575f5cf185922ecd9bd15713d7d21438"
},
"https://unicode.org/udhr/assemblies/udhr_txt.zip": {
"num_bytes": 2215441,
"checksum": "cb87ecb25b56f34e4fd6f22b323000524fd9c06ae2a29f122b048789cf17e9fe"
}
}
(hfdev) leon@blade:~/datasets/datasets/udhr$
```
--- is unicode.org a sustainable hosting solution for this dataset?
## Steps to reproduce the bug
```python
from datasets import load_dataset
udhr = load_dataset("udhr")
```
## Expected results
That a Dataset object containing the UDHR data will be returned.
## Actual results
```
>>> d = load_dataset('udhr')
Using custom data configuration default
Downloading and preparing dataset udhn/default (download: 4.18 MiB, generated: 6.15 MiB, post-processed: Unknown size, total: 10.33 MiB) to /home/leon/.cache/huggingface/datasets/udhn/default/0.0.0/ad74b91fa2b3c386e5751b0c52bdfda76d334f76731142fd432d4acc2e2fde66...
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/leon/.local/lib/python3.9/site-packages/datasets/load.py", line 1731, in load_dataset
builder_instance.download_and_prepare(
File "/home/leon/.local/lib/python3.9/site-packages/datasets/builder.py", line 613, in download_and_prepare
self._download_and_prepare(
File "/home/leon/.local/lib/python3.9/site-packages/datasets/builder.py", line 1117, in _download_and_prepare
super()._download_and_prepare(dl_manager, verify_infos, check_duplicate_keys=verify_infos)
File "/home/leon/.local/lib/python3.9/site-packages/datasets/builder.py", line 684, in _download_and_prepare
verify_checksums(
File "/home/leon/.local/lib/python3.9/site-packages/datasets/utils/info_utils.py", line 40, in verify_checksums
raise NonMatchingChecksumError(error_msg + str(bad_urls))
datasets.utils.info_utils.NonMatchingChecksumError: Checksums didn't match for dataset source files:
['https://unicode.org/udhr/assemblies/udhr_xml.zip', 'https://unicode.org/udhr/assemblies/udhr_txt.zip']
>>>
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.2.1 commit/4110fb6034f79c5fb470cf1043ff52180e9c63b7
- Platform: Linux Ubuntu 20.04
- Python version: 3.9.12
- PyArrow version: 8.0.0
| 4,361 |
https://github.com/huggingface/datasets/issues/4358 | Missing dataset tags and sections in some dataset cards | [
"@lhoestq I can take this issue. Please can you point out to me where I can find the other positional arguments?",
"Hi @RohitRathore1 :)\r\n\r\nYou can find all the YAML tags in the tagging app here: https://hf.co/spaces/huggingface/datasets-tagging). They're all passed as arguments to a DatasetMetadata object used to validate the tags."
] | Summary of CircleCI errors for different dataset metadata:
- **BoolQ**: missing 8 required positional arguments: 'annotations_creators', 'language_creators', 'licenses', 'multilinguality', 'size_categories', 'source_datasets', 'task_categories', and 'task_ids'
- **Conllpp**: expected some content in section `Citation Information` but it is empty.
- **GLUE**: 'annotations_creators', 'language_creators', 'source_datasets' :['unknown'] are not registered tags
- **ConLL2003**: field 'task_ids': ['part-of-speech-tagging'] are not registered tags for 'task_ids'
- **Hate_speech18:** Expected some content in section `Data Instances` but it is empty, Expected some content in section `Data Splits` but it is empty
- **Jjigsaw_toxicity_pred**: `Citation Information` but it is empty.
- **LIAR** : `Data Instances`,`Data Fields`, `Data Splits`, `Citation Information` are empty.
- **MSRA NER** : Dataset Summary`, `Data Instances`, `Data Fields`, `Data Splits`, `Citation Information` are empty.
- **sem_eval_2010_task_8**: missing 8 required positional arguments: 'annotations_creators', 'language_creators', 'licenses', 'multilinguality', 'size_categories', 'source_datasets', 'task_categories', and 'task_ids'
- **sms_spam**: `Data Instances` and`Data Splits` are empty.
- **Quora** : Expected some content in section `Citation Information` but it is empty, missing 8 required positional arguments: 'annotations_creators', 'language_creators', 'licenses', 'multilinguality', 'size_categories', 'source_datasets', 'task_categories', and 'task_ids'
- **sentiment140**: missing 8 required positional arguments: 'annotations_creators', 'language_creators', 'licenses', 'multilinguality', 'size_categories', 'source_datasets', 'task_categories', and 'task_ids' | 4,358 |
https://github.com/huggingface/datasets/issues/4354 | Problems with WMT dataset | [
"Hi! Yes, the docs are outdated. Expect this to be fixed soon. \r\n\r\nIn the meantime, you can try to fix the issue yourself.\r\n\r\nThese are the configs/language pairs supported by `wmt15` from which you can choose:\r\n* `cs-en` (Czech - English)\r\n* `de-en` (German - English)\r\n* `fi-en` (Finnish- English)\r\n* `fr-en` (French - English)\r\n* `ru-en` (Russian - English)\r\n\r\nAnd the current implementation always uses all the subsets available for a language, so to define custom subsets, you'll have to clone the repo from the Hub and replace the line https://huggingface.co/datasets/wmt15/blob/main/wmt_utils.py#L688 with:\r\n`for split, ss_names in (self._subsets if self.config.subsets is None else self.config.subsets).items()`\r\n\r\nThen, you can load the dataset as follows:\r\n```python\r\nfrom datasets import load_dataset\r\ndset = load_dataset(\"path/to/local/wmt15_folder\", \"<one of 5 available configs>\", subsets=...)",
"@mariosasko thanks a lot for the suggested fix! ",
"Hi @mariosasko \r\n\r\nAre the docs updated? If not, I would like to get on it. I am new around here, would we helpful, if you can guide.\r\n\r\nThanks",
"Hi @khushmeeet! The docs haven't been updated, so feel free to work on this issue. This is a tricky issue, so I'll give the steps you can follow to fix this:\r\n\r\nFirst, this code:\r\nhttps://github.com/huggingface/datasets/blob/7cff5b9726a223509dbd6224de3f5f452c8d924f/src/datasets/load.py#L113-L118\r\n\r\nneeds to be replaced with (makes the dataset builder search more robust and allows us to remove the ABC stuff from `wmt_utils.py`):\r\n```python\r\n for name, obj in module.__dict__.items():\r\n if inspect.isclass(obj) and issubclass(obj, main_cls_type):\r\n if inspect.isabstract(obj):\r\n continue\r\n module_main_cls = obj\r\n obj_module = inspect.getmodule(obj)\r\n if obj_module is not None and module == obj_module:\r\n break\r\n```\r\n\r\nThen, all the `wmt_utils.py` scripts need to be updated as follows (these are the diffs with the requiered changes):\r\n````diff\r\n import os\r\n import re\r\n import xml.etree.cElementTree as ElementTree\r\n-from abc import ABC, abstractmethod\r\n\r\n import datasets\r\n````\r\n\r\n````diff\r\nlogger = datasets.logging.get_logger(__name__)\r\n\r\n\r\n _DESCRIPTION = \"\"\"\\\r\n-Translate dataset based on the data from statmt.org.\r\n+Translation dataset based on the data from statmt.org.\r\n\r\n-Versions exists for the different years using a combination of multiple data\r\n-sources. The base `wmt_translate` allows you to create your own config to choose\r\n-your own data/language pair by creating a custom `datasets.translate.wmt.WmtConfig`.\r\n+Versions exist for different years using a combination of data\r\n+sources. The base `wmt` allows you to create a custom dataset by choosing\r\n+your own data/language pair. This can be done as follows:\r\n\r\n ```\r\n-config = datasets.wmt.WmtConfig(\r\n- version=\"0.0.1\",\r\n+from datasets import inspect_dataset, load_dataset_builder\r\n+\r\n+inspect_dataset(\"<insert the dataset name\", \"path/to/scripts\")\r\n+builder = load_dataset_builder(\r\n+ \"path/to/scripts/wmt_utils.py\",\r\n language_pair=(\"fr\", \"de\"),\r\n subsets={\r\n datasets.Split.TRAIN: [\"commoncrawl_frde\"],\r\n datasets.Split.VALIDATION: [\"euelections_dev2019\"],\r\n },\r\n )\r\n-builder = datasets.builder(\"wmt_translate\", config=config)\r\n-```\r\n\r\n+# Standard version\r\n+builder.download_and_prepare()\r\n+ds = builder.as_dataset()\r\n+\r\n+# Streamable version\r\n+ds = builder.as_streaming_dataset()\r\n+```\r\n \"\"\"\r\n````\r\n\r\n````diff\r\n+class Wmt(datasets.GeneratorBasedBuilder):\r\n \"\"\"WMT translation dataset.\"\"\"\r\n+\r\n+ BUILDER_CONFIG_CLASS = WmtConfig\r\n\r\n def __init__(self, *args, **kwargs):\r\n- if type(self) == Wmt and \"config\" not in kwargs: # pylint: disable=unidiomatic-typecheck\r\n- raise ValueError(\r\n- \"The raw `wmt_translate` can only be instantiated with the config \"\r\n- \"kwargs. You may want to use one of the `wmtYY_translate` \"\r\n- \"implementation instead to get the WMT dataset for a specific year.\"\r\n- )\r\n super(Wmt, self).__init__(*args, **kwargs)\r\n\r\n @property\r\n- @abstractmethod\r\n def _subsets(self):\r\n \"\"\"Subsets that make up each split of the dataset.\"\"\"\r\n````\r\n```diff\r\n \"\"\"Subsets that make up each split of the dataset for the language pair.\"\"\"\r\n source, target = self.config.language_pair\r\n filtered_subsets = {}\r\n- for split, ss_names in self._subsets.items():\r\n+ subsets = self._subsets if self.config.subsets is None else self.config.subsets\r\n+ for split, ss_names in subsets.items():\r\n filtered_subsets[split] = []\r\n for ss_name in ss_names:\r\n dataset = DATASET_MAP[ss_name]\r\n```\r\n\r\n`wmt14`, `wmt15`, `wmt16`, `wmt17`, `wmt18`, `wmt19` and `wmt_t2t` have this script, so all of them need to be updated. Also, the dataset summaries from the READMEs of these datasets need to be updated to match the new `_DESCRIPTION` string. And that's it! Let me know if you need additional help.",
"Hi @mariosasko ,\r\n\r\nI have made the changes as suggested by you and have opened a PR #4537.\r\n\r\nThanks",
"Resolved via #4554 "
] | ## Describe the bug
I am trying to load WMT15 dataset and to define which data-sources to use for train/validation/test splits, but unfortunately it seems that the official documentation at [https://huggingface.co/datasets/wmt15#:~:text=Versions%20exists%20for,wmt_translate%22%2C%20config%3Dconfig)](https://huggingface.co/datasets/wmt15#:~:text=Versions%20exists%20for,wmt_translate%22%2C%20config%3Dconfig)) doesn't work anymore.
## Steps to reproduce the bug
```shell
>>> import datasets
>>> a = datasets.translate.wmt.WmtConfig()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
AttributeError: module 'datasets' has no attribute 'translate'
>>> a = datasets.wmt.WmtConfig()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
AttributeError: module 'datasets' has no attribute 'wmt'
```
## Expected results
To load WMT15 with given data-sources.
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.0.0
- Platform: Linux-5.10.0-10-amd64-x86_64-with-glibc2.17
- Python version: 3.8.12
- PyArrow version: 7.0.0
- Pandas version: 1.4.1
| 4,354 |
https://github.com/huggingface/datasets/issues/4352 | When using `dataset.map()` if passed `Features` types do not match what is returned from the mapped function, execution does not except in an obvious way | [
"Hi ! Thanks for reporting :) `datasets` usually returns a `pa.lib.ArrowInvalid` error if the feature types don't match.\r\n\r\nIt would be awesome if we had a way to reproduce the `OverflowError` in this case, to better understand what happened and be able to provide the best error message"
] | ## Describe the bug
Recently I was trying to using `.map()` to preprocess a dataset. I defined the expected Features and passed them into `.map()` like `dataset.map(preprocess_data, features=features)`. My expected `Features` keys matched what came out of `preprocess_data`, but the types i had defined for them did not match the types that came back. Because of this, i ended up in tracebacks deep inside arrow_dataset.py and arrow_writer.py with exceptions that [did not make clear what the problem was](https://github.com/huggingface/datasets/issues/4349). In short i ended up with overflows and the OS killing processes when Arrow was attempting to write. It wasn't until I dug into `def write_batch` and the loop that loops over cols that I figured out what was going on.
It seems like `.map()` could set a boolean that it's checked that for at least 1 instance from the dataset, the returned data's types match the types provided by the `features` param and error out with a clear exception if they don't. This would make the cause of the issue much more understandable and save people time. This could be construed as a feature but it feels more like a bug to me.
## Steps to reproduce the bug
I don't have explicit code to repro the bug, but ill show an example
Code prior to the fix:
```python
def preprocess(examples):
# returns an encoded data dict with keys that match the features, but the types do not match
...
def get_encoded_data(data):
dataset = Dataset.from_pandas(data)
unique_labels = data['audit_type'].unique().tolist()
features = Features({
'image': Array3D(dtype="uint8", shape=(3, 224, 224))),
'input_ids': Sequence(feature=Value(dtype='int64'))),
'attention_mask': Sequence(Value(dtype='int64'))),
'token_type_ids': Sequence(Value(dtype='int64'))),
'bbox': Array2D(dtype="int64", shape=(512, 4))),
'label': ClassLabel(num_classes=len(unique_labels), names=unique_labels),
})
encoded_dataset = dataset.map(preprocess_data, features=features, remove_columns=dataset.column_names)
```
The Features set that fixed it:
```python
features = Features({
'image': Sequence(Array3D(dtype="uint8", shape=(3, 224, 224))),
'input_ids': Sequence(Sequence(feature=Value(dtype='int64'))),
'attention_mask': Sequence(Sequence(Value(dtype='int64'))),
'token_type_ids': Sequence(Sequence(Value(dtype='int64'))),
'bbox': Sequence(Array2D(dtype="int64", shape=(512, 4))),
'label': ClassLabel(num_classes=len(unique_labels), names=unique_labels),
})
```
The difference between my original code (which was based on documentation) and the working code is the addition of the `Sequence(...)` to 4/5 features as I am working with paginated data and the doc examples are not.
## Expected results
Dataset.map() attempts to validate the data types for each Feature on the first iteration and errors out if they are not validated.
## Actual results
Specify the actual results or traceback.
Based on the value of `writer_batch_size`, execution errors out when Arrow attempts to write because the types do not match, though its error messages dont make this obvious
Example errors:
```
OverflowError: There was an overflow with type <class 'list'>. Try to reduce writer_batch_size to have batches smaller than 2GB.
(offset overflow while concatenating arrays)
```
```
zsh: killed python doc_classification.py
UserWarning: resource_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
datasets version: 2.1.0
Platform: macOS-12.2.1-arm64-arm-64bit
Python version: 3.9.12
PyArrow version: 6.0.1
Pandas version: 1.4.2
| 4,352 |
https://github.com/huggingface/datasets/issues/4351 | Add optional progress bar for .save_to_disk(..) and .load_from_disk(..) when working with remote filesystems | [
"Hi! I like this idea. For consistency with `load_dataset`, we can use `fsspec`'s `TqdmCallback` in `.load_from_disk` to monitor the number of bytes downloaded, and in `.save_to_disk`, we can track the number of saved shards for consistency with `push_to_hub` (after we implement https://github.com/huggingface/datasets/issues/4196)."
] | **Is your feature request related to a problem? Please describe.**
When working with large datasets stored on remote filesystems(such as s3), the process of uploading a dataset could take really long time. For instance: I was uploading a re-processed version of wmt17 en-ru to my s3 bucket and it took like 35 minutes(and that's given that I have a fiber optic connection). The only output during that process was a progress bar for flattening indices and then ~35 minutes of complete silence.
**Describe the solution you'd like**
I want to be able to enable a progress bar when calling .save_to_disk(..) and .load_from_disk(..), it would track either amount of bytes sent/received or amount of records written/loaded, and will give some ETA. Basically just tqdm.
**Describe alternatives you've considered**
- Save dataset to tmp folder at the disk and then upload it using custom wrapper over botocore, which will work with progress bar, like [this](https://alexwlchan.net/2021/04/s3-progress-bars/). | 4,351 |
https://github.com/huggingface/datasets/issues/4349 | Dataset.map()'s fails at any value of parameter writer_batch_size | [
"Note that this same issue occurs even if i preprocess with the more default way of tokenizing that uses LayoutLMv2Processor's internal OCR:\r\n\r\n```python\r\n feature_extractor = LayoutLMv2FeatureExtractor()\r\n tokenizer = LayoutLMv2Tokenizer.from_pretrained(\"microsoft/layoutlmv2-base-uncased\")\r\n processor = LayoutLMv2Processor(feature_extractor, tokenizer)\r\n encoded_inputs = processor(images, padding=\"max_length\", truncation=True)\r\n encoded_inputs[\"image\"] = np.array(encoded_inputs[\"image\"])\r\n encoded_inputs[\"label\"] = examples['label_id']\r\n```",
"Wanted to make sure anyone that finds this also finds my other report: https://github.com/huggingface/datasets/issues/4352",
"Did you close it because you found that it was due to the incorrect Feature types ?",
"Yeah-- my analysis of the issue was wrong in this one so I just closed it while linking to the new issue",
"I met with the same problem when doing some experiments about layoutlm. I tried to set the writer_batch_size to 1, and the error still exists. Is there any solutions to this problem?",
"The problem lies in how your Features are defined. It's erroring out when it actually goes to write them to disk"
] | ## Describe the bug
If the the value of `writer_batch_size` is less than the total number of instances in the dataset it will fail at that same number of instances. If it is greater than the total number of instances, it fails on the last instance.
Context:
I am attempting to fine-tune a pre-trained HuggingFace transformers model called LayoutLMv2. This model takes three inputs: document images, words and word bounding boxes. [The Processor for this model has two options](https://huggingface.co/docs/transformers/model_doc/layoutlmv2#usage-layoutlmv2processor), the default is passing a document to the Processor and allowing it to create images of the document and use PyTesseract to perform OCR and generate words/bounding boxes. The other option is to provide `revision="no_ocr"` to the pre-trained model which allows you to use your own OCR results (in my case, Amazon Textract) so you have to provide the image, words and bounding boxes yourself. I am using this second option which might be good context for the bug.
I am using the Dataset.map() paradigm to create these three inputs, encode them and save the dataset. Note that my documents (data instances) on average are fairly large and can range from 1 page up to 300 pages.
Code I am using is provided below
## Steps to reproduce the bug
I do not have explicit sample code, but I will paste the code I'm using in case reading it helps. When `.map()` is called, the dataset has 2933 rows, many of which represent large pdf documents.
```python
def get_encoded_data(data):
dataset = Dataset.from_pandas(data)
unique_labels = data['label'].unique()
features = Features({
'image': Array3D(dtype="int64", shape=(3, 224, 224)),
'input_ids': Sequence(feature=Value(dtype='int64')),
'attention_mask': Sequence(Value(dtype='int64')),
'token_type_ids': Sequence(Value(dtype='int64')),
'bbox': Array2D(dtype="int64", shape=(512, 4)),
'label': ClassLabel(num_classes=len(unique_labels), names=unique_labels),
})
encoded_dataset = dataset.map(preprocess_data, features=features, remove_columns=dataset.column_names, writer_batch_size=dataset.num_rows+1)
encoded_dataset.save_to_disk(TRAINING_DATA_PATH + ENCODED_DATASET_NAME)
encoded_dataset.set_format(type="torch")
return encoded_dataset
```
```python
PROCESSOR = LayoutLMv2Processor.from_pretrained(MODEL_PATH, revision="no_ocr", use_fast=False)
def preprocess_data(examples):
directory = os.path.join(FILES_PATH, examples['file_location'])
images_dir = os.path.join(directory, PDF_IMAGE_DIR)
textract_response_path = os.path.join(directory, 'textract.json')
doc_meta_path = os.path.join(directory, 'doc_meta.json')
textract_document = get_textract_document(textract_response_path, doc_meta_path)
images, words, bboxes = get_doc_training_data(images_dir, textract_document)
encoded_inputs = PROCESSOR(images, words, boxes=bboxes, padding="max_length", truncation=True)
# https://github.com/NielsRogge/Transformers-Tutorials/issues/36
encoded_inputs["image"] = np.array(encoded_inputs["image"])
encoded_inputs["label"] = examples['label_id']
return encoded_inputs
```
## Expected results
My expectation is that `writer_batch_size` allows one to simply trade off performance and memory requirements, not that it must be a specific number for `.map()` to function correctly.
## Actual results
If writer_batch_size is set to a value less than the number of rows, I get either:
```
OverflowError: There was an overflow with type <class 'list'>. Try to reduce writer_batch_size to have batches smaller than 2GB.
(offset overflow while concatenating arrays)
```
or simply
```
zsh: killed python doc_classification.py
UserWarning: resource_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown
```
If it is greater than the number of rows, i get the `zsh: killed` error above
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.1.0
- Platform: macOS-12.2.1-arm64-arm-64bit
- Python version: 3.9.12
- PyArrow version: 6.0.1
- Pandas version: 1.4.2
| 4,349 |
https://github.com/huggingface/datasets/issues/4348 | `inspect` functions can't fetch dataset script from the Hub | [
"Hi, thanks for reporting! `git bisect` points to #2986 as the PR that introduced the bug. Since then, there have been some additional changes to the loading logic, and in the current state, `force_local_path` (set via `local_path`) forbids pulling a script from the internet instead of downloading it: https://github.com/huggingface/datasets/blob/cfae0545b2ba05452e16136cacc7d370b4b186a1/src/datasets/inspect.py#L89-L91\r\n\r\ncc @lhoestq: `force_local_path` is only used in `inspect_dataset` and `inspect_metric`. Is it OK if we revert the behavior to match the old one?",
"Good catch ! Yea I think it's fine :)"
] | The `inspect_dataset` and `inspect_metric` functions are unable to retrieve a dataset or metric script from the Hub and store it locally at the specified `local_path`:
```py
>>> from datasets import inspect_dataset
>>> inspect_dataset('rotten_tomatoes', local_path='path/to/my/local/folder')
FileNotFoundError: Couldn't find a dataset script at /content/rotten_tomatoes/rotten_tomatoes.py or any data file in the same directory.
``` | 4,348 |
https://github.com/huggingface/datasets/issues/4346 | GH Action to build documentation never ends | [] | ## Describe the bug
See: https://github.com/huggingface/datasets/runs/6418035586?check_suite_focus=true
I finally forced the cancel of the workflow. | 4,346 |
https://github.com/huggingface/datasets/issues/4343 | Metrics documentation is not accessible in the datasets doc UI | [
"Hey @fxmarty :) Yes we are working on showing the docs of all the metrics on the Hugging face website. If you want to follow the advancements you can check the [evaluate](https://github.com/huggingface/evaluate) repository cc @lvwerra @sashavor "
] | **Is your feature request related to a problem? Please describe.**
Search for a metric name like "seqeval" yields no results on https://huggingface.co/docs/datasets/master/en/index . One needs to go look in `datasets/metrics/README.md` to find the doc. Even in the `README.md`, it can be hard to understand what the metric expects as an input, for example for `squad` there is a [key `id`](https://github.com/huggingface/datasets/blob/1a4c185663a6958f48ec69624473fdc154a36a9d/metrics/squad/squad.py#L42) documented only in the function doc but not in the `README.md`, and one needs to go look into the code to understand what the metric expects.
**Describe the solution you'd like**
Have the documentation for metrics appear as well in the doc UI, e.g. this https://github.com/huggingface/datasets/blob/1a4c185663a6958f48ec69624473fdc154a36a9d/metrics/squad/squad.py#L21-L63
I know there are plans to migrate metrics to the evaluate library, but just pointing this out.
| 4,343 |
https://github.com/huggingface/datasets/issues/4341 | Failing CI on Windows for sari and wiki_split metrics | [] | ## Describe the bug
Our CI is failing from yesterday on Windows for metrics: sari and wiki_split
```
FAILED tests/test_metric_common.py::LocalMetricTest::test_load_metric_sari - ...
FAILED tests/test_metric_common.py::LocalMetricTest::test_load_metric_wiki_split
```
See: https://app.circleci.com/pipelines/github/huggingface/datasets/11928/workflows/79daa5e7-65c9-4e85-829b-00d2bfbd076a/jobs/71594 | 4,341 |
https://github.com/huggingface/datasets/issues/4327 | `wikipedia` pre-processed datasets | [
"Hi @vpj, thanks for reporting.\r\n\r\nI'm sorry, but I can't reproduce your bug: I load \"20220301.simple\"in 9 seconds:\r\n```shell\r\ntime python -c \"from datasets import load_dataset; load_dataset('wikipedia', '20220301.simple')\"\r\n\r\nDownloading and preparing dataset wikipedia/20220301.simple (download: 228.58 MiB, generated: 224.18 MiB, post-processed: Unknown size, total: 452.76 MiB) to .../.cache/huggingface/datasets/wikipedia/20220301.simple/2.0.0/aa542ed919df55cc5d3347f42dd4521d05ca68751f50dbc32bae2a7f1e167559...\r\nDownloading: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1.66k/1.66k [00:00<00:00, 1.02MB/s]\r\nDownloading: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 235M/235M [00:02<00:00, 82.8MB/s]\r\nDataset wikipedia downloaded and prepared to .../.cache/huggingface/datasets/wikipedia/20220301.simple/2.0.0/aa542ed919df55cc5d3347f42dd4521d05ca68751f50dbc32bae2a7f1e167559. Subsequent calls will reuse this data.\r\n100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 290.75it/s]\r\n\r\nreal\t0m9.693s\r\nuser\t0m6.002s\r\nsys\t0m3.260s\r\n```\r\n\r\nCould you please check your environment info, as requested when opening this issue?\r\n```\r\n## Environment info\r\n<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->\r\n- `datasets` version:\r\n- Platform:\r\n- Python version:\r\n- PyArrow version:\r\n```\r\nMaybe you are using an old version of `datasets`...",
"Downloading and processing `wikipedia simple` dataset completed in under 11sec on M1 Mac. Could you please check `dataset` version as mentioned by @albertvillanova? Also check system specs, if system is under load processing could take some time I guess."
] | ## Describe the bug
[Wikipedia](https://huggingface.co/datasets/wikipedia) dataset readme says that certain subsets are preprocessed. However it seems like they are not available. When I try to load them it takes a really long time, and it seems like it's processing them.
## Steps to reproduce the bug
```python
from datasets import load_dataset
load_dataset("wikipedia", "20220301.en")
```
## Expected results
To load the dataset
## Actual results
Takes a very long time to load (after downloading)
After `Downloading data files: 100%`. It takes hours and gets killed.
Tried `wikipedia.simple` and it got processed after ~30mins. | 4,327 |
https://github.com/huggingface/datasets/issues/4325 | Dataset Viewer issue for strombergnlp/offenseval_2020, strombergnlp/polstance | [
"Not sure if it's related... I was going to raise an issue for https://huggingface.co/datasets/domenicrosati/TruthfulQA which also has the same issue... https://huggingface.co/datasets/domenicrosati/TruthfulQA/viewer/domenicrosati--TruthfulQA/train \r\n\r\n",
"Yes, it's related. The backend behind the dataset viewer is currently under too much load, and these datasets are still in the jobs queue. We're actively working on this issue, and we expect to fix the issue permanently soon. Thanks for your patience 🙏 ",
"Thanks @severo and no worries! - a suggestion for a UI usability thing maybe is to indicate that the dataset processing is in the job queue (rather than no data?)",
"Thanks, these are working great now (including @domenicrosati 's, afaics!)"
] | ### Link
https://huggingface.co/datasets/strombergnlp/offenseval_2020/viewer/ar/train
### Description
The viewer isn't running for these two datasets. I left it overnight because a wait sometimes helps things get loaded, and the error messages have all gone, but the datasets are still turning up blank in viewer. Maybe it needs a bit more time.
* https://huggingface.co/datasets/strombergnlp/polstance/viewer/PolStance/train
* https://huggingface.co/datasets/strombergnlp/offenseval_2020/viewer/ar/train
While offenseval_2020 is gated w. prompt, the other gated previews I have run fine in Viewer, e.g. https://huggingface.co/datasets/strombergnlp/shaj , so I'm a bit stumped!
### Owner
Yes | 4,325 |
https://github.com/huggingface/datasets/issues/4324 | Support >1 PWC dataset per dataset card | [
"Hi @leondz, I agree it would be nice. We'll see what we can do ;)"
] | **Is your feature request related to a problem? Please describe.**
Some datasets cover more than one dataset on PapersWithCode. For example, the OffensEval 2020 challenge involved five languages, and there's one dataset to cover all five datasets, [`strombergnlp/offenseval_2020`](https://huggingface.co/datasets/strombergnlp/offenseval_2020). However, the yaml `paperswithcode_id:` dataset card entry only supports one value; when multiple are added, the PWC link disappears from the dataset page.
Because the link from a PapersWithCode dataset to a Hugging Face Hub entry can't be entered manually and seems to be scraped, this means end users don't have a way of getting a dataset reader link to appear on all the PWC datasets supported by one HF Hub Dataset reader.
It's not super unusual to have papers introduce multiple parallel variants of a dataset and would be handy to reflect this, so e.g. dataset maintainers can DRY, and so dataset users can keep what they're doing simple.
**Describe the solution you'd like**
I'd like `paperswithcode_id:` to support lists and be able to connect with multiple PWC datasets.
**Describe alternatives you've considered**
De-normalising the datasets on HF Hub to create multiple readers for each variation on a task, i.e. instead of a single `offenseval_2020`, having `offenseval_2020_ar`, `offenseval_2020_da`, `offenseval_2020_gr`, ...
**Additional context**
Hope that's enough
**Priority**
Low | 4,324 |
https://github.com/huggingface/datasets/issues/4323 | Audio can not find value["bytes"] | [
"\r\n\r\nthat is reason my bytes`s empty\r\nbut i have some confused why path prior is higher than bytes?\r\n\r\nif you can make bytes in _generate_examples , you don`t have to make bytes to path?\r\nbecause we have path and bytes already",
"> but i have some confused why path prior is higher than bytes?\r\n\r\nIf the audio file is already available locally, we don't need to store the bytes again.\r\n\r\nIf you don't specify a \"path\" to a local file, then the bytes are stored. You can set \"path\" to None for example.\r\n\r\n> if you can make bytes in _generate_examples , you don`t have to make bytes to path?\r\n> because we have path and bytes already\r\n\r\nIt's useful to pass both \"path\" and \"bytes\" in `_generate_examples`:\r\n- when the dataset has been downloaded, then the \"path\" to the audio files are stored and we can ignore \"bytes\" in order to save disk space.\r\n- when the dataset is loaded in streaming mode, the audio files are not available on your disk and therefore we use the \"bytes\" ",
"@lhoestq \r\nFirst of all, thx for reply\r\n\r\nbut, if i put in \"bytes\" and \"path\"\r\nex) {\"bytes\":\"blah blah~\", \"path\":\"blah blah~\"}\r\n\r\nthat source working that my bytes to empty first,\r\nand then, re-calculate my bytes!\r\n\r\n\r\nif you have some pcm file, pcm is can read bytes.\r\nso, i put in bytes and paths.\r\nbut bytes is been None why encode_example func make None\r\nand then, on decode_example func, we no have bytes. so, calculate bytes to path.\r\npcm is not support librosa or soundfile, error occured!\r\n\r\nthe most important thing is not announced anywhere this situation can be reproduced\r\n\r\nis that truly right process flow?",
"I don't think we support PCM files, feel free to convert your data to WAV for now.\r\n\r\nIt would be awesome to support PCM files though, let me know if you'd like to contribute this feature, I'd be happy to help",
"@lhoestq oh, how can i contribute?",
"You can clone the repository (see the guide on [how to contribute](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-create-a-pull-request)) and see how we can make the `Image.encode_example` method work with PCM data.\r\n\r\nThere might be other ways to approach this problem, but here is what I think is a reasonable one:\r\n\r\nI think `Image.encode_example` should be able to take PCM bytes as input and the sampling rate, and return the WAV bytes (built by combining the PCM bytes and the sampling rate info), so that `Image.decode_example` can read it.\r\n\r\nTo check if the input bytes are PCM data, you can just check if the extension of the `path` is \".pcm\".\r\n",
"maybe i can start to contribute on this sunday!\r\n@lhoestq ",
"@lhoestq plz check my pr #4409 \r\n\r\nam i wrong somting?",
"Thanks, I reviewed your PR :)"
] | ## Describe the bug
I wrote down _generate_examples like:
but where is the bytes?
## Expected results
value["bytes"] is not None, so i can make datasets with bytes, not path
## bytes looks like:
blah blah~~
\xfe\x03\x00\xfb\x06\x1c\x0bo\x074\x03\xaf\x01\x13\x04\xbc\x06\x8c\x05y\x05,\t7\x08\xaf\x03\xc0\xfe\xe8\xfc\x94\xfe\xb7\xfd\xea\xfa\xd5\xf9$\xf9>\xf9\x1f\xf8\r\xf5F\xf49\xf4\xda\xf5-\xf8\n\xf8k\xf8\x07\xfb\x18\xfd\xd9\xfdv\xfd"\xfe\xcc\x01\x1c\x04\x08\x04@\x04{\x06^\tf\t\x1e\x07\x8b\x06\x02\x08\x13\t\x07\x08 \x06g\x06"\x06\xa0\x03\xc6\x002\xff \xff\x1d\xff\x19\xfd?\xfb\xdb\xfa\xfc\xfa$\xfb}\xf9\xe5\xf7\xf9\xf7\xce\xf8.\xf9b\xf9\xc5\xf9\xc0\xfb\xfa\xfcP\xfc\xba\xfbQ\xfc1\xfe\x9f\xff\x12\x00\xa2\x00\x18\x02Z\x03\x02\x04\xb1\x03\xc5\x03W\x04\x82\x04\x8f\x04U\x04\xb6\x04\x10\x05{\x04\x83\x02\x17\x01\x1d\x00\xa0\xff\xec\xfe\x03\xfe#\xfe\xc2\xfe2\xff\xe6\xfe\x9a\xfe~\x01\x91\x08\xb3\tU\x05\x10\x024\x02\xe4\x05\xa8\x07\xa7\x053\x07I\n\x91\x07v\x02\x95\xfd\xbb\xfd\x96\xff\x01\xfe\x1e\xfb\xbb\xf9S\xf8!\xf8\xf4\xf5\xd6\xf3\xf7\xf3l\xf4d\xf6l\xf7d\xf6b\xf7\xc1\xfa(\xfd\xcf\xfd*\xfdq\xfe\xe9\x01\xa8\x03t\x03\x17\x04B\x07\xce\t\t\t\xeb\x06\x0c\x07\x95\x08\x92\t\xbc\x07O\x06\xfb\x06\xd2\x06U\x04\x00\x02\x92\x00\xdc\x00\x84\x00 \xfeT\xfc\xf1\xfb\x82\xfc\x97\xfb}\xf9\x00\xf8_\xf8\x0b\xf9\xe5\xf8\xe2\xf7\xaa\xf8\xb2\xfa\x10\xfbl\xfa\xf5\xf9Y\xfb\xc0\xfd\xe8\xfe\xec\xfe1\x00\xad\x01\xec\x02E\x03\x13\x03\x9b\x03o\x04\xce\x04\xa8\x04\xb2\x04\x1b\x05\xc0\x05\xd2\x04\xe8\x02z\x01\xbe\x00\xae\x00\x07\x00$\xff|\xff\x8e\x00\x13\x00\x10\xff\x98\xff0\x05{\x0b\x05\t\xaa\x03\x82\x01n\x03
blah blah~~
that function not return None
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version:2.2.1
- Platform:ubuntu 18.04
- Python version:3.6.9
- PyArrow version:6.0.1
| 4,323 |
https://github.com/huggingface/datasets/issues/4320 | Multi-news dataset loader attempts to strip wrong character from beginning of summaries | [
"Hi ! Thanks for reporting :)\r\n\r\nThis dataset was simply converted from [tensorflow datasets](https://github.com/tensorflow/datasets/blob/master/tensorflow_datasets/summarization/multi_news.py)\r\n\r\nI think we can just remove the `.strip(\"- \")` and keep this character",
"Cool! I made a PR."
] | ## Describe the bug
The `multi_news.py` data loader has [a line which attempts to strip `"- "` from the beginning of summaries](https://github.com/huggingface/datasets/blob/aa743886221d76afb409d263e1b136e7a71fe2b4/datasets/multi_news/multi_news.py#L97). The actual character in the multi-news dataset, however, is `"– "`, which is different, e.g. `"– " != "- "`.
I would have just opened a PR to fix the mistake, but I am wondering what the motivation for stripping this character is? AFAICT most approaches just leave it in, e.g. the current SOTA on this dataset, [PRIMERA](https://huggingface.co/allenai/PRIMERA-multinews) (you can see its in the generated summaries of the model in their [example notebook](https://github.com/allenai/PRIMER/blob/main/Evaluation_Example.ipynb)).
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.2.0
- Platform: Linux-5.4.188+-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.13
- PyArrow version: 6.0.1
- Pandas version: 1.3.5
| 4,320 |
https://github.com/huggingface/datasets/issues/4310 | Loading dataset with streaming: '_io.BufferedReader' object has no attribute 'loc' | [] | ## Describe the bug
Loading a datasets with `load_dataset` and `streaming=True` returns `AttributeError: '_io.BufferedReader' object has no attribute 'loc'`. Notice that loading with `streaming=False` works fine.
In the following steps we load parquet files but the same happens with pickle files. The problem seems to come from `fsspec` lib, I put in the environment info also `s3fs` and `fsspec` versions since I'm loading from an s3 bucket.
## Steps to reproduce the bug
```python
from datasets import load_dataset
# path is the path to parquet files
data_files = {"train": path + "meta_train.parquet.gzip", "test": path + "meta_test.parquet.gzip"}
dataset = load_dataset("parquet", data_files=data_files, streaming=True)
```
## Expected results
A dataset object `datasets.dataset_dict.DatasetDict`
## Actual results
```
AttributeError Traceback (most recent call last)
<command-562086> in <module>
11
12 data_files = {"train": path + "meta_train.parquet.gzip", "test": path + "meta_test.parquet.gzip"}
---> 13 dataset = load_dataset("parquet", data_files=data_files, streaming=True)
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/load.py in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, ignore_verifications, keep_in_memory, save_infos, revision, use_auth_token, task, streaming, **config_kwargs)
1679 if streaming:
1680 extend_dataset_builder_for_streaming(builder_instance, use_auth_token=use_auth_token)
-> 1681 return builder_instance.as_streaming_dataset(
1682 split=split,
1683 use_auth_token=use_auth_token,
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/builder.py in as_streaming_dataset(self, split, base_path, use_auth_token)
904 )
905 self._check_manual_download(dl_manager)
--> 906 splits_generators = {sg.name: sg for sg in self._split_generators(dl_manager)}
907 # By default, return all splits
908 if split is None:
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/packaged_modules/parquet/parquet.py in _split_generators(self, dl_manager)
30 if not self.config.data_files:
31 raise ValueError(f"At least one data file must be specified, but got data_files={self.config.data_files}")
---> 32 data_files = dl_manager.download_and_extract(self.config.data_files)
33 if isinstance(data_files, (str, list, tuple)):
34 files = data_files
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/utils/streaming_download_manager.py in download_and_extract(self, url_or_urls)
798
799 def download_and_extract(self, url_or_urls):
--> 800 return self.extract(self.download(url_or_urls))
801
802 def iter_archive(self, urlpath_or_buf: Union[str, io.BufferedReader]) -> Iterable[Tuple]:
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/utils/streaming_download_manager.py in extract(self, path_or_paths)
776
777 def extract(self, path_or_paths):
--> 778 urlpaths = map_nested(self._extract, path_or_paths, map_tuple=True)
779 return urlpaths
780
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/utils/py_utils.py in map_nested(function, data_struct, dict_only, map_list, map_tuple, map_numpy, num_proc, types, disable_tqdm, desc)
312 num_proc = 1
313 if num_proc <= 1 or len(iterable) <= num_proc:
--> 314 mapped = [
315 _single_map_nested((function, obj, types, None, True, None))
316 for obj in logging.tqdm(iterable, disable=disable_tqdm, desc=desc)
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/utils/py_utils.py in <listcomp>(.0)
313 if num_proc <= 1 or len(iterable) <= num_proc:
314 mapped = [
--> 315 _single_map_nested((function, obj, types, None, True, None))
316 for obj in logging.tqdm(iterable, disable=disable_tqdm, desc=desc)
317 ]
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/utils/py_utils.py in _single_map_nested(args)
267 return {k: _single_map_nested((function, v, types, None, True, None)) for k, v in pbar}
268 else:
--> 269 mapped = [_single_map_nested((function, v, types, None, True, None)) for v in pbar]
270 if isinstance(data_struct, list):
271 return mapped
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/utils/py_utils.py in <listcomp>(.0)
267 return {k: _single_map_nested((function, v, types, None, True, None)) for k, v in pbar}
268 else:
--> 269 mapped = [_single_map_nested((function, v, types, None, True, None)) for v in pbar]
270 if isinstance(data_struct, list):
271 return mapped
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/utils/py_utils.py in _single_map_nested(args)
249 # Singleton first to spare some computation
250 if not isinstance(data_struct, dict) and not isinstance(data_struct, types):
--> 251 return function(data_struct)
252
253 # Reduce logging to keep things readable in multiprocessing with tqdm
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/utils/streaming_download_manager.py in _extract(self, urlpath)
781 def _extract(self, urlpath: str) -> str:
782 urlpath = str(urlpath)
--> 783 protocol = _get_extraction_protocol(urlpath, use_auth_token=self.download_config.use_auth_token)
784 if protocol is None:
785 # no extraction
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/utils/streaming_download_manager.py in _get_extraction_protocol(urlpath, use_auth_token)
371 urlpath, kwargs = urlpath, {}
372 with fsspec.open(urlpath, **kwargs) as f:
--> 373 return _get_extraction_protocol_with_magic_number(f)
374
375
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/datasets/utils/streaming_download_manager.py in _get_extraction_protocol_with_magic_number(f)
335 def _get_extraction_protocol_with_magic_number(f) -> Optional[str]:
336 """read the magic number from a file-like object and return the compression protocol"""
--> 337 prev_loc = f.loc
338 magic_number = f.read(MAGIC_NUMBER_MAX_LENGTH)
339 f.seek(prev_loc)
/local_disk0/.ephemeral_nfs/envs/pythonEnv-a7e72260-221c-472b-85f4-bec801aee66d/lib/python3.8/site-packages/fsspec/implementations/local.py in __getattr__(self, item)
337
338 def __getattr__(self, item):
--> 339 return getattr(self.f, item)
340
341 def __enter__(self):
AttributeError: '_io.BufferedReader' object has no attribute 'loc'
```
## Environment info
- `datasets` version: 2.1.0
- Platform: Linux-5.4.0-1071-aws-x86_64-with-glibc2.29
- Python version: 3.8.10
- PyArrow version: 8.0.0
- Pandas version: 1.4.2
- `fsspec` version: 2021.08.1
- `s3fs` version: 2021.08.1 | 4,310 |
https://github.com/huggingface/datasets/issues/4306 | `load_dataset` does not work with certain filename. | [
"Never mind. It is because of the caching of datasets..."
] | ## Describe the bug
This is a weird bug that took me some time to find out.
I have a JSON dataset that I want to load with `load_dataset` like this:
```
data_files = dict(train="train.json.zip", val="val.json.zip")
dataset = load_dataset("json", data_files=data_files, field="data")
```
## Expected results
No error.
## Actual results
The val file is loaded as expected, but the train file throws JSON decoding error:
```
╭──────────────────────────── Traceback (most recent call last) ────────────────────────────╮
│ <ipython-input-74-97947e92c100>:5 in <module> │
│ │
│ /home/tiankang/software/anaconda3/lib/python3.8/site-packages/datasets/load.py:1687 in │
│ load_dataset │
│ │
│ 1684 │ try_from_hf_gcs = path not in _PACKAGED_DATASETS_MODULES │
│ 1685 │ │
│ 1686 │ # Download and prepare data │
│ ❱ 1687 │ builder_instance.download_and_prepare( │
│ 1688 │ │ download_config=download_config, │
│ 1689 │ │ download_mode=download_mode, │
│ 1690 │ │ ignore_verifications=ignore_verifications, │
│ │
│ /home/tiankang/software/anaconda3/lib/python3.8/site-packages/datasets/builder.py:605 in │
│ download_and_prepare │
│ │
│ 602 │ │ │ │ │ │ except ConnectionError: │
│ 603 │ │ │ │ │ │ │ logger.warning("HF google storage unreachable. Downloa │
│ 604 │ │ │ │ │ if not downloaded_from_gcs: │
│ ❱ 605 │ │ │ │ │ │ self._download_and_prepare( │
│ 606 │ │ │ │ │ │ │ dl_manager=dl_manager, verify_infos=verify_infos, **do │
│ 607 │ │ │ │ │ │ ) │
│ 608 │ │ │ │ │ # Sync info │
│ │
│ /home/tiankang/software/anaconda3/lib/python3.8/site-packages/datasets/builder.py:694 in │
│ _download_and_prepare │
│ │
│ 691 │ │ │ │
│ 692 │ │ │ try: │
│ 693 │ │ │ │ # Prepare split will record examples associated to the split │
│ ❱ 694 │ │ │ │ self._prepare_split(split_generator, **prepare_split_kwargs) │
│ 695 │ │ │ except OSError as e: │
│ 696 │ │ │ │ raise OSError( │
│ 697 │ │ │ │ │ "Cannot find data file. " │
│ │
│ /home/tiankang/software/anaconda3/lib/python3.8/site-packages/datasets/builder.py:1151 in │
│ _prepare_split │
│ │
│ 1148 │ │ │
│ 1149 │ │ generator = self._generate_tables(**split_generator.gen_kwargs) │
│ 1150 │ │ with ArrowWriter(features=self.info.features, path=fpath) as writer: │
│ ❱ 1151 │ │ │ for key, table in logging.tqdm( │
│ 1152 │ │ │ │ generator, unit=" tables", leave=False, disable=True # not loggin │
│ 1153 │ │ │ ): │
│ 1154 │ │ │ │ writer.write_table(table) │
│ │
│ /home/tiankang/software/anaconda3/lib/python3.8/site-packages/tqdm/notebook.py:257 in │
│ __iter__ │
│ │
│ 254 │ │
│ 255 │ def __iter__(self): │
│ 256 │ │ try: │
│ ❱ 257 │ │ │ for obj in super(tqdm_notebook, self).__iter__(): │
│ 258 │ │ │ │ # return super(tqdm...) will not catch exception │
│ 259 │ │ │ │ yield obj │
│ 260 │ │ # NB: except ... [ as ...] breaks IPython async KeyboardInterrupt │
│ │
│ /home/tiankang/software/anaconda3/lib/python3.8/site-packages/tqdm/std.py:1183 in │
│ __iter__ │
│ │
│ 1180 │ │ # If the bar is disabled, then just walk the iterable │
│ 1181 │ │ # (note: keep this check outside the loop for performance) │
│ 1182 │ │ if self.disable: │
│ ❱ 1183 │ │ │ for obj in iterable: │
│ 1184 │ │ │ │ yield obj │
│ 1185 │ │ │ return │
│ │
│ /home/tiankang/software/anaconda3/lib/python3.8/site-packages/datasets/packaged_modules/j │
│ son/json.py:90 in _generate_tables │
│ │
│ 87 │ │ │ # If the file is one json object and if we need to look at the list of │
│ 88 │ │ │ if self.config.field is not None: │
│ 89 │ │ │ │ with open(file, encoding="utf-8") as f: │
│ ❱ 90 │ │ │ │ │ dataset = json.load(f) │
│ 91 │ │ │ │ │
│ 92 │ │ │ │ # We keep only the field we are interested in │
│ 93 │ │ │ │ dataset = dataset[self.config.field] │
│ │
│ /home/tiankang/software/anaconda3/lib/python3.8/json/__init__.py:293 in load │
│ │
│ 290 │ To use a custom ``JSONDecoder`` subclass, specify it with the ``cls`` │
│ 291 │ kwarg; otherwise ``JSONDecoder`` is used. │
│ 292 │ """ │
│ ❱ 293 │ return loads(fp.read(), │
│ 294 │ │ cls=cls, object_hook=object_hook, │
│ 295 │ │ parse_float=parse_float, parse_int=parse_int, │
│ 296 │ │ parse_constant=parse_constant, object_pairs_hook=object_pairs_hook, **kw) │
│ │
│ /home/tiankang/software/anaconda3/lib/python3.8/json/__init__.py:357 in loads │
│ │
│ 354 │ if (cls is None and object_hook is None and │
│ 355 │ │ │ parse_int is None and parse_float is None and │
│ 356 │ │ │ parse_constant is None and object_pairs_hook is None and not kw): │
│ ❱ 357 │ │ return _default_decoder.decode(s) │
│ 358 │ if cls is None: │
│ 359 │ │ cls = JSONDecoder │
│ 360 │ if object_hook is not None: │
│ │
│ /home/tiankang/software/anaconda3/lib/python3.8/json/decoder.py:337 in decode │
│ │
│ 334 │ │ containing a JSON document). │
│ 335 │ │ │
│ 336 │ │ """ │
│ ❱ 337 │ │ obj, end = self.raw_decode(s, idx=_w(s, 0).end()) │
│ 338 │ │ end = _w(s, end).end() │
│ 339 │ │ if end != len(s): │
│ 340 │ │ │ raise JSONDecodeError("Extra data", s, end) │
│ │
│ /home/tiankang/software/anaconda3/lib/python3.8/json/decoder.py:353 in raw_decode │
│ │
│ 350 │ │ │
│ 351 │ │ """ │
│ 352 │ │ try: │
│ ❱ 353 │ │ │ obj, end = self.scan_once(s, idx) │
│ 354 │ │ except StopIteration as err: │
│ 355 │ │ │ raise JSONDecodeError("Expecting value", s, err.value) from None │
│ 356 │ │ return obj, end │
╰───────────────────────────────────────────────────────────────────────────────────────────╯
JSONDecodeError: Unterminated string starting at: line 85 column 20 (char 60051)
```
However, when I rename the `train.json.zip` to other names (like `training.json.zip`, or even to `train.json`), everything works fine; when I unzip the file to `train.json`, it works as well.
## Environment info
```
- `datasets` version: 2.1.0
- Platform: Linux-4.4.0-131-generic-x86_64-with-glibc2.10
- Python version: 3.8.5
- PyArrow version: 7.0.0
- Pandas version: 1.4.2
``` | 4,306 |
https://github.com/huggingface/datasets/issues/4304 | Language code search does direct matches | [
"Thanks for reporting ! I forwarded the issue to the front-end team :)\r\n\r\nWill keep you posted !\r\n\r\nI also changed the tagging app to suggest two letters code for now."
] | ## Describe the bug
Hi. Searching for bcp47 tags that are just the language prefix (e.g. `sq` or `da`) excludes datasets that have added extra information in their language metadata (e.g. `sq-AL` or `da-bornholm`). The example codes given in the [tagging app](https://huggingface.co/spaces/huggingface/datasets-tagging) encourages addition of the additional codes ("_expected format is BCP47 tags separated for ';' e.g. 'en-US;fr-FR'_") but this would lead to those datasets being hidden in datasets search.
## Steps to reproduce the bug
1. Add a dataset using a variant tag (e.g. [`sq-AL`](https://huggingface.co/datasets?languages=languages:sq-AL))
2. Look for datasets using the full code
3. Note that they're missing when just the language is searched for (e.g. [`sq`](https://huggingface.co/datasets?languages=languages:sq))
Some datasets are already affected by this - e.g. `AmazonScience/massive` is listed under `sq-AL` but not `sq`.
One workaround is for dataset creators to add an additional root language tag to dataset YAML metadata, but it's unclear how to communicate this. It might be possible to index the search on `languagecode.split('-')[0]` but I wanted to float this issue before trying to write any code :)
## Expected results
Datasets using longer bcp47 tags also appear under searches for just the language code; e.g. Quebecois datasets (`fr-CA`) would come up when looking for French datasets with no region specification (`fr`), or US English (`en-US`) datasets would come up when searching for English datasets (`en`).
## Actual results
The language codes seem to be directly string matched, excluding datasets with specific language tags from non-specific searches.
## Environment info
(web app) | 4,304 |
https://github.com/huggingface/datasets/issues/4298 | Normalise license names | [
"we'll add the same server-side metadata validation system as for hf.co/models soon-ish\r\n\r\n(you can check on hf.co/models that licenses are \"clean\")",
"Fixed by #4367."
] | **Is your feature request related to a problem? Please describe.**
When browsing datasets, the Licenses tag cloud (bottom left of e.g. https://huggingface.co/datasets) has multiple variants of the same license. This means the options exclude datasets arbitrarily, giving users artificially low recall. The cause of the dupes is probably due to a bit of variation in metadata.
**Describe the solution you'd like**
I'd like the licenses in metadata to follow the same standard as much as possible, to remove this problem. I'd like to go ahead and normalise the dataset metadata to follow the format & values given in [src/datasets/utils/resources/licenses.json](https://github.com/huggingface/datasets/blob/master/src/datasets/utils/resources/licenses.json) .
**Describe alternatives you've considered**
None
**Additional context**
None
**Priority**
Low
| 4,298 |
https://github.com/huggingface/datasets/issues/4297 | Datasets YAML tagging space is down | [
"@lhoestq @albertvillanova `update-task-list` branch does not exist anymore, should point to `main` now i guess",
"Thanks for reporting, fixing it now",
"It's up again :)"
] | ## Describe the bug
The neat hf spaces app for generating YAML tags for dataset `README.md`s is down
## Steps to reproduce the bug
1. Visit https://huggingface.co/spaces/huggingface/datasets-tagging
## Expected results
There'll be a HF spaces web app for generating dataset metadata YAML
## Actual results
There's an error message; here's the step where it breaks:
```
Step 18/29 : RUN pip install -r requirements.txt
---> Running in e88bfe7e7e0c
Defaulting to user installation because normal site-packages is not writeable
Collecting git+https://github.com/huggingface/datasets.git@update-task-list (from -r requirements.txt (line 4))
Cloning https://github.com/huggingface/datasets.git (to revision update-task-list) to /tmp/pip-req-build-bm8t0r0k
Running command git clone --filter=blob:none --quiet https://github.com/huggingface/datasets.git /tmp/pip-req-build-bm8t0r0k
WARNING: Did not find branch or tag 'update-task-list', assuming revision or ref.
Running command git checkout -q update-task-list
error: pathspec 'update-task-list' did not match any file(s) known to git
error: subprocess-exited-with-error
× git checkout -q update-task-list did not run successfully.
│ exit code: 1
╰─> See above for output.
note: This error originates from a subprocess, and is likely not a problem with pip.
error: subprocess-exited-with-error
× git checkout -q update-task-list did not run successfully.
│ exit code: 1
╰─> See above for output.
```
## Environment info
- Platform: Linux / Brave
| 4,297 |
https://github.com/huggingface/datasets/issues/4291 | Dataset Viewer issue for strombergnlp/ipm_nel : preview is empty, no error message | [
"Hi @leondz, thanks for reporting.\r\n\r\nIndeed, the dataset viewer relies on the dataset being streamable (passing `streaming=True` to `load_dataset`). Whereas most of the datastes are streamable out of the box (thanks to our implementation of streaming), there are still some exceptions.\r\n\r\nIn particular, in your case, that is due to the data file being TAR. This format is not streamable out of the box (it does not allow random access to the archived files), but we use a trick to allow streaming: using `dl_manager.iter_archive`.\r\n\r\nLet me know if you need some help: I could push a commit to your repo with the fix.",
"Ah, right! The preview is working now, but this explanation is good to know, thank you. I'll prefer formats with random file access supported in datasets.utils.extract in future, and try out this fix for the tarfiles :)"
] | ### Link
https://huggingface.co/datasets/strombergnlp/ipm_nel/viewer/ipm_nel/train
### Description
The viewer is blank. I tried my best to emulate a dataset with a working viewer, but this one just doesn't seem to want to come up. What did I miss?
### Owner
Yes | 4,291 |
https://github.com/huggingface/datasets/issues/4287 | "NameError: name 'faiss' is not defined" on `.add_faiss_index` when `device` is not None | [
"So I managed to solve this by adding a missing `import faiss` in the `@staticmethod` defined in https://github.com/huggingface/datasets/blob/f51b6994db27ea69261ef919fb7775928f9ec10b/src/datasets/search.py#L305, triggered from https://github.com/huggingface/datasets/blob/f51b6994db27ea69261ef919fb7775928f9ec10b/src/datasets/search.py#L249 when trying to `ds_with_embeddings.add_faiss_index(column='embeddings', device=0)` with the code above.\r\n\r\nAs it seems that the `@staticmethod` doesn't recognize the `import faiss` defined in https://github.com/huggingface/datasets/blob/f51b6994db27ea69261ef919fb7775928f9ec10b/src/datasets/search.py#L261, so whenever the value of `device` is not None in https://github.com/huggingface/datasets/blob/71f76e0bdeaddadedc4f9c8d15cfff5a36d62f66/src/datasets/search.py#L438, that exception is triggered.\r\n\r\nSo on, adding `import faiss` inside https://github.com/huggingface/datasets/blob/71f76e0bdeaddadedc4f9c8d15cfff5a36d62f66/src/datasets/search.py#L305 right after the check of `device`'s value, solves the issue and lets you calculate the indices in GPU.\r\n\r\nI'll add the code in a PR linked to this issue in case you want to merge it!",
"Adding here the complete error traceback!\r\n\r\n```\r\nTraceback (most recent call last):\r\n File \"/home/alvarobartt/lol.py\", line 12, in <module>\r\n ds_with_embeddings.add_faiss_index(column='embeddings', device=0) # default `device=None`\r\n File \"/home/alvarobartt/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py\", line 3656, in add_faiss_index\r\n super().add_faiss_index(\r\n File \"/home/alvarobartt/.local/lib/python3.9/site-packages/datasets/search.py\", line 478, in add_faiss_index\r\n faiss_index.add_vectors(self, column=column, train_size=train_size, faiss_verbose=True)\r\n File \"/home/alvarobartt/.local/lib/python3.9/site-packages/datasets/search.py\", line 281, in add_vectors\r\n self.faiss_index = self._faiss_index_to_device(index, self.device)\r\n File \"/home/alvarobartt/.local/lib/python3.9/site-packages/datasets/search.py\", line 327, in _faiss_index_to_device\r\n faiss_res = faiss.StandardGpuResources()\r\nNameError: name 'faiss' is not defined\r\n```",
"Closed as https://github.com/huggingface/datasets/pull/4288 already merged! :hugs:"
] | ## Describe the bug
When using `datasets` to calculate the FAISS indices of a dataset, the exception `NameError: name 'faiss' is not defined` is triggered when trying to calculate those on a device (GPU), so `.add_faiss_index(..., device=0)` fails with that exception.
All that assuming that `datasets` is properly installed and `faiss-gpu` too, as well as all the CUDA drivers required.
## Steps to reproduce the bug
```python
# Sample code to reproduce the bug
from transformers import DPRContextEncoder, DPRContextEncoderTokenizer
import torch
torch.set_grad_enabled(False)
ctx_encoder = DPRContextEncoder.from_pretrained("facebook/dpr-ctx_encoder-single-nq-base")
ctx_tokenizer = DPRContextEncoderTokenizer.from_pretrained("facebook/dpr-ctx_encoder-single-nq-base")
from datasets import load_dataset
ds = load_dataset('crime_and_punish', split='train[:100]')
ds_with_embeddings = ds.map(lambda example: {'embeddings': ctx_encoder(**ctx_tokenizer(example["line"], return_tensors="pt"))[0][0].numpy()})
ds_with_embeddings.add_faiss_index(column='embeddings', device=0) # default `device=None`
```
## Expected results
A new column named `embeddings` in the dataset that we're adding the index to.
## Actual results
An exception is triggered with the following message `NameError: name 'faiss' is not defined`.
## Environment info
- `datasets` version: 2.1.0
- Platform: Linux-5.13.0-1022-azure-x86_64-with-glibc2.31
- Python version: 3.9.12
- PyArrow version: 7.0.0
- Pandas version: 1.4.2
| 4,287 |
https://github.com/huggingface/datasets/issues/4284 | Issues in processing very large datasets | [
"Hi ! `datasets` doesn't load the dataset in memory. Instead it uses memory mapping to load your dataset from your disk (it is stored as arrow files). Do you know at what point you have RAM issues exactly ?\r\n\r\nHow big are your graph_data_train dictionaries btw ?",
"Closing this issue due to inactivity."
] | ## Describe the bug
I'm trying to add a feature called "subgraph" to CNN/DM dataset (modifications on run_summarization.py of Huggingface Transformers script) --- I'm not quite sure if I'm doing it the right way, though--- but the main problem appears when the training starts where the error ` [OSError: [Errno 12] Cannot allocate memory]` appears. I suppose this problem roots in RAM issues and how the dataset is loaded during training, but I have no clue of what I can do to fix it. Observing the dataset's cache directory, I see that it takes ~600GB of memory and that's why I believe special care is needed when loading it into the memory.
Here are my modifications to `run_summarization.py` code.
```
# loading pre-computed dictionary where keys are 'id' of article and values are corresponding subgraph
graph_data_train = get_graph_data('train')
graph_data_validation = get_graph_data('val')
...
...
with training_args.main_process_first(desc="train dataset map pre-processing"):
train_dataset = train_dataset.map(
preprocess_function_train,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
desc="Running tokenizer on train dataset",
)
```
And here is the modified preprocessed function:
```
def preprocess_function_train(examples):
inputs, targets, sub_graphs, ids = [], [], [], []
for i in range(len(examples[text_column])):
if examples[text_column][i] is not None and examples[summary_column][i] is not None:
# if examples['doc_id'][i] in graph_data.keys():
inputs.append(examples[text_column][i])
targets.append(examples[summary_column][i])
sub_graphs.append(graph_data_train[examples['id'][i]])
ids.append(examples['id'][i])
inputs = [prefix + inp for inp in inputs]
model_inputs = tokenizer(inputs, max_length=data_args.max_source_length, padding=padding, truncation=True,
sub_graphs=sub_graphs, ids=ids)
# Setup the tokenizer for targets
with tokenizer.as_target_tokenizer():
labels = tokenizer(targets, max_length=max_target_length, padding=padding, truncation=True)
# If we are padding here, replace all tokenizer.pad_token_id in the labels by -100 when we want to ignore
# padding in the loss.
if padding == "max_length" and data_args.ignore_pad_token_for_loss:
labels["input_ids"] = [
[(l if l != tokenizer.pad_token_id else -100) for l in label] for label in labels["input_ids"]
]
model_inputs["labels"] = labels["input_ids"]
return model_inputs
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.1.0
- Platform: Linux Ubuntu
- Python version: 3.6
- PyArrow version: 6.0.1
| 4,284 |
https://github.com/huggingface/datasets/issues/4276 | OpenBookQA has missing and inconsistent field names | [
"Thanks for reporting, @vblagoje.\r\n\r\nIndeed, I noticed some of these issues while reviewing this PR:\r\n- #4259 \r\n\r\nThis is in my TODO list. ",
"Ok, awesome @albertvillanova How about #4275 ?",
"On the other hand, I am not sure if we should always preserve the original nested structure. I think we should also consider other factors as convenience or consistency.\r\n\r\nFor example, other datasets also flatten \"question.stem\" into \"question\":\r\n- ai2_arc:\r\n ```python\r\n question = data[\"question\"][\"stem\"]\r\n choices = data[\"question\"][\"choices\"]\r\n text_choices = [choice[\"text\"] for choice in choices]\r\n label_choices = [choice[\"label\"] for choice in choices]\r\n yield id_, {\r\n \"id\": id_,\r\n \"answerKey\": answerkey,\r\n \"question\": question,\r\n \"choices\": {\"text\": text_choices, \"label\": label_choices},\r\n }\r\n ```\r\n- commonsense_qa:\r\n ```python\r\n question = data[\"question\"]\r\n stem = question[\"stem\"]\r\n yield id_, {\r\n \"answerKey\": answerkey,\r\n \"question\": stem,\r\n \"choices\": {\"label\": labels, \"text\": texts},\r\n }\r\n ```\r\n- cos_e:\r\n ```python\r\n \"question\": cqa[\"question\"][\"stem\"],\r\n ```\r\n- qasc\r\n- quartz\r\n- wiqa\r\n\r\nExceptions:\r\n- exams\r\n\r\nI think we should agree on a CONVENIENT format for QA and use always CONSISTENTLY the same.",
"@albertvillanova I agree that we should be consistent. In the last month, I have come across tons of code that deals with OpenBookQA and CommonSenseQA and all of that code relies on the original data format structure. We can't expect users to adopt HF Datasets if we arbitrarily change the structure of the format just because we think something makes more sense. I am in that position now (downloading original data rather than using HF Datasets) and undoubtedly it hinders HF Datasets' widespread use and adoption. Missing fields like in the case of #4275 is definitely bad and not even up for a discussion IMHO! cc @lhoestq ",
"I'm opening a PR that adds the missing fields.\r\n\r\nLet's agree on the feature structure: @lhoestq @mariosasko @polinaeterna ",
"IMO we should always try to preserve the original structure unless there is a good reason not to (and I don't see one in this case).",
"I agree with @mariosasko . The transition to the original format could be done in one PR for the next minor release, clearly documenting all dataset changes just as @albertvillanova outlined them above and perhaps even providing a per dataset util method to convert the new valid format to the old for backward compatibility. Users who relied on the old format will update their code with either the util method for a quick fix or slightly more elaborate for the new. ",
"I don't have a strong opinion on this, besides the fact that whatever decision we agree on, should be applied to all datasets.\r\n\r\nThere is always the tension between:\r\n- preserving each dataset original structure (which has the advantage of not forcing users to learn other structure for the same dataset),\r\n- and on the other hand performing some kind of standardization/harmonization depending on the task (this has the advantage that once learnt, the same structure applies to all datasets; this has been done for e.g. POS tagging: all datasets have been adapted to a certain \"standard\" structure).\r\n - Another advantage: datasets can easily be interchanged (or joined) to be used by the same model\r\n\r\nRecently, in the BigScience BioMedical hackathon, they adopted a different approach:\r\n- they implement a \"source\" config, respecting the original structure as much as possible\r\n- they implement additional config for each task, with a \"standard\" nested structure per task, which is most useful for users.",
"@albertvillanova, thanks for the detailed answer and the new perspectives. I understand the friction for the best design approach much better now. Ultimately, it is essential to include all the missing fields and the correct data first. Whatever approach is determined to be optimal is important but not as crucial once all the data is there, and users can create lambda functions to create whatever structure serves them best. ",
"Datasets are not tracked in this repository anymore. I think we must move this thread to the [discussions tab of the dataset](https://huggingface.co/datasets/openbookqa/discussions)",
"Indeed @osbm thanks. I'm closing this issue if it's fine for you all then"
] | ## Describe the bug
OpenBookQA implementation is inconsistent with the original dataset.
We need to:
1. The dataset field [question][stem] is flattened into question_stem. Unflatten it to match the original format.
2. Add missing additional fields:
- 'fact1': row['fact1'],
- 'humanScore': row['humanScore'],
- 'clarity': row['clarity'],
- 'turkIdAnonymized': row['turkIdAnonymized']
3. Ensure the structure and every data item in the original OpenBookQA matches our OpenBookQA version.
## Expected results
The structure and every data item in the original OpenBookQA matches our OpenBookQA version.
## Actual results
TBD
## Environment info
- `datasets` version: 2.1.0
- Platform: macOS-10.15.7-x86_64-i386-64bit
- Python version: 3.8.13
- PyArrow version: 7.0.0
- Pandas version: 1.4.2 | 4,276 |
https://github.com/huggingface/datasets/issues/4275 | CommonSenseQA has missing and inconsistent field names | [
"Thanks for reporting, @vblagoje.\r\n\r\nI'm opening a PR to address this. "
] | ## Describe the bug
In short, CommonSenseQA implementation is inconsistent with the original dataset.
More precisely, we need to:
1. Add the dataset matching "id" field. The current dataset, instead, regenerates monotonically increasing id.
2. The [“question”][“stem”] field is flattened into "question". We should match the original dataset and unflatten it
3. Add the missing "question_concept" field in the question tree node
4. Anything else? Go over the data structure of the newly repaired CommonSenseQA and make sure it matches the original
## Expected results
Every data item of the CommonSenseQA should structurally and data-wise match the original CommonSenseQA dataset.
## Actual results
TBD
## Environment info
- `datasets` version: 2.1.0
- Platform: macOS-10.15.7-x86_64-i386-64bit
- Python version: 3.8.13
- PyArrow version: 7.0.0
- Pandas version: 1.4.2 | 4,275 |
https://github.com/huggingface/datasets/issues/4271 | A typo in docs of datasets.disable_progress_bar | [
"Hi! Thanks for catching and reporting the typo, a PR has been opened to fix it :)"
] | ## Describe the bug
in the docs of V2.1.0 datasets.disable_progress_bar, we should replace "enable" with "disable". | 4,271 |
https://github.com/huggingface/datasets/issues/4268 | error downloading bigscience-catalogue-lm-data/lm_en_wiktionary_filtered | [
"It would help a lot to be able to preview the dataset - I'd like to see if the pronunciations are in the dataset, eg. for [\"word\"](https://en.wiktionary.org/wiki/word),\r\n\r\nPronunciation\r\n([Received Pronunciation](https://en.wikipedia.org/wiki/Received_Pronunciation)) [IPA](https://en.wiktionary.org/wiki/Wiktionary:International_Phonetic_Alphabet)([key](https://en.wiktionary.org/wiki/Appendix:English_pronunciation)): /wɜːd/\r\n([General American](https://en.wikipedia.org/wiki/General_American)) [enPR](https://en.wiktionary.org/wiki/Appendix:English_pronunciation): wûrd, [IPA](https://en.wiktionary.org/wiki/Wiktionary:International_Phonetic_Alphabet)([key](https://en.wiktionary.org/wiki/Appendix:English_pronunciation)): /wɝd/",
"Hi @i-am-neo, thanks for reporting.\r\n\r\nNormally this dataset should be private and not accessible for public use. @cakiki, @lvwerra, any reason why is it public? I see many other Wikimedia datasets are also public.\r\n\r\nAlso note that last commit \"Add metadata\" (https://huggingface.co/datasets/bigscience-catalogue-lm-data/lm_en_wiktionary_filtered/commit/dc2f458dab50e00f35c94efb3cd4009996858609) introduced buggy data files (`data/file-01.jsonl.gz.lock`, `data/file-01.jsonl.gz.lock.lock`). The same bug appears in other datasets as well.\r\n\r\n@i-am-neo, please note that in the near future we are planning to make public all datasets used for the BigScience project (at least all of them whose license allows to do that). Once public, they will be accessible for all the NLP community.",
"Ah this must be a bug introduced at creation time since the repos were created programmatically; I'll go ahead and make them private; sorry about that!",
"All datasets are private now. \r\n\r\nRe:that bug I think we're currently avoiding it by avoiding verifications. (i.e. `ignore_verifications=True`)",
"Thanks a lot, @cakiki.\r\n\r\n@i-am-neo, I'm closing this issue for now because the dataset is not publicly available yet. Just stay tuned, as we will soon release all the BigScience open-license datasets. ",
"Thanks for letting me know, @albertvillanova @cakiki.\r\nAny chance of having a subset alpha version in the meantime? \r\nI only need two dicts out of wiktionary: 1) phoneme(as key): word, and 2) word(as key): its phonemes.\r\n\r\nWould like to use it for a mini-poc [Robust ASR](https://github.com/huggingface/transformers/issues/13162#issuecomment-1096881290) decoding, cc @patrickvonplaten. \r\n\r\n(Patrick, possible to email you so as not to litter github with comments? I have some observations after experiments training hubert on some YT AMI-like data (11.44% wer). Also wonder if a robust ASR is on your/HG's roadmap). Thanks!",
"Hey @i-am-neo,\r\n\r\nCool to hear that you're working on Robust ASR! Feel free to drop me a mail :-)",
"@i-am-neo This particular subset of the dataset was taken from the [CirrusSearch dumps](https://dumps.wikimedia.org/other/cirrussearch/current/)\r\nYou're specifically after the [enwiktionary-20220425-cirrussearch-content.json.gz](https://dumps.wikimedia.org/other/cirrussearch/current/enwiktionary-20220425-cirrussearch-content.json.gz) file",
"thanks @cakiki ! <del>I could access the gz file yesterday (but neglected to tuck it away somewhere safe), and today the link is throwing a 404. Can you help? </del> Never mind, got it!",
"thanks @patrickvonplaten. will do - getting my observations together."
] | ## Describe the bug
Error generated when attempting to download dataset
## Steps to reproduce the bug
```python
from datasets import load_dataset
dataset = load_dataset("bigscience-catalogue-lm-data/lm_en_wiktionary_filtered")
```
## Expected results
A clear and concise description of the expected results.
## Actual results
```
ExpectedMoreDownloadedFiles Traceback (most recent call last)
[<ipython-input-62-4ac5cf959477>](https://localhost:8080/#) in <module>()
1 from datasets import load_dataset
2
----> 3 dataset = load_dataset("bigscience-catalogue-lm-data/lm_en_wiktionary_filtered")
3 frames
[/usr/local/lib/python3.7/dist-packages/datasets/utils/info_utils.py](https://localhost:8080/#) in verify_checksums(expected_checksums, recorded_checksums, verification_name)
31 return
32 if len(set(expected_checksums) - set(recorded_checksums)) > 0:
---> 33 raise ExpectedMoreDownloadedFiles(str(set(expected_checksums) - set(recorded_checksums)))
34 if len(set(recorded_checksums) - set(expected_checksums)) > 0:
35 raise UnexpectedDownloadedFile(str(set(recorded_checksums) - set(expected_checksums)))
ExpectedMoreDownloadedFiles: {'/home/leandro/catalogue_data/datasets/lm_en_wiktionary_filtered/data/file-01.jsonl.gz', '/home/leandro/catalogue_data/datasets/lm_en_wiktionary_filtered/data/file-01.jsonl.gz.lock'}
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.18.3
- Platform: Linux-5.4.188+-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.13
- PyArrow version: 6.0.1
| 4,268 |
https://github.com/huggingface/datasets/issues/4261 | data leakage in `webis/conclugen` dataset | [
"Hi @xflashxx, thanks for reporting.\r\n\r\nPlease note that this dataset was generated and shared by Webis Group: https://huggingface.co/webis\r\n\r\nWe are contacting the dataset owners to inform them about the issue you found. We'll keep you updated of their reply.",
"i'd suggest just pinging the authors here in the issue if possible?",
"Thanks for reporting this @xflashxx. I'll have a look and get back to you on this.",
"Hi @xflashxx and @albertvillanova,\r\n\r\nI have updated the files with de-duplicated splits. Apparently the debate portals from which part of the examples were sourced had unique timestamps for some examples (up to 6%; updated counts in the README) without any actual content updated that lead to \"new\" items. The length of `ids_validation` and `ids_testing` is zero.\r\n\r\nRegarding impact on scores:\r\n1. We employed automatic evaluation (on a separate set of 1000 examples) only to justify the exclusion of the smaller models for manual evaluation (due to budget constraints). I am confident the ranking still stands (unsurprisingly, the bigger models doing better than those trained on the smaller splits). We also highlight this in the paper. \r\n\r\n2. The examples used for manual evaluation have no overlap with any splits (also because they do not have any ground truth as we applied the trained models on an unlabeled sample to test its practical usage). I've added these two files to the dataset repository.\r\n\r\nHope this helps!",
"Thanks @shahbazsyed for your fast fix.\r\n\r\nAs a side note:\r\n- Your email appearing as Point of Contact in the dataset README has a typo: @uni.leipzig.de instead of @uni-leipzig.de\r\n- Your commits on the Hub are not linked to your profile on the Hub: this is because we use the email address to make this link; the email address used in your commit author and the email address set on your Hub account settings."
] | ## Describe the bug
Some samples (argument-conclusion pairs) in the *training* split of the `webis/conclugen` dataset are present in both the *validation* and *test* splits, creating data leakage and distorting model results.
Furthermore, all splits contain duplicate samples.
## Steps to reproduce the bug
```python
from datasets import load_dataset
training = load_dataset("webis/conclugen", "base", split="train")
validation = load_dataset("webis/conclugen", "base", split="validation")
testing = load_dataset("webis/conclugen", "base", split="test")
# collect which sample id's are present in the training split
ids_validation = list()
ids_testing = list()
for train_sample in training:
train_argument = train_sample["argument"]
train_conclusion = train_sample["conclusion"]
train_id = train_sample["id"]
# test if current sample is in validation split
if train_argument in validation["argument"]:
for validation_sample in validation:
validation_argument = validation_sample["argument"]
validation_conclusion = validation_sample["conclusion"]
validation_id = validation_sample["id"]
if train_argument == validation_argument and train_conclusion == validation_conclusion:
ids_validation.append(validation_id)
# test if current sample is in test split
if train_argument in testing["argument"]:
for testing_sample in testing:
testing_argument = testing_sample["argument"]
testing_conclusion = testing_sample["conclusion"]
testing_id = testing_sample["id"]
if train_argument == testing_argument and train_conclusion == testing_conclusion:
ids_testing.append(testing_id)
```
## Expected results
Length of both lists `ids_validation` and `ids_testing` should be zero.
## Actual results
Length of `ids_validation` = `2556`
Length of `ids_testing` = `287`
Furthermore, there seems to be duplicate samples in (at least) the *training* split, since:
`print(len(set(ids_validation)))` = `950`
`print(len(set(ids_testing)))` = `101`
All in all, around 7% of the samples of each the *validation* and *test* split seems to be present in the *training* split.
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.18.4
- Platform: macOS-12.3.1-arm64-arm-64bit
- Python version: 3.9.10
- PyArrow version: 7.0.0 | 4,261 |
https://github.com/huggingface/datasets/issues/4248 | conll2003 dataset loads original data. | [
"Thanks for reporting @sue99.\r\n\r\nUnfortunately. I'm not able to reproduce your problem:\r\n```python\r\nIn [1]: import datasets\r\n ...: from datasets import load_dataset\r\n ...: dataset = load_dataset(\"conll2003\")\r\n\r\nIn [2]: dataset\r\nOut[2]: \r\nDatasetDict({\r\n train: Dataset({\r\n features: ['id', 'tokens', 'pos_tags', 'chunk_tags', 'ner_tags'],\r\n num_rows: 14042\r\n })\r\n validation: Dataset({\r\n features: ['id', 'tokens', 'pos_tags', 'chunk_tags', 'ner_tags'],\r\n num_rows: 3251\r\n })\r\n test: Dataset({\r\n features: ['id', 'tokens', 'pos_tags', 'chunk_tags', 'ner_tags'],\r\n num_rows: 3454\r\n })\r\n})\r\n\r\nIn [3]: dataset[\"train\"][0]\r\nOut[3]: \r\n{'id': '0',\r\n 'tokens': ['EU',\r\n 'rejects',\r\n 'German',\r\n 'call',\r\n 'to',\r\n 'boycott',\r\n 'British',\r\n 'lamb',\r\n '.'],\r\n 'pos_tags': [22, 42, 16, 21, 35, 37, 16, 21, 7],\r\n 'chunk_tags': [11, 21, 11, 12, 21, 22, 11, 12, 0],\r\n 'ner_tags': [3, 0, 7, 0, 0, 0, 7, 0, 0]}\r\n```\r\n\r\nJust guessing: might be the case that you are calling `load_dataset` from a working directory that contains a local folder named `conll2003` (containing the raw data files)? If that is the case, `datasets` library gives precedence to the local folder over the dataset on the Hub. "
] | ## Describe the bug
I load `conll2003` dataset to use refined data like [this](https://huggingface.co/datasets/conll2003/viewer/conll2003/train) preview, but it is original data that contains `'-DOCSTART- -X- -X- O'` text.
Is this a bug or should I use another dataset_name like `lhoestq/conll2003` ?
## Steps to reproduce the bug
```python
import datasets
from datasets import load_dataset
dataset = load_dataset("conll2003")
```
## Expected results
{
"chunk_tags": [11, 12, 12, 21, 13, 11, 11, 21, 13, 11, 12, 13, 11, 21, 22, 11, 12, 17, 11, 21, 17, 11, 12, 12, 21, 22, 22, 13, 11, 0],
"id": "0",
"ner_tags": [0, 3, 4, 0, 0, 0, 0, 0, 0, 7, 0, 0, 0, 0, 0, 7, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
"pos_tags": [12, 22, 22, 38, 15, 22, 28, 38, 15, 16, 21, 35, 24, 35, 37, 16, 21, 15, 24, 41, 15, 16, 21, 21, 20, 37, 40, 35, 21, 7],
"tokens": ["The", "European", "Commission", "said", "on", "Thursday", "it", "disagreed", "with", "German", "advice", "to", "consumers", "to", "shun", "British", "lamb", "until", "scientists", "determine", "whether", "mad", "cow", "disease", "can", "be", "transmitted", "to", "sheep", "."]
}
## Actual results
```python
print(dataset)
DatasetDict({
train: Dataset({
features: ['text'],
num_rows: 219554
})
test: Dataset({
features: ['text'],
num_rows: 50350
})
validation: Dataset({
features: ['text'],
num_rows: 55044
})
})
```
```python
for i in range(20):
print(dataset['train'][i])
{'text': '-DOCSTART- -X- -X- O'}
{'text': ''}
{'text': 'EU NNP B-NP B-ORG'}
{'text': 'rejects VBZ B-VP O'}
{'text': 'German JJ B-NP B-MISC'}
{'text': 'call NN I-NP O'}
{'text': 'to TO B-VP O'}
{'text': 'boycott VB I-VP O'}
{'text': 'British JJ B-NP B-MISC'}
{'text': 'lamb NN I-NP O'}
{'text': '. . O O'}
{'text': ''}
{'text': 'Peter NNP B-NP B-PER'}
{'text': 'Blackburn NNP I-NP I-PER'}
{'text': ''}
{'text': 'BRUSSELS NNP B-NP B-LOC'}
{'text': '1996-08-22 CD I-NP O'}
{'text': ''}
{'text': 'The DT B-NP O'}
{'text': 'European NNP I-NP B-ORG'}
```
| 4,248 |
https://github.com/huggingface/datasets/issues/4247 | The data preview of XGLUE | [
"\r\n",
"Thanks for reporting @czq1999.\r\n\r\nNote that the dataset viewer uses the dataset in streaming mode and that not all datasets support streaming yet.\r\n\r\nThat is the case for XGLUE dataset (as the error message points out): this must be refactored to support streaming. ",
"Fixed, thanks @albertvillanova !\r\n\r\nhttps://huggingface.co/datasets/xglue\r\n\r\n<img width=\"824\" alt=\"Capture d’écran 2022-04-29 à 10 23 14\" src=\"https://user-images.githubusercontent.com/1676121/165909391-9f98d98a-665a-4e57-822d-8baa2dc9b7c9.png\">\r\n"
] | It seems that something wrong with the data previvew of XGLUE | 4,247 |
https://github.com/huggingface/datasets/issues/4241 | NonMatchingChecksumError when attempting to download GLUE | [
"Hi :)\r\n\r\nI think your issue may be related to the older `nlp` library. I was able to download `glue` with the latest version of `datasets`. Can you try updating with:\r\n\r\n```py\r\npip install -U datasets\r\n```\r\n\r\nThen you can download:\r\n\r\n```py\r\nfrom datasets import load_dataset\r\nds = load_dataset(\"glue\", \"rte\")\r\n```",
"This appears to work. Thank you!\n\nOn Wed, Apr 27, 2022, 1:18 PM Steven Liu ***@***.***> wrote:\n\n> Hi :)\n>\n> I think your issue may be related to the older nlp library. I was able to\n> download glue with the latest version of datasets. Can you try updating\n> with:\n>\n> pip install -U datasets\n>\n> Then you can download:\n>\n> from datasets import load_datasetds = load_dataset(\"glue\", \"rte\")\n>\n> —\n> Reply to this email directly, view it on GitHub\n> <https://github.com/huggingface/datasets/issues/4241#issuecomment-1111267650>,\n> or unsubscribe\n> <https://github.com/notifications/unsubscribe-auth/ACJUEKLUP2EL7ES3RRWJRPTVHFZHBANCNFSM5UPJBYXA>\n> .\n> You are receiving this because you authored the thread.Message ID:\n> ***@***.***>\n>\n"
] | ## Describe the bug
I am trying to download the GLUE dataset from the NLP module but get an error (see below).
## Steps to reproduce the bug
```python
import nlp
nlp.__version__ # '0.2.0'
nlp.load_dataset('glue', name="rte", download_mode="force_redownload")
```
## Expected results
I expect the dataset to download without an error.
## Actual results
```
INFO:nlp.load:Checking /home/richier/.cache/huggingface/datasets/5fe6ab0df8a32a3371b2e6a969d31d855a19563724fb0d0f163748c270c0ac60.2ea96febf19981fae5f13f0a43d4e2aa58bc619bc23acf06de66675f425a5538.py for additional imports.
INFO:nlp.load:Found main folder for dataset https://s3.amazonaws.com/datasets.huggingface.co/nlp/datasets/glue/glue.py at /home/richier/anaconda3/envs/py36_bert_ee_torch1_11/lib/python3.6/site-packages/nlp/datasets/glue
INFO:nlp.load:Found specific version folder for dataset https://s3.amazonaws.com/datasets.huggingface.co/nlp/datasets/glue/glue.py at /home/richier/anaconda3/envs/py36_bert_ee_torch1_11/lib/python3.6/site-packages/nlp/datasets/glue/637080968c182118f006d3ea39dd9937940e81cfffc8d79836eaae8bba307fc4
INFO:nlp.load:Found script file from https://s3.amazonaws.com/datasets.huggingface.co/nlp/datasets/glue/glue.py to /home/richier/anaconda3/envs/py36_bert_ee_torch1_11/lib/python3.6/site-packages/nlp/datasets/glue/637080968c182118f006d3ea39dd9937940e81cfffc8d79836eaae8bba307fc4/glue.py
INFO:nlp.load:Found dataset infos file from https://s3.amazonaws.com/datasets.huggingface.co/nlp/datasets/glue/dataset_infos.json to /home/richier/anaconda3/envs/py36_bert_ee_torch1_11/lib/python3.6/site-packages/nlp/datasets/glue/637080968c182118f006d3ea39dd9937940e81cfffc8d79836eaae8bba307fc4/dataset_infos.json
INFO:nlp.load:Found metadata file for dataset https://s3.amazonaws.com/datasets.huggingface.co/nlp/datasets/glue/glue.py at /home/richier/anaconda3/envs/py36_bert_ee_torch1_11/lib/python3.6/site-packages/nlp/datasets/glue/637080968c182118f006d3ea39dd9937940e81cfffc8d79836eaae8bba307fc4/glue.json
INFO:nlp.info:Loading Dataset Infos from /home/richier/anaconda3/envs/py36_bert_ee_torch1_11/lib/python3.6/site-packages/nlp/datasets/glue/637080968c182118f006d3ea39dd9937940e81cfffc8d79836eaae8bba307fc4
INFO:nlp.builder:Generating dataset glue (/home/richier/.cache/huggingface/datasets/glue/rte/1.0.0)
INFO:nlp.builder:Dataset not on Hf google storage. Downloading and preparing it from source
INFO:nlp.utils.file_utils:Couldn't get ETag version for url https://firebasestorage.googleapis.com/v0/b/mtl-sentence-representations.appspot.com/o/data%2FRTE.zip?alt=media&token=5efa7e85-a0bb-4f19-8ea2-9e1840f077fb
INFO:nlp.utils.file_utils:https://firebasestorage.googleapis.com/v0/b/mtl-sentence-representations.appspot.com/o/data%2FRTE.zip?alt=media&token=5efa7e85-a0bb-4f19-8ea2-9e1840f077fb not found in cache or force_download set to True, downloading to /home/richier/.cache/huggingface/datasets/downloads/tmpldt3n805
Downloading and preparing dataset glue/rte (download: 680.81 KiB, generated: 1.83 MiB, total: 2.49 MiB) to /home/richier/.cache/huggingface/datasets/glue/rte/1.0.0...
Downloading: 100%|██████████| 73.0/73.0 [00:00<00:00, 73.9kB/s]
INFO:nlp.utils.file_utils:storing https://firebasestorage.googleapis.com/v0/b/mtl-sentence-representations.appspot.com/o/data%2FRTE.zip?alt=media&token=5efa7e85-a0bb-4f19-8ea2-9e1840f077fb in cache at /home/richier/.cache/huggingface/datasets/downloads/e8b62ee44e6f8b6aea761935928579ffe1aa55d161808c482e0725abbdcf9c64
INFO:nlp.utils.file_utils:creating metadata file for /home/richier/.cache/huggingface/datasets/downloads/e8b62ee44e6f8b6aea761935928579ffe1aa55d161808c482e0725abbdcf9c64
---------------------------------------------------------------------------
NonMatchingChecksumError Traceback (most recent call last)
<ipython-input-7-669a8343dcc1> in <module>
----> 1 nlp.load_dataset('glue', name="rte", download_mode="force_redownload")
~/anaconda3/envs/py36_bert_ee_torch1_11/lib/python3.6/site-packages/nlp/load.py in load_dataset(path, name, version, data_dir, data_files, split, cache_dir, download_config, download_mode, ignore_verifications, save_infos, **config_kwargs)
518 download_mode=download_mode,
519 ignore_verifications=ignore_verifications,
--> 520 save_infos=save_infos,
521 )
522
~/anaconda3/envs/py36_bert_ee_torch1_11/lib/python3.6/site-packages/nlp/builder.py in download_and_prepare(self, download_config, download_mode, ignore_verifications, save_infos, try_from_hf_gcs, dl_manager, **download_and_prepare_kwargs)
418 verify_infos = not save_infos and not ignore_verifications
419 self._download_and_prepare(
--> 420 dl_manager=dl_manager, verify_infos=verify_infos, **download_and_prepare_kwargs
421 )
422 # Sync info
~/anaconda3/envs/py36_bert_ee_torch1_11/lib/python3.6/site-packages/nlp/builder.py in _download_and_prepare(self, dl_manager, verify_infos, **prepare_split_kwargs)
458 # Checksums verification
459 if verify_infos:
--> 460 verify_checksums(self.info.download_checksums, dl_manager.get_recorded_sizes_checksums())
461 for split_generator in split_generators:
462 if str(split_generator.split_info.name).lower() == "all":
~/anaconda3/envs/py36_bert_ee_torch1_11/lib/python3.6/site-packages/nlp/utils/info_utils.py in verify_checksums(expected_checksums, recorded_checksums)
34 bad_urls = [url for url in expected_checksums if expected_checksums[url] != recorded_checksums[url]]
35 if len(bad_urls) > 0:
---> 36 raise NonMatchingChecksumError(str(bad_urls))
37 logger.info("All the checksums matched successfully.")
38
NonMatchingChecksumError: ['https://firebasestorage.googleapis.com/v0/b/mtl-sentence-representations.appspot.com/o/data%2FRTE.zip?alt=media&token=5efa7e85-a0bb-4f19-8ea2-9e1840f077fb']
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.0.0
- Platform: Linux-4.18.0-348.20.1.el8_5.x86_64-x86_64-with-redhat-8.5-Ootpa
- Python version: 3.6.13
- PyArrow version: 6.0.1
- Pandas version: 1.1.5
| 4,241 |
https://github.com/huggingface/datasets/issues/4238 | Dataset caching policy | [
"Hi @loretoparisi, thanks for reporting.\r\n\r\nThere is an option to force the redownload of the data files (and thus not using previously download and cached data files): `load_dataset(..., download_mode=\"force_redownload\")`.\r\n\r\nPlease, let me know if this fixes your problem.\r\n\r\nI can confirm you that your dataset loads without any problem for me:\r\n```python\r\nIn [2]: ds = load_dataset(\"loretoparisi/tatoeba-sentences\", data_files={\"train\": \"train.csv\", \"test\": \"test.csv\"}, delimiter=\"\\t\", column_names=['label', 'text'])\r\n\r\nIn [3]: ds\r\nOut[3]: \r\nDatasetDict({\r\n train: Dataset({\r\n features: ['label', 'text'],\r\n num_rows: 8256449\r\n })\r\n test: Dataset({\r\n features: ['label', 'text'],\r\n num_rows: 2061204\r\n })\r\n})\r\n``` ",
"@albertvillanova thank you, it seems it still does not work using:\r\n\r\n```python\r\nsentences = load_dataset(\r\n \"loretoparisi/tatoeba-sentences\",\r\n data_files=data_files,\r\n delimiter='\\t', \r\n column_names=['label', 'text'],\r\n download_mode=\"force_redownload\"\r\n)\r\n```\r\n[This](https://colab.research.google.com/drive/1EA6FWo5pHxU8rPHHRn24NlHqRPiOlPTr?usp=sharing) is my notebook!\r\n\r\nThe problem is that the download file's revision for `test.csv` is not correctly parsed\r\n\r\n\r\n\r\nIf you download that file `test.csv` from the repo, the line `\\\\N` is not there anymore (it was there at the first file upload).\r\n\r\nMy impression is that the Apache Arrow file is still cached - so server side, despite of enabling a forced download. For what I can see I get those two arrow files, but I cannot grep the bad line (`\\\\N`) since are binary files:\r\n\r\n```\r\n!ls -l /root/.cache/huggingface/datasets/csv/loretoparisi--tatoeba-sentences-efeff8965c730a2c/0.0.0/433e0ccc46f9880962cc2b12065189766fbb2bee57a221866138fb9203c83519\r\n!ls -l /root/.cache/huggingface/datasets/csv/loretoparisi--tatoeba-sentences-efeff8965c730a2c/0.0.0/433e0ccc46f9880962cc2b12065189766fbb2bee57a221866138fb9203c83519/csv-test.arrow\r\n!head /root/.cache/huggingface/datasets/csv/loretoparisi--tatoeba-sentences-efeff8965c730a2c/0.0.0/433e0ccc46f9880962cc2b12065189766fbb2bee57a221866138fb9203c83519/dataset_info.json\r\n```\r\n",
"SOLVED! The problem was the with the file itself, using caching parameter helped indeed.\r\nThanks for helping!"
] | ## Describe the bug
I cannot clean cache of my datasets files, despite I have updated the `csv` files on the repository [here](https://huggingface.co/datasets/loretoparisi/tatoeba-sentences). The original file had a line with bad characters, causing the following error
```
[/usr/local/lib/python3.7/dist-packages/datasets/features/features.py](https://localhost:8080/#) in str2int(self, values)
852 if value not in self._str2int:
853 value = str(value).strip()
--> 854 output.append(self._str2int[str(value)])
855 else:
856 # No names provided, try to integerize
KeyError: '\\N'
```
The file now is cleanup up, but I still get the error. This happens even if I inspect the local cached contents, and cleanup the files locally:
```python
from datasets import load_dataset_builder
dataset_builder = load_dataset_builder("loretoparisi/tatoeba-sentences")
print(dataset_builder.cache_dir)
print(dataset_builder.info.features)
print(dataset_builder.info.splits)
```
```
Using custom data configuration loretoparisi--tatoeba-sentences-e59b8ad92f1bb8dd
/root/.cache/huggingface/datasets/csv/loretoparisi--tatoeba-sentences-e59b8ad92f1bb8dd/0.0.0/433e0ccc46f9880962cc2b12065189766fbb2bee57a221866138fb9203c83519
None
None
```
and removing files located at `/root/.cache/huggingface/datasets/csv/loretoparisi--tatoeba-sentences-*`.
Is there any remote file caching policy in place? If so, is it possibile to programmatically disable it?
Currently it seems that the file `test.csv` on the repo [here](https://huggingface.co/datasets/loretoparisi/tatoeba-sentences/blob/main/test.csv) is cached remotely. In fact I download locally the file from raw link, the file is up-to-date; but If I use it within `datasets` as shown above, it gives to me always the first revision of the file, not the last.
Thank you.
## Steps to reproduce the bug
```python
from datasets import load_dataset,Features,Value,ClassLabel
class_names = ["cmn","deu","rus","fra","eng","jpn","spa","ita","kor","vie","nld","epo","por","tur","heb","hun","ell","ind","ara","arz","fin","bul","yue","swe","ukr","bel","que","ces","swh","nno","wuu","nob","zsm","est","kat","pol","lat","urd","sqi","isl","fry","afr","ron","fao","san","bre","tat","yid","uig","uzb","srp","qya","dan","pes","slk","eus","cycl","acm","tgl","lvs","kaz","hye","hin","lit","ben","cat","bos","hrv","tha","orv","cha","mon","lzh","scn","gle","mkd","slv","frm","glg","vol","ain","jbo","tok","ina","nds","mal","tlh","roh","ltz","oss","ido","gla","mlt","sco","ast","jav","oci","ile","ota","xal","tel","sjn","nov","khm","tpi","ang","aze","tgk","tuk","chv","hsb","dsb","bod","sme","cym","mri","ksh","kmr","ewe","kab","ber","tpw","udm","lld","pms","lad","grn","mlg","xho","pnb","grc","hat","lao","npi","cor","nah","avk","mar","guj","pan","kir","myv","prg","sux","crs","ckt","bak","zlm","hil","cbk","chr","nav","lkt","enm","arq","lin","abk","pcd","rom","gsw","tam","zul","awa","wln","amh","bar","hbo","mhr","bho","mrj","ckb","osx","pfl","mgm","sna","mah","hau","kan","nog","sin","glv","dng","kal","liv","vro","apc","jdt","fur","che","haw","yor","crh","pdc","ppl","kin","shs","mnw","tet","sah","kum","ngt","nya","pus","hif","mya","moh","wol","tir","ton","lzz","oar","lug","brx","non","mww","hak","nlv","ngu","bua","aym","vec","ibo","tkl","bam","kha","ceb","lou","fuc","smo","gag","lfn","arg","umb","tyv","kjh","oji","cyo","urh","kzj","pam","srd","lmo","swg","mdf","gil","snd","tso","sot","zza","tsn","pau","som","egl","ady","asm","ori","dtp","cho","max","kam","niu","sag","ilo","kaa","fuv","nch","hoc","iba","gbm","sun","war","mvv","pap","ary","kxi","csb","pag","cos","rif","kek","krc","aii","ban","ssw","tvl","mfe","tah","bvy","bcl","hnj","nau","nst","afb","quc","min","tmw","mad","bjn","mai","cjy","got","hsn","gan","tzl","dws","ldn","afh","sgs","krl","vep","rue","tly","mic","ext","izh","sma","jam","cmo","mwl","kpv","koi","bis","ike","run","evn","ryu","mnc","aoz","otk","kas","aln","akl","yua","shy","fkv","gos","fij","thv","zgh","gcf","cay","xmf","tig","div","lij","rap","hrx","cpi","tts","gaa","tmr","iii","ltg","bzt","syc","emx","gom","chg","osp","stq","frr","fro","nys","toi","new","phn","jpa","rel","drt","chn","pli","laa","bal","hdn","hax","mik","ajp","xqa","pal","crk","mni","lut","ayl","ood","sdh","ofs","nus","kiu","diq","qxq","alt","bfz","klj","mus","srn","guc","lim","zea","shi","mnr","bom","sat","szl"]
features = Features({ 'label': ClassLabel(names=class_names), 'text': Value('string')})
num_labels = features['label'].num_classes
data_files = { "train": "train.csv", "test": "test.csv" }
sentences = load_dataset(
"loretoparisi/tatoeba-sentences",
data_files=data_files,
delimiter='\t',
column_names=['label', 'text'],
)
# You can make this part faster with num_proc=<some int>
sentences = sentences.map(lambda ex: {"label" : features["label"].str2int(ex["label"]) if ex["label"] is not None else None}, features=features)
sentences = sentences.shuffle()
```
## Expected results
Properly tokenize dataset file `test.csv` without issues.
## Actual results
Specify the actual results or traceback.
```
Downloading data files: 100%
2/2 [00:16<00:00, 7.34s/it]
Downloading data: 100%
391M/391M [00:12<00:00, 36.6MB/s]
Downloading data: 100%
92.4M/92.4M [00:02<00:00, 40.0MB/s]
Extracting data files: 100%
2/2 [00:00<00:00, 47.66it/s]
Dataset csv downloaded and prepared to /root/.cache/huggingface/datasets/csv/loretoparisi--tatoeba-sentences-efeff8965c730a2c/0.0.0/433e0ccc46f9880962cc2b12065189766fbb2bee57a221866138fb9203c83519. Subsequent calls will reuse this data.
100%
2/2 [00:00<00:00, 25.94it/s]
11%
942339/8256449 [01:55<13:11, 9245.85ex/s]
---------------------------------------------------------------------------
KeyError Traceback (most recent call last)
[<ipython-input-3-6a9867fad8d6>](https://localhost:8080/#) in <module>()
12 )
13 # You can make this part faster with num_proc=<some int>
---> 14 sentences = sentences.map(lambda ex: {"label" : features["label"].str2int(ex["label"]) if ex["label"] is not None else None}, features=features)
15 sentences = sentences.shuffle()
10 frames
[/usr/local/lib/python3.7/dist-packages/datasets/features/features.py](https://localhost:8080/#) in str2int(self, values)
852 if value not in self._str2int:
853 value = str(value).strip()
--> 854 output.append(self._str2int[str(value)])
855 else:
856 # No names provided, try to integerize
KeyError: '\\N'
```
## Environment info
```
- `datasets` version: 2.1.0
- Platform: Linux-5.4.144+-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.13
- PyArrow version: 6.0.1
- Pandas version: 1.3.5
- ```
```
- `transformers` version: 4.18.0
- Platform: Linux-5.4.144+-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.13
- Huggingface_hub version: 0.5.1
- PyTorch version (GPU?): 1.11.0+cu113 (True)
- Tensorflow version (GPU?): 2.8.0 (True)
- Flax version (CPU?/GPU?/TPU?): not installed (NA)
- Jax version: not installed
- JaxLib version: not installed
- Using GPU in script?: <fill in>
- Using distributed or parallel set-up in script?: <fill in>
- ```
| 4,238 |
https://github.com/huggingface/datasets/issues/4237 | Common Voice 8 doesn't show datasets viewer | [
"Thanks for reporting. I understand it's an error in the dataset script. To reproduce:\r\n\r\n```python\r\n>>> import datasets as ds\r\n>>> split_names = ds.get_dataset_split_names(\"mozilla-foundation/common_voice_8_0\", use_auth_token=\"**********\")\r\nDownloading builder script: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 10.9k/10.9k [00:00<00:00, 10.9MB/s]\r\nDownloading extra modules: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2.98k/2.98k [00:00<00:00, 3.36MB/s]\r\nDownloading extra modules: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 53.1k/53.1k [00:00<00:00, 650kB/s]\r\nNo config specified, defaulting to: common_voice/en\r\nTraceback (most recent call last):\r\n File \"/home/slesage/hf/datasets-preview-backend/libs/libmodels/.venv/lib/python3.9/site-packages/datasets/inspect.py\", line 280, in get_dataset_config_info\r\n for split_generator in builder._split_generators(\r\n File \"/home/slesage/.cache/huggingface/modules/datasets_modules/datasets/mozilla-foundation--common_voice_8_0/720589e6e5ad674019008b719053303a71716db1b27e63c9846df02fdf93f2f3/common_voice_8_0.py\", line 153, in _split_generators\r\n self._log_download(self.config.name, bundle_version, hf_auth_token)\r\n File \"/home/slesage/.cache/huggingface/modules/datasets_modules/datasets/mozilla-foundation--common_voice_8_0/720589e6e5ad674019008b719053303a71716db1b27e63c9846df02fdf93f2f3/common_voice_8_0.py\", line 139, in _log_download\r\n email = HfApi().whoami(auth_token)[\"email\"]\r\nKeyError: 'email'\r\n\r\nThe above exception was the direct cause of the following exception:\r\n\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/home/slesage/hf/datasets-preview-backend/libs/libmodels/.venv/lib/python3.9/site-packages/datasets/inspect.py\", line 323, in get_dataset_split_names\r\n info = get_dataset_config_info(\r\n File \"/home/slesage/hf/datasets-preview-backend/libs/libmodels/.venv/lib/python3.9/site-packages/datasets/inspect.py\", line 285, in get_dataset_config_info\r\n raise SplitsNotFoundError(\"The split names could not be parsed from the dataset config.\") from err\r\ndatasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.\r\n```",
"Thanks for reporting @patrickvonplaten and thanks for the investigation @severo.\r\n\r\nUnfortunately I'm not able to reproduce the error.\r\n\r\nI think the error has to do with authentication with `huggingface_hub`, because the exception is thrown from these code lines: https://huggingface.co/datasets/mozilla-foundation/common_voice_8_0/blob/main/common_voice_8_0.py#L137-L139\r\n```python\r\nfrom huggingface_hub import HfApi, HfFolder\r\n\r\nif isinstance(auth_token, bool):\r\n email = HfApi().whoami(auth_token)\r\nemail = HfApi().whoami(auth_token)[\"email\"]\r\n```\r\n\r\nCould you please verify the previous code with the `auth_token` you pass to `load_dataset(..., use_auth_token=auth_token,...`?",
"OK, thanks for digging a bit into it. Indeed, the error occurs with the dataset-viewer, but not with a normal user token, because we use an app token, and it does not have a related email!\r\n\r\n```python\r\n>>> from huggingface_hub import HfApi, HfFolder\r\n>>> auth_token = \"hf_app_******\"\r\n>>> t = HfApi().whoami(auth_token)\r\n>>> t\r\n{'type': 'app', 'name': 'dataset-preview-backend'}\r\n>>> t[\"email\"]\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\nKeyError: 'email'\r\n```\r\n\r\nNote also that the doc (https://huggingface.co/docs/huggingface_hub/package_reference/hf_api#huggingface_hub.HfApi.whoami) does not state that `whoami` should return an `email` key.\r\n\r\n@SBrandeis @julien-c: do you think the app token should have an email associated, like the users?",
"We can workaround this with\r\n```python\r\nemail = HfApi().whoami(auth_token).get(\"email\", \"[email protected]\")\r\n```\r\nin the common voice scripts",
"Hmmm, does this mean that any person who downloads the common voice dataset will be logged as \"[email protected]\"? If so, it would defeat the purpose of sending the user's email to the commonvoice API, right?",
"I agree with @severo: we cannot set our system email as default, allowing anybody not authenticated to by-pass the Common Voice usage policy.\r\n\r\nAdditionally, looking at the code, I think we should implement a more robust way to send user email to Common Voice: currently anybody can tweak the script and send somebody else email instead.\r\n\r\nCC: @patrickvonplaten @lhoestq @SBrandeis @julien-c ",
"Hmm I don't agree here. \r\n\r\nAnybody can always just bypass the system by setting whatever email. As soon as someone has access to the downloading script it's trivial to tweak the code to not send the \"correct\" email but to just whatever and it would work.\r\n\r\nNote that someone only has visibility on the code after having \"signed\" the access-mechanism so I think we can expect the users to have agreed to not do anything malicious. \r\n\r\nI'm fine with both @lhoestq's solution or we find a way that forces the user to be logged in + being able to load the data for the datasets viewer. Wdyt @lhoestq @severo @albertvillanova ?",
"> Additionally, looking at the code, I think we should implement a more robust way to send user email to Common Voice: currently anybody can tweak the script and send somebody else email instead.\r\n\r\nYes, I agree we can forget about this @patrickvonplaten. After having had a look at Common Voice website, I've seen they only require sending an email (no auth is inplace on their side, contrary to what I had previously thought). Therefore, currently we impose stronger requirements than them: we require the user having logged in and accepted the access mechanism.\r\n\r\nCurrently the script as it is already requires the user being logged in:\r\n```python\r\nHfApi().whoami(auth_token)\r\n```\r\nthrows an exception if None/invalid auth_token is passed.\r\n\r\nOn the other hand, we should agree on the way to allow the viewer to stream the data.",
"The preview is back now, thanks !"
] | https://huggingface.co/datasets/mozilla-foundation/common_voice_8_0 | 4,237 |
https://github.com/huggingface/datasets/issues/4235 | How to load VERY LARGE dataset? | [
"The `Trainer` support `IterableDataset`, not just datasets."
] | ### System Info
```shell
I am using transformer trainer while meeting the issue.
The trainer requests torch.utils.data.Dataset as input, which loads the whole dataset into the memory at once. Therefore, when the dataset is too large to load, there's nothing I can do except using IterDataset, which loads samples of data seperately, and results in low efficiency.
I wonder if there are any tricks like Sharding in huggingface trainer.
Looking forward to your reply.
```
### Who can help?
Trainer: @sgugger
### Information
- [ ] The official example scripts
- [ ] My own modified scripts
### Tasks
- [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...)
- [ ] My own task or dataset (give details below)
### Reproduction
None
### Expected behavior
```shell
I wonder if there are any tricks like fairseq Sharding very large datasets https://fairseq.readthedocs.io/en/latest/getting_started.html.
Thanks a lot!
```
| 4,235 |
https://github.com/huggingface/datasets/issues/4230 | Why the `conll2003` dataset on huggingface only contains the `en` subset? Where is the German data? | [
"Thanks for reporting @beyondguo.\r\n\r\nIndeed, we generate this dataset from this raw data file URL: https://data.deepai.org/conll2003.zip\r\nAnd that URL only contains the English version.",
"The German data requires payment\r\n\r\nThe [original task page](https://www.clips.uantwerpen.be/conll2003/ner/) states \"The German data is a collection of articles from the Frankfurter Rundschau. The named entities have been annotated by people of the University of Antwerp. Only the annotations are available here. In order to build these data sets you need access to the ECI Multilingual Text Corpus. It can be ordered from the Linguistic Data Consortium (2003 non-member price: US$ 35.00).\"\r\n\r\nInflation since 2003 has also affected LDC's prices, and today the dataset [LDC94T5](https://catalog.ldc.upenn.edu/LDC94T5) is available under license for $75 a copy. The [license](https://catalog.ldc.upenn.edu/license/eci-slash-mci-user-agreement.pdf) includes a non-distribution condition, which is probably why the data has not turned up openly.\r\n\r\nThe ACL hold copyright of this data; I'll mail them and anyone I can find at ECI to see if they'll open this up now. After all, it worked with Microsoft 3DMM, why not here too, after 28 years? :)\r\n",
"Closing this issue as we are not allowed to share publicly the German subset."
] | 
But on huggingface datasets:
Where is the German data? | 4,230 |
https://github.com/huggingface/datasets/issues/4221 | Dictionary Feature | [
"Hi @jordiae,\r\n\r\nInstead of the `Sequence` feature, you can use just a regular list: put the dict between `[` and `]`:\r\n```python\r\n\"list_of_dict_feature\": [\r\n {\r\n \"key1_in_dict\": datasets.Value(\"string\"),\r\n \"key2_in_dict\": datasets.Value(\"int32\"),\r\n ...\r\n }\r\n],\r\n```\r\n\r\nFeel free to re-open this issue if that does not work for your use case.",
"> Hi @jordiae,\r\n> \r\n> Instead of the `Sequence` feature, you can use just a regular list: put the dict between `[` and `]`:\r\n> \r\n> ```python\r\n> \"list_of_dict_feature\": [\r\n> {\r\n> \"key1_in_dict\": datasets.Value(\"string\"),\r\n> \"key2_in_dict\": datasets.Value(\"int32\"),\r\n> ...\r\n> }\r\n> ],\r\n> ```\r\n> \r\n> Feel free to re-open this issue if that does not work for your use case.\r\n\r\nThank you"
] | Hi, I'm trying to create the loading script for a dataset in which one feature is a list of dictionaries, which afaik doesn't fit very well the values and structures supported by Value and Sequence. Is there any suggested workaround, am I missing something?
Thank you in advance. | 4,221 |
https://github.com/huggingface/datasets/issues/4217 | Big_Patent dataset broken | [
"Thanks for reporting. The issue seems not to be directly related to the dataset viewer or the `datasets` library, but instead to it being hosted on Google Drive.\r\n\r\nSee related issues: https://github.com/huggingface/datasets/issues?q=is%3Aissue+is%3Aopen+drive.google.com\r\n\r\nTo quote [@lhoestq](https://github.com/huggingface/datasets/issues/4075#issuecomment-1087362551):\r\n\r\n> PS: if possible, please try to not use Google Drive links in your dataset script, since Google Drive has download quotas and is not always reliable.\r\n\r\n",
"We should find out if the dataset license allows redistribution and contact the data owners to propose them to host their data on our Hub.",
"The data owners have agreed on hosting their data on the Hub."
] | ## Dataset viewer issue for '*big_patent*'
**Link:** *[link to the dataset viewer page](https://huggingface.co/datasets/big_patent/viewer/all/train)*
*Unable to view because it says FileNotFound, also cannot download it through the python API*
Am I the one who added this dataset ? No
| 4,217 |
https://github.com/huggingface/datasets/issues/4211 | DatasetDict containing Datasets with different features when pushed to hub gets remapped features | [
"Hi @pietrolesci, thanks for reporting.\r\n\r\nPlease note that this is a design purpose: a `DatasetDict` has the same features for all its datasets. Normally, a `DatasetDict` is composed of several sub-datasets each corresponding to a different **split**.\r\n\r\nTo handle sub-datasets with different features, we use another approach: use different **configurations** instead of **splits**.\r\n\r\nHowever, for the moment `push_to_hub` does not support specifying different configurations. IMHO, we should implement this.",
"Hi @albertvillanova,\r\n\r\nThanks a lot for your reply! I got it now. The strange thing for me was to have it correctly working (i.e., DatasetDict with different features in some datasets) locally and not on the Hub. It would be great to have configuration supported by `push_to_hub`. Personally, this latter functionality allowed me to iterate rather quickly on dataset curation.\r\n\r\nAgain, thanks for your time @albertvillanova!\r\n\r\nBest,\r\nPietro",
"Hi! Yes, we should override `DatasetDict.__setitem__` and throw an error if features dictionaries are different. `DatasetDict` is a subclass of `dict`, so `DatasetDict.{update/setdefault}` need to be overridden as well. We could avoid this by subclassing `UserDict`, but then we would get the name collision - `DatasetDict.data` vs. `UserDict.data`. This makes me think we should rename the `data` attribute of `DatasetDict`/`Dataset` for easier dict subclassing (would also simplify https://github.com/huggingface/datasets/pull/3997) and to follow good Python practices. Another option is to have a custom `UserDict` class in `py_utils`, but it can be hard to keep this class consistent with the built-in `UserDict`. \r\n\r\n@albertvillanova @lhoestq wdyt?",
"I would keep things simple and keep subclassing dict. Regarding the features check, I guess this can be done only for `push_to_hub` right ? It is the only function right now that requires the underlying datasets to be splits (e.g. train/test) and have the same features.\r\n\r\nNote that later you will be able to push datasets with different features as different dataset **configurations** (similarly to the [GLUE subsets](https://huggingface.co/datasets/glue) for example). We will work on this soon",
"Hi @lhoestq,\r\n\r\nReturning to this thread to ask whether the possibility to create `DatasetDict` with different configurations will be supported in the future.\r\n\r\nBest,\r\nPietro",
"DatasetDict is likely to always require the datasets to have the same columns and types, while different configurations may have different columns and types.\r\n\r\nWhy would you like to see that ?\r\nIf it's related to push_to_hub, we plan to allow pushing several configs, but not using DatasetDict",
"Hi @lhoestq and @pietrolesci,\r\n\r\nI have been curious about this question as well. I don't have experience working with different configurations, but I can give a bit more detail on the work flow that I have been using with `Dataset_dict`.\r\n\r\nAs @pietrolesci mentions, I have been using `push_to_hub` to quickly iterate on dataset curation for different ML experiments - locally I create a set of dataset splits e.g. `train/val/test/inference`, then convert them to `HF_Datasets` and finally a to `Dataset_Dict` to `push_to_hub`. Where I have run into issues is when I want to include different metadata for different splits. For example, I have situations where I only have meta-data for one of the splits (e.g. test) or situations where I am working with `inference` data that does not have labels. Currently I use a rather hacky work around by adding \"dummy\" columns for missing columns to avoid the error:\r\n\r\n```\r\nValueError: All datasets in `DatasetDict` should have the same features\r\n```\r\n\r\nI am curious why `DatasetDict` will likely not support this functionality? I don't know much about working with different configurations, but allowing for different columns between datasets / splits would be a very helpful use-case for me. Are there any docs for using different configuration OR a more info about incorporating it with `push_to_hub`.\r\n\r\nBest wishes,\r\nJonathan\r\n\r\n",
"+1",
"> I am curious why DatasetDict will likely not support this functionality?\r\n\r\nThere's a possibility we may merge the Dataset and DatasetDict classes. The DatasetDict purpose was to define a way to get the train/test splits of a dataset.\r\n\r\nsee the discussions at https://github.com/huggingface/datasets/issues/5189\r\n\r\n> Are there any docs for using different configuration OR a more info about incorporating it with push_to_hub.\r\n\r\nThere's a PR open to allow to upload a dataset with a certain configuration name. Then later you can reload this specific configuration using `load_dataset(ds_name, config_name)`\r\n\r\nsee the PR at https://github.com/huggingface/datasets/pull/5213",
"Hi, regarding the following information:\r\n\r\n> Please note that this is a design purpose: a `DatasetDict` has the same features for all its datasets. Normally, a `DatasetDict` is composed of several sub-datasets each corresponding to a different **split**.\r\n> \r\n> To handle sub-datasets with different features, we use another approach: use different **configurations** instead of **splits**.\r\n\r\nAltough this is often implied (such as how else would `DatasetDict` be able to process multiple splits in the same way?), I would expect it to be written somewhere in the docs plainly and maybe even in bold. Also I would expect to see it in multiple places such as:\r\n\r\n- in docstring of `DatasetDict`\r\n- in nlp/image/audio guides on how to create a dataset\r\n- [in conceptual guide on how to create a loading script](https://huggingface.co/docs/datasets/main/en/about_dataset_load)\r\n\r\n\r\nI think this addition would benefit the docs, especially when you guide a newbie (such as me) through the process of creating a dataset. As I said, you somehow suspect that this is in fact the case, but without reading it in the docs you cannot be sure."
] | Hi there,
I am trying to load a dataset to the Hub. This dataset is a `DatasetDict` composed of various splits. Some splits have a different `Feature` mapping. Locally, the DatasetDict preserves the individual features but if I `push_to_hub` and then `load_dataset`, the features are all the same.
Dataset and code to reproduce available [here](https://huggingface.co/datasets/pietrolesci/robust_nli).
In short:
I have 3 feature mapping
```python
Tri_features = Features(
{
"idx": Value(dtype="int64"),
"premise": Value(dtype="string"),
"hypothesis": Value(dtype="string"),
"label": ClassLabel(num_classes=3, names=["entailment", "neutral", "contradiction"]),
}
)
Ent_features = Features(
{
"idx": Value(dtype="int64"),
"premise": Value(dtype="string"),
"hypothesis": Value(dtype="string"),
"label": ClassLabel(num_classes=2, names=["non-entailment", "entailment"]),
}
)
Con_features = Features(
{
"idx": Value(dtype="int64"),
"premise": Value(dtype="string"),
"hypothesis": Value(dtype="string"),
"label": ClassLabel(num_classes=2, names=["non-contradiction", "contradiction"]),
}
)
```
Then I create different datasets
```python
dataset_splits = {}
for split in df["split"].unique():
print(split)
df_split = df.loc[df["split"] == split].copy()
if split in Tri_dataset:
df_split["label"] = df_split["label"].map({"entailment": 0, "neutral": 1, "contradiction": 2})
ds = Dataset.from_pandas(df_split, features=Tri_features)
elif split in Ent_bin_dataset:
df_split["label"] = df_split["label"].map({"non-entailment": 0, "entailment": 1})
ds = Dataset.from_pandas(df_split, features=Ent_features)
elif split in Con_bin_dataset:
df_split["label"] = df_split["label"].map({"non-contradiction": 0, "contradiction": 1})
ds = Dataset.from_pandas(df_split, features=Con_features)
else:
print("ERROR:", split)
dataset_splits[split] = ds
datasets = DatasetDict(dataset_splits)
```
I then push to hub
```python
datasets.push_to_hub("pietrolesci/robust_nli", token="<token>")
```
Finally, I load it from the hub
```python
datasets_loaded_from_hub = load_dataset("pietrolesci/robust_nli")
```
And I get that
```python
datasets["LI_TS"].features != datasets_loaded_from_hub["LI_TS"].features
```
since
```python
"label": ClassLabel(num_classes=2, names=["non-contradiction", "contradiction"])
```
gets remapped to
```python
"label": ClassLabel(num_classes=3, names=["entailment", "neutral", "contradiction"])
``` | 4,211 |
https://github.com/huggingface/datasets/issues/4210 | TypeError: Cannot cast array data from dtype('O') to dtype('int64') according to the rule 'safe' | [
"Hi! Casting class labels from strings is currently not supported in the CSV loader, but you can get the same result with an additional map as follows:\r\n```python\r\nfrom datasets import load_dataset,Features,Value,ClassLabel\r\nclass_names = [\"cmn\",\"deu\",\"rus\",\"fra\",\"eng\",\"jpn\",\"spa\",\"ita\",\"kor\",\"vie\",\"nld\",\"epo\",\"por\",\"tur\",\"heb\",\"hun\",\"ell\",\"ind\",\"ara\",\"arz\",\"fin\",\"bul\",\"yue\",\"swe\",\"ukr\",\"bel\",\"que\",\"ces\",\"swh\",\"nno\",\"wuu\",\"nob\",\"zsm\",\"est\",\"kat\",\"pol\",\"lat\",\"urd\",\"sqi\",\"isl\",\"fry\",\"afr\",\"ron\",\"fao\",\"san\",\"bre\",\"tat\",\"yid\",\"uig\",\"uzb\",\"srp\",\"qya\",\"dan\",\"pes\",\"slk\",\"eus\",\"cycl\",\"acm\",\"tgl\",\"lvs\",\"kaz\",\"hye\",\"hin\",\"lit\",\"ben\",\"cat\",\"bos\",\"hrv\",\"tha\",\"orv\",\"cha\",\"mon\",\"lzh\",\"scn\",\"gle\",\"mkd\",\"slv\",\"frm\",\"glg\",\"vol\",\"ain\",\"jbo\",\"tok\",\"ina\",\"nds\",\"mal\",\"tlh\",\"roh\",\"ltz\",\"oss\",\"ido\",\"gla\",\"mlt\",\"sco\",\"ast\",\"jav\",\"oci\",\"ile\",\"ota\",\"xal\",\"tel\",\"sjn\",\"nov\",\"khm\",\"tpi\",\"ang\",\"aze\",\"tgk\",\"tuk\",\"chv\",\"hsb\",\"dsb\",\"bod\",\"sme\",\"cym\",\"mri\",\"ksh\",\"kmr\",\"ewe\",\"kab\",\"ber\",\"tpw\",\"udm\",\"lld\",\"pms\",\"lad\",\"grn\",\"mlg\",\"xho\",\"pnb\",\"grc\",\"hat\",\"lao\",\"npi\",\"cor\",\"nah\",\"avk\",\"mar\",\"guj\",\"pan\",\"kir\",\"myv\",\"prg\",\"sux\",\"crs\",\"ckt\",\"bak\",\"zlm\",\"hil\",\"cbk\",\"chr\",\"nav\",\"lkt\",\"enm\",\"arq\",\"lin\",\"abk\",\"pcd\",\"rom\",\"gsw\",\"tam\",\"zul\",\"awa\",\"wln\",\"amh\",\"bar\",\"hbo\",\"mhr\",\"bho\",\"mrj\",\"ckb\",\"osx\",\"pfl\",\"mgm\",\"sna\",\"mah\",\"hau\",\"kan\",\"nog\",\"sin\",\"glv\",\"dng\",\"kal\",\"liv\",\"vro\",\"apc\",\"jdt\",\"fur\",\"che\",\"haw\",\"yor\",\"crh\",\"pdc\",\"ppl\",\"kin\",\"shs\",\"mnw\",\"tet\",\"sah\",\"kum\",\"ngt\",\"nya\",\"pus\",\"hif\",\"mya\",\"moh\",\"wol\",\"tir\",\"ton\",\"lzz\",\"oar\",\"lug\",\"brx\",\"non\",\"mww\",\"hak\",\"nlv\",\"ngu\",\"bua\",\"aym\",\"vec\",\"ibo\",\"tkl\",\"bam\",\"kha\",\"ceb\",\"lou\",\"fuc\",\"smo\",\"gag\",\"lfn\",\"arg\",\"umb\",\"tyv\",\"kjh\",\"oji\",\"cyo\",\"urh\",\"kzj\",\"pam\",\"srd\",\"lmo\",\"swg\",\"mdf\",\"gil\",\"snd\",\"tso\",\"sot\",\"zza\",\"tsn\",\"pau\",\"som\",\"egl\",\"ady\",\"asm\",\"ori\",\"dtp\",\"cho\",\"max\",\"kam\",\"niu\",\"sag\",\"ilo\",\"kaa\",\"fuv\",\"nch\",\"hoc\",\"iba\",\"gbm\",\"sun\",\"war\",\"mvv\",\"pap\",\"ary\",\"kxi\",\"csb\",\"pag\",\"cos\",\"rif\",\"kek\",\"krc\",\"aii\",\"ban\",\"ssw\",\"tvl\",\"mfe\",\"tah\",\"bvy\",\"bcl\",\"hnj\",\"nau\",\"nst\",\"afb\",\"quc\",\"min\",\"tmw\",\"mad\",\"bjn\",\"mai\",\"cjy\",\"got\",\"hsn\",\"gan\",\"tzl\",\"dws\",\"ldn\",\"afh\",\"sgs\",\"krl\",\"vep\",\"rue\",\"tly\",\"mic\",\"ext\",\"izh\",\"sma\",\"jam\",\"cmo\",\"mwl\",\"kpv\",\"koi\",\"bis\",\"ike\",\"run\",\"evn\",\"ryu\",\"mnc\",\"aoz\",\"otk\",\"kas\",\"aln\",\"akl\",\"yua\",\"shy\",\"fkv\",\"gos\",\"fij\",\"thv\",\"zgh\",\"gcf\",\"cay\",\"xmf\",\"tig\",\"div\",\"lij\",\"rap\",\"hrx\",\"cpi\",\"tts\",\"gaa\",\"tmr\",\"iii\",\"ltg\",\"bzt\",\"syc\",\"emx\",\"gom\",\"chg\",\"osp\",\"stq\",\"frr\",\"fro\",\"nys\",\"toi\",\"new\",\"phn\",\"jpa\",\"rel\",\"drt\",\"chn\",\"pli\",\"laa\",\"bal\",\"hdn\",\"hax\",\"mik\",\"ajp\",\"xqa\",\"pal\",\"crk\",\"mni\",\"lut\",\"ayl\",\"ood\",\"sdh\",\"ofs\",\"nus\",\"kiu\",\"diq\",\"qxq\",\"alt\",\"bfz\",\"klj\",\"mus\",\"srn\",\"guc\",\"lim\",\"zea\",\"shi\",\"mnr\",\"bom\",\"sat\",\"szl\"]\r\nfeatures = Features({ 'label': ClassLabel(names=class_names), 'text': Value('string')})\r\nnum_labels = features['label'].num_classes\r\ndata_files = { \"train\": \"train.csv\", \"test\": \"test.csv\" }\r\nsentences = load_dataset(\r\n \"loretoparisi/tatoeba-sentences\",\r\n data_files=data_files,\r\n delimiter='\\t', \r\n column_names=['label', 'text'],\r\n)\r\n# You can make this part faster with num_proc=<some int>\r\nsentences = sentences.map(lambda ex: features[\"label\"].str2int(ex[\"label\"]) if ex[\"label\"] is not None else None, features=features)\r\n```\r\n\r\n@lhoestq IIRC, I suggested adding `cast_to_storage` to `ClassLabel` + `table_cast` to the packaged loaders if the `ClassLabel`/`Image`/`Audio` type is present in `features` to avoid this kind of error, but your concern was speed. IMO shouldn't be a problem if we do `table_cast` only when these features are present.",
"I agree packaged loaders should support `ClassLabel` feature without throwing an error.",
"@albertvillanova @mariosasko thank you, with that change now I get\r\n```\r\n---------------------------------------------------------------------------\r\nTypeError Traceback (most recent call last)\r\n[<ipython-input-9-eeb68eeb9bec>](https://localhost:8080/#) in <module>()\r\n 11 )\r\n 12 # You can make this part faster with num_proc=<some int>\r\n---> 13 sentences = sentences.map(lambda ex: features[\"label\"].str2int(ex[\"label\"]) if ex[\"label\"] is not None else None, features=features)\r\n 14 sentences = sentences.shuffle()\r\n\r\n8 frames\r\n[/usr/local/lib/python3.7/dist-packages/datasets/arrow_dataset.py](https://localhost:8080/#) in validate_function_output(processed_inputs, indices)\r\n 2193 if processed_inputs is not None and not isinstance(processed_inputs, (Mapping, pa.Table)):\r\n 2194 raise TypeError(\r\n-> 2195 f\"Provided `function` which is applied to all elements of table returns a variable of type {type(processed_inputs)}. Make sure provided `function` returns a variable of type `dict` (or a pyarrow table) to update the dataset or `None` if you are only interested in side effects.\"\r\n 2196 )\r\n 2197 elif isinstance(indices, list) and isinstance(processed_inputs, Mapping):\r\n\r\nTypeError: Provided `function` which is applied to all elements of table returns a variable of type <class 'int'>. Make sure provided `function` returns a variable of type `dict` (or a pyarrow table) to update the dataset or `None` if you are only interested in side effects.\r\n```\r\n\r\nthe error is raised by [this](https://github.com/huggingface/datasets/blob/master/src/datasets/arrow_dataset.py#L2221)\r\n\r\n```\r\n[/usr/local/lib/python3.7/dist-packages/datasets/arrow_dataset.py](https://localhost:8080/#) in validate_function_output(processed_inputs, indices)\r\n```",
"@mariosasko changed it like\r\n\r\n```python\r\nsentences = sentences.map(lambda ex: {\"label\" : features[\"label\"].str2int(ex[\"label\"]) if ex[\"label\"] is not None else None}, features=features)\r\n```\r\n\r\nto avoid the above errorr.",
"Any update on this? Is this correct ?\r\n> @mariosasko changed it like\r\n> \r\n> ```python\r\n> sentences = sentences.map(lambda ex: {\"label\" : features[\"label\"].str2int(ex[\"label\"]) if ex[\"label\"] is not None else None}, features=features)\r\n> ```\r\n> \r\n> to avoid the above errorr.\r\n\r\n"
] | ### System Info
```shell
- `transformers` version: 4.18.0
- Platform: Linux-5.4.144+-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.13
- Huggingface_hub version: 0.5.1
- PyTorch version (GPU?): 1.10.0+cu111 (True)
- Tensorflow version (GPU?): 2.8.0 (True)
- Flax version (CPU?/GPU?/TPU?): not installed (NA)
- Jax version: not installed
- JaxLib version: not installed
- Using GPU in script?: <fill in>
- Using distributed or parallel set-up in script?: <fill in>
```
### Who can help?
@LysandreJik
### Information
- [ ] The official example scripts
- [X] My own modified scripts
### Tasks
- [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...)
- [X] My own task or dataset (give details below)
### Reproduction
```python
from datasets import load_dataset,Features,Value,ClassLabel
class_names = ["cmn","deu","rus","fra","eng","jpn","spa","ita","kor","vie","nld","epo","por","tur","heb","hun","ell","ind","ara","arz","fin","bul","yue","swe","ukr","bel","que","ces","swh","nno","wuu","nob","zsm","est","kat","pol","lat","urd","sqi","isl","fry","afr","ron","fao","san","bre","tat","yid","uig","uzb","srp","qya","dan","pes","slk","eus","cycl","acm","tgl","lvs","kaz","hye","hin","lit","ben","cat","bos","hrv","tha","orv","cha","mon","lzh","scn","gle","mkd","slv","frm","glg","vol","ain","jbo","tok","ina","nds","mal","tlh","roh","ltz","oss","ido","gla","mlt","sco","ast","jav","oci","ile","ota","xal","tel","sjn","nov","khm","tpi","ang","aze","tgk","tuk","chv","hsb","dsb","bod","sme","cym","mri","ksh","kmr","ewe","kab","ber","tpw","udm","lld","pms","lad","grn","mlg","xho","pnb","grc","hat","lao","npi","cor","nah","avk","mar","guj","pan","kir","myv","prg","sux","crs","ckt","bak","zlm","hil","cbk","chr","nav","lkt","enm","arq","lin","abk","pcd","rom","gsw","tam","zul","awa","wln","amh","bar","hbo","mhr","bho","mrj","ckb","osx","pfl","mgm","sna","mah","hau","kan","nog","sin","glv","dng","kal","liv","vro","apc","jdt","fur","che","haw","yor","crh","pdc","ppl","kin","shs","mnw","tet","sah","kum","ngt","nya","pus","hif","mya","moh","wol","tir","ton","lzz","oar","lug","brx","non","mww","hak","nlv","ngu","bua","aym","vec","ibo","tkl","bam","kha","ceb","lou","fuc","smo","gag","lfn","arg","umb","tyv","kjh","oji","cyo","urh","kzj","pam","srd","lmo","swg","mdf","gil","snd","tso","sot","zza","tsn","pau","som","egl","ady","asm","ori","dtp","cho","max","kam","niu","sag","ilo","kaa","fuv","nch","hoc","iba","gbm","sun","war","mvv","pap","ary","kxi","csb","pag","cos","rif","kek","krc","aii","ban","ssw","tvl","mfe","tah","bvy","bcl","hnj","nau","nst","afb","quc","min","tmw","mad","bjn","mai","cjy","got","hsn","gan","tzl","dws","ldn","afh","sgs","krl","vep","rue","tly","mic","ext","izh","sma","jam","cmo","mwl","kpv","koi","bis","ike","run","evn","ryu","mnc","aoz","otk","kas","aln","akl","yua","shy","fkv","gos","fij","thv","zgh","gcf","cay","xmf","tig","div","lij","rap","hrx","cpi","tts","gaa","tmr","iii","ltg","bzt","syc","emx","gom","chg","osp","stq","frr","fro","nys","toi","new","phn","jpa","rel","drt","chn","pli","laa","bal","hdn","hax","mik","ajp","xqa","pal","crk","mni","lut","ayl","ood","sdh","ofs","nus","kiu","diq","qxq","alt","bfz","klj","mus","srn","guc","lim","zea","shi","mnr","bom","sat","szl"]
features = Features({ 'label': ClassLabel(names=class_names), 'text': Value('string')})
num_labels = features['label'].num_classes
data_files = { "train": "train.csv", "test": "test.csv" }
sentences = load_dataset("loretoparisi/tatoeba-sentences",
data_files=data_files,
delimiter='\t',
column_names=['label', 'text'],
features = features
```
ERROR:
```
ClassLabel(num_classes=403, names=['cmn', 'deu', 'rus', 'fra', 'eng', 'jpn', 'spa', 'ita', 'kor', 'vie', 'nld', 'epo', 'por', 'tur', 'heb', 'hun', 'ell', 'ind', 'ara', 'arz', 'fin', 'bul', 'yue', 'swe', 'ukr', 'bel', 'que', 'ces', 'swh', 'nno', 'wuu', 'nob', 'zsm', 'est', 'kat', 'pol', 'lat', 'urd', 'sqi', 'isl', 'fry', 'afr', 'ron', 'fao', 'san', 'bre', 'tat', 'yid', 'uig', 'uzb', 'srp', 'qya', 'dan', 'pes', 'slk', 'eus', 'cycl', 'acm', 'tgl', 'lvs', 'kaz', 'hye', 'hin', 'lit', 'ben', 'cat', 'bos', 'hrv', 'tha', 'orv', 'cha', 'mon', 'lzh', 'scn', 'gle', 'mkd', 'slv', 'frm', 'glg', 'vol', 'ain', 'jbo', 'tok', 'ina', 'nds', 'mal', 'tlh', 'roh', 'ltz', 'oss', 'ido', 'gla', 'mlt', 'sco', 'ast', 'jav', 'oci', 'ile', 'ota', 'xal', 'tel', 'sjn', 'nov', 'khm', 'tpi', 'ang', 'aze', 'tgk', 'tuk', 'chv', 'hsb', 'dsb', 'bod', 'sme', 'cym', 'mri', 'ksh', 'kmr', 'ewe', 'kab', 'ber', 'tpw', 'udm', 'lld', 'pms', 'lad', 'grn', 'mlg', 'xho', 'pnb', 'grc', 'hat', 'lao', 'npi', 'cor', 'nah', 'avk', 'mar', 'guj', 'pan', 'kir', 'myv', 'prg', 'sux', 'crs', 'ckt', 'bak', 'zlm', 'hil', 'cbk', 'chr', 'nav', 'lkt', 'enm', 'arq', 'lin', 'abk', 'pcd', 'rom', 'gsw', 'tam', 'zul', 'awa', 'wln', 'amh', 'bar', 'hbo', 'mhr', 'bho', 'mrj', 'ckb', 'osx', 'pfl', 'mgm', 'sna', 'mah', 'hau', 'kan', 'nog', 'sin', 'glv', 'dng', 'kal', 'liv', 'vro', 'apc', 'jdt', 'fur', 'che', 'haw', 'yor', 'crh', 'pdc', 'ppl', 'kin', 'shs', 'mnw', 'tet', 'sah', 'kum', 'ngt', 'nya', 'pus', 'hif', 'mya', 'moh', 'wol', 'tir', 'ton', 'lzz', 'oar', 'lug', 'brx', 'non', 'mww', 'hak', 'nlv', 'ngu', 'bua', 'aym', 'vec', 'ibo', 'tkl', 'bam', 'kha', 'ceb', 'lou', 'fuc', 'smo', 'gag', 'lfn', 'arg', 'umb', 'tyv', 'kjh', 'oji', 'cyo', 'urh', 'kzj', 'pam', 'srd', 'lmo', 'swg', 'mdf', 'gil', 'snd', 'tso', 'sot', 'zza', 'tsn', 'pau', 'som', 'egl', 'ady', 'asm', 'ori', 'dtp', 'cho', 'max', 'kam', 'niu', 'sag', 'ilo', 'kaa', 'fuv', 'nch', 'hoc', 'iba', 'gbm', 'sun', 'war', 'mvv', 'pap', 'ary', 'kxi', 'csb', 'pag', 'cos', 'rif', 'kek', 'krc', 'aii', 'ban', 'ssw', 'tvl', 'mfe', 'tah', 'bvy', 'bcl', 'hnj', 'nau', 'nst', 'afb', 'quc', 'min', 'tmw', 'mad', 'bjn', 'mai', 'cjy', 'got', 'hsn', 'gan', 'tzl', 'dws', 'ldn', 'afh', 'sgs', 'krl', 'vep', 'rue', 'tly', 'mic', 'ext', 'izh', 'sma', 'jam', 'cmo', 'mwl', 'kpv', 'koi', 'bis', 'ike', 'run', 'evn', 'ryu', 'mnc', 'aoz', 'otk', 'kas', 'aln', 'akl', 'yua', 'shy', 'fkv', 'gos', 'fij', 'thv', 'zgh', 'gcf', 'cay', 'xmf', 'tig', 'div', 'lij', 'rap', 'hrx', 'cpi', 'tts', 'gaa', 'tmr', 'iii', 'ltg', 'bzt', 'syc', 'emx', 'gom', 'chg', 'osp', 'stq', 'frr', 'fro', 'nys', 'toi', 'new', 'phn', 'jpa', 'rel', 'drt', 'chn', 'pli', 'laa', 'bal', 'hdn', 'hax', 'mik', 'ajp', 'xqa', 'pal', 'crk', 'mni', 'lut', 'ayl', 'ood', 'sdh', 'ofs', 'nus', 'kiu', 'diq', 'qxq', 'alt', 'bfz', 'klj', 'mus', 'srn', 'guc', 'lim', 'zea', 'shi', 'mnr', 'bom', 'sat', 'szl'], id=None)
Value(dtype='string', id=None)
Using custom data configuration loretoparisi--tatoeba-sentences-7b2c5e991f398f39
Downloading and preparing dataset csv/loretoparisi--tatoeba-sentences to /root/.cache/huggingface/datasets/csv/loretoparisi--tatoeba-sentences-7b2c5e991f398f39/0.0.0/433e0ccc46f9880962cc2b12065189766fbb2bee57a221866138fb9203c83519...
Downloading data files: 100%
2/2 [00:18<00:00, 8.06s/it]
Downloading data: 100%
391M/391M [00:13<00:00, 35.3MB/s]
Downloading data: 100%
92.4M/92.4M [00:02<00:00, 36.5MB/s]
Failed to read file '/root/.cache/huggingface/datasets/downloads/933132df9905194ea9faeb30cabca8c49318795612f6495fcb941a290191dd5d' with error <class 'ValueError'>: invalid literal for int() with base 10: 'cmn'
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
/usr/local/lib/python3.7/dist-packages/pandas/_libs/parsers.pyx in pandas._libs.parsers.TextReader._convert_tokens()
TypeError: Cannot cast array data from dtype('O') to dtype('int64') according to the rule 'safe'
During handling of the above exception, another exception occurred:
ValueError Traceback (most recent call last)
15 frames
/usr/local/lib/python3.7/dist-packages/pandas/_libs/parsers.pyx in pandas._libs.parsers.TextReader._convert_tokens()
ValueError: invalid literal for int() with base 10: 'cmn'
```
while loading without `features` it loads without errors
```
sentences = load_dataset("loretoparisi/tatoeba-sentences",
data_files=data_files,
delimiter='\t',
column_names=['label', 'text']
)
```
but the `label` col seems to be wrong (without the `ClassLabel` object):
```
sentences['train'].features
{'label': Value(dtype='string', id=None),
'text': Value(dtype='string', id=None)}
```
The dataset was https://huggingface.co/datasets/loretoparisi/tatoeba-sentences
Dataset format is:
```
ces Nechci vědět, co je tam uvnitř.
ces Kdo o tom chce slyšet?
deu Tom sagte, er fühle sich nicht wohl.
ber Mel-iyi-d anida-t tura ?
hun Gondom lesz rá rögtön.
ber Mel-iyi-d anida-tt tura ?
deu Ich will dich nicht reden hören.
```
### Expected behavior
```shell
correctly load train and test files.
``` | 4,210 |
https://github.com/huggingface/datasets/issues/4199 | Cache miss during reload for datasets using image fetch utilities through map | [
"Hi ! Maybe one of the objects in the function is not deterministic across sessions ? You can read more about it and how to investigate here: https://huggingface.co/docs/datasets/about_cache",
"Hi @apsdehal! Can you verify that replacing\r\n```python\r\ndef fetch_single_image(image_url, timeout=None, retries=0):\r\n for _ in range(retries + 1):\r\n try:\r\n request = urllib.request.Request(\r\n image_url,\r\n data=None,\r\n headers={\"user-agent\": get_datasets_user_agent()},\r\n )\r\n with urllib.request.urlopen(request, timeout=timeout) as req:\r\n image = PIL.Image.open(io.BytesIO(req.read()))\r\n break\r\n except Exception:\r\n image = None\r\n return image\r\n```\r\nwith \r\n```python\r\nUSER_AGENT = get_datasets_user_agent()\r\n\r\ndef fetch_single_image(image_url, timeout=None, retries=0):\r\n for _ in range(retries + 1):\r\n try:\r\n request = urllib.request.Request(\r\n image_url,\r\n data=None,\r\n headers={\"user-agent\": USER_AGENT},\r\n )\r\n with urllib.request.urlopen(request, timeout=timeout) as req:\r\n image = PIL.Image.open(io.BytesIO(req.read()))\r\n break\r\n except Exception:\r\n image = None\r\n return image\r\n```\r\nfixes the issue?",
"Thanks @mariosasko. That does fix the issue. In general, I think these image downloading utilities since they are being used by a lot of image dataset should be provided as a part of `datasets` library right to keep the logic consistent and READMEs smaller? If they already exists, that is also great, please point me to those. I saw that `http_get` does exist.",
"You can find my rationale (and a proposed solution) for why these utilities are not a part of `datasets` here: https://github.com/huggingface/datasets/pull/4100#issuecomment-1097994003.",
"Makes sense. But, I think as the number of image datasets as grow, more people are copying pasting original code from docs to work as it is while we make fixes to them later. I think we do need a central place for these to avoid that confusion as well as more easier access to image datasets. Should we restart that discussion, possible on slack?"
] | ## Describe the bug
It looks like that result of `.map` operation dataset are missing the cache when you reload the script and always run from scratch. In same interpretor session, they are able to find the cache and reload it. But, when you exit the interpretor and reload it, the downloading starts from scratch.
## Steps to reproduce the bug
Using the example provided in `red_caps` dataset.
```python
from concurrent.futures import ThreadPoolExecutor
from functools import partial
import io
import urllib
import PIL.Image
import datasets
from datasets import load_dataset
from datasets.utils.file_utils import get_datasets_user_agent
def fetch_single_image(image_url, timeout=None, retries=0):
for _ in range(retries + 1):
try:
request = urllib.request.Request(
image_url,
data=None,
headers={"user-agent": get_datasets_user_agent()},
)
with urllib.request.urlopen(request, timeout=timeout) as req:
image = PIL.Image.open(io.BytesIO(req.read()))
break
except Exception:
image = None
return image
def fetch_images(batch, num_threads, timeout=None, retries=0):
fetch_single_image_with_args = partial(fetch_single_image, timeout=timeout, retries=retries)
with ThreadPoolExecutor(max_workers=num_threads) as executor:
batch["image"] = list(executor.map(lambda image_urls: [fetch_single_image_with_args(image_url) for image_url in image_urls], batch["image_url"]))
return batch
def process_image_urls(batch):
processed_batch_image_urls = []
for image_url in batch["image_url"]:
processed_example_image_urls = []
image_url_splits = re.findall(r"http\S+", image_url)
for image_url_split in image_url_splits:
if "imgur" in image_url_split and "," in image_url_split:
for image_url_part in image_url_split.split(","):
if not image_url_part:
continue
image_url_part = image_url_part.strip()
root, ext = os.path.splitext(image_url_part)
if not root.startswith("http"):
root = "http://i.imgur.com/" + root
root = root.split("#")[0]
if not ext:
ext = ".jpg"
ext = re.split(r"[?%]", ext)[0]
image_url_part = root + ext
processed_example_image_urls.append(image_url_part)
else:
processed_example_image_urls.append(image_url_split)
processed_batch_image_urls.append(processed_example_image_urls)
batch["image_url"] = processed_batch_image_urls
return batch
dset = load_dataset("red_caps", "jellyfish")
dset = dset.map(process_image_urls, batched=True, num_proc=4)
features = dset["train"].features.copy()
features["image"] = datasets.Sequence(datasets.Image())
num_threads = 5
dset = dset.map(fetch_images, batched=True, batch_size=50, features=features, fn_kwargs={"num_threads": num_threads})
```
Run this in an interpretor or as a script twice and see that the cache is missed the second time.
## Expected results
At reload there should not be any cache miss
## Actual results
Every time script is run, cache is missed and dataset is built from scratch.
## Environment info
- `datasets` version: 2.1.1.dev0
- Platform: Linux-4.19.0-20-cloud-amd64-x86_64-with-glibc2.10
- Python version: 3.8.13
- PyArrow version: 7.0.0
- Pandas version: 1.4.1
| 4,199 |
https://github.com/huggingface/datasets/issues/4198 | There is no dataset | [] | ## Dataset viewer issue for '*name of the dataset*'
**Link:** *link to the dataset viewer page*
*short description of the issue*
Am I the one who added this dataset ? Yes-No
| 4,198 |
https://github.com/huggingface/datasets/issues/4196 | Embed image and audio files in `save_to_disk` | [] | Following https://github.com/huggingface/datasets/pull/4184, currently a dataset saved using `save_to_disk` doesn't actually contain the bytes of the image or audio files. Instead it stores the path to your local files.
Adding `embed_external_files` and set it to True by default to save_to_disk would be kind of a breaking change since some users will get bigger Arrow files when updating the lib, but the advantages are nice:
- the resulting dataset is self contained, in case you want to delete your cache for example or share it with someone else
- users also upload these Arrow files to cloud storage via the fs parameter, and in this case they would expect to upload a self-contained dataset
- consistency with push_to_hub
This can be implemented at the same time as sharding for `save_to_disk` for efficiency, and reuse the helpers from `push_to_hub` to embed the external files.
cc @mariosasko | 4,196 |
https://github.com/huggingface/datasets/issues/4192 | load_dataset can't load local dataset,Unable to find ... | [
"Hi! :)\r\n\r\nI believe that should work unless `dataset_infos.json` isn't actually a dataset. For Hugging Face datasets, there is usually a file named `dataset_infos.json` which contains metadata about the dataset (eg. the dataset citation, license, description, etc). Can you double-check that `dataset_infos.json` isn't just metadata please?",
"Hi @ahf876828330, \r\n\r\nAs @stevhliu pointed out, the proper way to load a dataset is not trying to load its metadata file.\r\n\r\nIn your case, as the dataset script is local, you should better point to your local loading script:\r\n```python\r\ndataset = load_dataset(\"dataset/opus_books.py\")\r\n```\r\n\r\nPlease, feel free to re-open this issue if the previous code snippet does not work for you.",
"> Hi! :)\r\n> \r\n> I believe that should work unless `dataset_infos.json` isn't actually a dataset. For Hugging Face datasets, there is usually a file named `dataset_infos.json` which contains metadata about the dataset (eg. the dataset citation, license, description, etc). Can you double-check that `dataset_infos.json` isn't just metadata please?\r\n\r\nYes，you are right!So if I have a metadata dataset local,How can I turn it to a dataset that can be used by the load_dataset() function？Are there some examples?",
"The metadata file isn't a dataset so you can't turn it into one. You should try @albertvillanova's code snippet above (now merged in the docs [here](https://huggingface.co/docs/datasets/master/en/loading#local-loading-script)), which uses your local loading script `opus_books.py` to:\r\n\r\n1. Download the actual dataset. \r\n2. Once the dataset is downloaded, `load_dataset` will load it for you."
] |
Traceback (most recent call last):
File "/home/gs603/ahf/pretrained/model.py", line 48, in <module>
dataset = load_dataset("json",data_files="dataset/dataset_infos.json")
File "/home/gs603/miniconda3/envs/coderepair/lib/python3.7/site-packages/datasets/load.py", line 1675, in load_dataset
**config_kwargs,
File "/home/gs603/miniconda3/envs/coderepair/lib/python3.7/site-packages/datasets/load.py", line 1496, in load_dataset_builder
data_files=data_files,
File "/home/gs603/miniconda3/envs/coderepair/lib/python3.7/site-packages/datasets/load.py", line 1155, in dataset_module_factory
download_mode=download_mode,
File "/home/gs603/miniconda3/envs/coderepair/lib/python3.7/site-packages/datasets/load.py", line 800, in get_module
data_files = DataFilesDict.from_local_or_remote(patterns, use_auth_token=self.downnload_config.use_auth_token)
File "/home/gs603/miniconda3/envs/coderepair/lib/python3.7/site-packages/datasets/data_files.py", line 582, in from_local_or_remote
if not isinstance(patterns_for_key, DataFilesList)
File "/home/gs603/miniconda3/envs/coderepair/lib/python3.7/site-packages/datasets/data_files.py", line 544, in from_local_or_remote
data_files = resolve_patterns_locally_or_by_urls(base_path, patterns, allowed_extensions)
File "/home/gs603/miniconda3/envs/coderepair/lib/python3.7/site-packages/datasets/data_files.py", line 194, in resolve_patterns_locally_or_by_urls
for path in _resolve_single_pattern_locally(base_path, pattern, allowed_extensions):
File "/home/gs603/miniconda3/envs/coderepair/lib/python3.7/site-packages/datasets/data_files.py", line 144, in _resolve_single_pattern_locally
raise FileNotFoundError(error_msg)
FileNotFoundError: Unable to find '/home/gs603/ahf/pretrained/dataset/dataset_infos.json' at /home/gs603/ahf/pretrained
the code is in the model.py,why I can't use the load_dataset function to load my local dataset? | 4,192 |
https://github.com/huggingface/datasets/issues/4191 | feat: create an `Array3D` column from a list of arrays of dimension 2 | [
"Hi @SaulLu, thanks for your proposal.\r\n\r\nJust I got a bit confused about the dimensions...\r\n- For the 2D case, you mention it is possible to create an `Array2D` from a list of arrays of dimension 1\r\n- However, you give an example of creating an `Array2D` from arrays of dimension 2:\r\n - the values of `data_map` are arrays of dimension 2\r\n - the outer list in `prepare_dataset_2D` should not be taken into account in the dimension counting, as it is used because in `map` you pass `batched=True`\r\n\r\nNote that for the 3D alternatives you mention:\r\n- In `prepare_dataset_3D_ter`, you create an `Array3D` from arrays of dimension 3:\r\n - the array `data_map[index][np.newaxis, :, :]` has dimension 3\r\n - the outer list in `prepare_dataset_3D_ter` is the one used by `batched=True`\r\n- In `prepare_dataset_3D_bis`, you create an `Array3D` from a list of list of lists:\r\n - the value of `data_map[index].tolist()` is a list of lists\r\n - it is enclosed by another list `[data_map[index].tolist()]`, thus giving a list of list of lists\r\n - the outer list is the one used by `batched=True`\r\n\r\nTherefore, if I understand correctly, your request would be to be able to create an `Array3D` from a list of an array of dimension 2:\r\n- In `prepare_dataset_3D`, `data_map[index]` is an array of dimension 2\r\n- it is enclosed by a list `[data_map[index]]`, thus giving a list of an array of dimension 2\r\n- the outer list is the one used by `batched=True`\r\n\r\nPlease, feel free to tell me if I did not understand you correctly.",
"Hi @albertvillanova ,\r\n\r\nIndeed my message was confusing and you guessed right :smile: : I think would be interesting to be able to create an Array3D from a list of an array of dimension 2. \r\n\r\nFor the 2D case I should have given as a \"similar\" example:\r\n```python\r\n\r\ndata_map_1D = {\r\n 1: np.array([0.2, 0.4]),\r\n 2: np.array([0.1, 0.4]),\r\n}\r\n\r\ndef prepare_dataset_2D(batch):\r\n batch[\"pixel_values\"] = [[data_map_1D[index]] for index in batch[\"id\"]]\r\n return batch\r\n \r\nds_2D = ds.map(\r\n prepare_dataset_2D, \r\n batched=True, \r\n remove_columns=ds.column_names, \r\n features=features.Features({\"pixel_values\": features.Array2D(shape=(1, 2), dtype=\"float32\")})\r\n)\r\n```"
] | **Is your feature request related to a problem? Please describe.**
It is possible to create an `Array2D` column from a list of arrays of dimension 1. Similarly, I think it might be nice to be able to create a `Array3D` column from a list of lists of arrays of dimension 1.
To illustrate my proposal, let's take the following toy dataset t:
```python
import numpy as np
from datasets import Dataset, features
data_map = {
1: np.array([[0.2, 0,4],[0.19, 0,3]]),
2: np.array([[0.1, 0,4],[0.19, 0,3]]),
}
def create_toy_ds():
my_dict = {"id":[1, 2]}
return Dataset.from_dict(my_dict)
ds = create_toy_ds()
```
The following 2D processing works without any errors raised:
```python
def prepare_dataset_2D(batch):
batch["pixel_values"] = [data_map[index] for index in batch["id"]]
return batch
ds_2D = ds.map(
prepare_dataset_2D,
batched=True,
remove_columns=ds.column_names,
features=features.Features({"pixel_values": features.Array2D(shape=(2, 3), dtype="float32")})
)
```
The following 3D processing doesn't work:
```python
def prepare_dataset_3D(batch):
batch["pixel_values"] = [[data_map[index]] for index in batch["id"]]
return batch
ds_3D = ds.map(
prepare_dataset_3D,
batched=True,
remove_columns=ds.column_names,
features=features.Features({"pixel_values": features.Array3D(shape=(1, 2, 3, dtype="float32")})
)
```
The error raised is:
```
---------------------------------------------------------------------------
ArrowInvalid Traceback (most recent call last)
[<ipython-input-6-676547e4cd41>](https://localhost:8080/#) in <module>()
3 batched=True,
4 remove_columns=ds.column_names,
----> 5 features=features.Features({"pixel_values": features.Array3D(shape=(1, 2, 3), dtype="float32")})
6 )
12 frames
[/usr/local/lib/python3.7/dist-packages/datasets/arrow_dataset.py](https://localhost:8080/#) in map(self, function, with_indices, with_rank, input_columns, batched, batch_size, drop_last_batch, remove_columns, keep_in_memory, load_from_cache_file, cache_file_name, writer_batch_size, features, disable_nullable, fn_kwargs, num_proc, suffix_template, new_fingerprint, desc)
1971 new_fingerprint=new_fingerprint,
1972 disable_tqdm=disable_tqdm,
-> 1973 desc=desc,
1974 )
1975 else:
[/usr/local/lib/python3.7/dist-packages/datasets/arrow_dataset.py](https://localhost:8080/#) in wrapper(*args, **kwargs)
518 self: "Dataset" = kwargs.pop("self")
519 # apply actual function
--> 520 out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
521 datasets: List["Dataset"] = list(out.values()) if isinstance(out, dict) else [out]
522 for dataset in datasets:
[/usr/local/lib/python3.7/dist-packages/datasets/arrow_dataset.py](https://localhost:8080/#) in wrapper(*args, **kwargs)
485 }
486 # apply actual function
--> 487 out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
488 datasets: List["Dataset"] = list(out.values()) if isinstance(out, dict) else [out]
489 # re-apply format to the output
[/usr/local/lib/python3.7/dist-packages/datasets/fingerprint.py](https://localhost:8080/#) in wrapper(*args, **kwargs)
456 # Call actual function
457
--> 458 out = func(self, *args, **kwargs)
459
460 # Update fingerprint of in-place transforms + update in-place history of transforms
[/usr/local/lib/python3.7/dist-packages/datasets/arrow_dataset.py](https://localhost:8080/#) in _map_single(self, function, with_indices, with_rank, input_columns, batched, batch_size, drop_last_batch, remove_columns, keep_in_memory, load_from_cache_file, cache_file_name, writer_batch_size, features, disable_nullable, fn_kwargs, new_fingerprint, rank, offset, disable_tqdm, desc, cache_only)
2354 writer.write_table(batch)
2355 else:
-> 2356 writer.write_batch(batch)
2357 if update_data and writer is not None:
2358 writer.finalize() # close_stream=bool(buf_writer is None)) # We only close if we are writing in a file
[/usr/local/lib/python3.7/dist-packages/datasets/arrow_writer.py](https://localhost:8080/#) in write_batch(self, batch_examples, writer_batch_size)
505 col_try_type = try_features[col] if try_features is not None and col in try_features else None
506 typed_sequence = OptimizedTypedSequence(batch_examples[col], type=col_type, try_type=col_try_type, col=col)
--> 507 arrays.append(pa.array(typed_sequence))
508 inferred_features[col] = typed_sequence.get_inferred_type()
509 schema = inferred_features.arrow_schema if self.pa_writer is None else self.schema
/usr/local/lib/python3.7/dist-packages/pyarrow/array.pxi in pyarrow.lib.array()
/usr/local/lib/python3.7/dist-packages/pyarrow/array.pxi in pyarrow.lib._handle_arrow_array_protocol()
[/usr/local/lib/python3.7/dist-packages/datasets/arrow_writer.py](https://localhost:8080/#) in __arrow_array__(self, type)
175 storage = list_of_np_array_to_pyarrow_listarray(data, type=pa_type.value_type)
176 else:
--> 177 storage = pa.array(data, pa_type.storage_dtype)
178 return pa.ExtensionArray.from_storage(pa_type, storage)
179
/usr/local/lib/python3.7/dist-packages/pyarrow/array.pxi in pyarrow.lib.array()
/usr/local/lib/python3.7/dist-packages/pyarrow/array.pxi in pyarrow.lib._sequence_to_array()
/usr/local/lib/python3.7/dist-packages/pyarrow/error.pxi in pyarrow.lib.pyarrow_internal_check_status()
/usr/local/lib/python3.7/dist-packages/pyarrow/error.pxi in pyarrow.lib.check_status()
ArrowInvalid: Can only convert 1-dimensional array values
```
**Describe the solution you'd like**
No error in the second scenario and an identical result to the following snippets.
**Describe alternatives you've considered**
There are other alternatives that work such as:
```python
def prepare_dataset_3D_bis(batch):
batch["pixel_values"] = [[data_map[index].tolist()] for index in batch["id"]]
return batch
ds_3D_bis = ds.map(
prepare_dataset_3D_bis,
batched=True,
remove_columns=ds.column_names,
features=features.Features({"pixel_values": features.Array3D(shape=(1, 2, 3), dtype="float32")})
)
```
or
```python
def prepare_dataset_3D_ter(batch):
batch["pixel_values"] = [data_map[index][np.newaxis, :, :] for index in batch["id"]]
return batch
ds_3D_ter = ds.map(
prepare_dataset_3D_ter,
batched=True,
remove_columns=ds.column_names,
features=features.Features({"pixel_values": features.Array3D(shape=(1, 2, 3), dtype="float32")})
)
```
But both solutions require the user to be aware that `data_map[index]` is an `np.array` type.
cc @lhoestq as we discuss this offline :smile: | 4,191 |
https://github.com/huggingface/datasets/issues/4185 | Librispeech documentation, clarification on format | [
"(@patrickvonplaten )",
"Also cc @lhoestq here",
"The documentation in the code is definitely outdated - thanks for letting me know, I'll remove it in https://github.com/huggingface/datasets/pull/4184 .\r\n\r\nYou're exactly right `audio` `array` already decodes the audio file to the correct waveform. This is done on the fly, which is also why one should **not** do `ds[\"audio\"][\"array\"][0]` as this will decode all dataset samples, but instead `ds[0][\"audio\"][\"array\"]` see: https://huggingface.co/docs/datasets/audio_process#audio-datasets\r\n\r\n",
"So, again to clarify: On disk, only the raw flac file content is stored? Is this also the case after `save_to_disk`?\r\n\r\nAnd is it simple to also store it re-encoded as ogg or mp3 instead?\r\n",
"Hey, \r\n\r\nSorry yeah I was just about to look into this! We actually had an outdated version of Librispeech ASR that didn't save any files, but instead converted the audio files to a byte string, then was then decoded on-the-fly. This however is not very user-friendly so we recently decided to instead show the full path of the audio files with the `path` parameter.\r\n\r\nI'm currently changing this for Librispeech here: https://github.com/huggingface/datasets/pull/4184 .\r\nYou should be able to see the audio file in the original `flac` format under `path` then. I don't think it's a good idea to convert to MP3 out-of-the-box, but we could maybe think about some kind of convert function for audio datasets cc @lhoestq ? ",
"> I don't think it's a good idea to convert to MP3 out-of-the-box, but we could maybe think about some kind of convert function for audio datasets cc @lhoestq ?\r\n\r\nSure, I would expect that `load_dataset(\"librispeech_asr\")` would give you the original (not re-encoded) data (flac or already decoded). So such re-encoding logic would be some separate generic function. So I could do sth like `dataset.reencode_as_ogg(**ogg_encode_opts).save_to_disk(...)` or so.\r\n",
"A follow-up question: I wonder whether a Parquet dataset is maybe more what we actually want to have? (Following also my comment here: https://github.com/huggingface/datasets/pull/4184#issuecomment-1105045491.) Because I think we actually would prefer to embed the data content in the dataset.\r\n\r\nSo, instead of `save_to_disk`/`load_from_disk`, we would use `to_parquet`,`from_parquet`? Is there any downside? Are arrow files more efficient?\r\n\r\nRelated is also the doc update in #4193.\r\n",
"`save_to_disk` saves the dataset as an Arrow file, which is the format we use to load a dataset using memory mapping. This way the dataset does not fill your RAM, but is read from your disk instead.\r\n\r\nTherefore you can directly reload a dataset saved with `save_to_disk` using `load_from_disk`.\r\n\r\nParquet files are used for cold storage: to use memory mapping on a Parquet dataset, you first have to convert it to Arrow. We use Parquet to reduce the I/O when pushing/downloading data from the Hugging face Hub. When you load a Parquet file from the Hub, it is converted to Arrow on the fly during the download."
] | https://github.com/huggingface/datasets/blob/cd3ce34ab1604118351e1978d26402de57188901/datasets/librispeech_asr/librispeech_asr.py#L53
> Note that in order to limit the required storage for preparing this dataset, the audio
> is stored in the .flac format and is not converted to a float32 array. To convert, the audio
> file to a float32 array, please make use of the `.map()` function as follows:
>
> ```python
> import soundfile as sf
> def map_to_array(batch):
> speech_array, _ = sf.read(batch["file"])
> batch["speech"] = speech_array
> return batch
> dataset = dataset.map(map_to_array, remove_columns=["file"])
> ```
Is this still true?
In my case, `ds["train.100"]` returns:
```
Dataset({
features: ['file', 'audio', 'text', 'speaker_id', 'chapter_id', 'id'],
num_rows: 28539
})
```
and taking the first instance yields:
```
{'file': '374-180298-0000.flac',
'audio': {'path': '374-180298-0000.flac',
'array': array([ 7.01904297e-04, 7.32421875e-04, 7.32421875e-04, ...,
-2.74658203e-04, -1.83105469e-04, -3.05175781e-05]),
'sampling_rate': 16000},
'text': 'CHAPTER SIXTEEN I MIGHT HAVE TOLD YOU OF THE BEGINNING OF THIS LIAISON IN A FEW LINES BUT I WANTED YOU TO SEE EVERY STEP BY WHICH WE CAME I TO AGREE TO WHATEVER MARGUERITE WISHED',
'speaker_id': 374,
'chapter_id': 180298,
'id': '374-180298-0000'}
```
The `audio` `array` seems to be already decoded. So such convert/decode code as mentioned in the doc is wrong?
But I wonder, is it actually stored as flac on disk, and the decoding is done on-the-fly? Or was it decoded already during the preparation and is stored as raw samples on disk?
Note that I also used `datasets.load_dataset("librispeech_asr", "clean").save_to_disk(...)` and then `datasets.load_from_disk(...)` in this example. Does this change anything on how it is stored on disk?
A small related question: Actually I would prefer to even store it as mp3 or ogg on disk. Is this easy to convert? | 4,185 |
https://github.com/huggingface/datasets/issues/4182 | Zenodo.org download is not responding | [
"[Off topic but related: Is the uptime of S3 provably better than Zenodo's?]",
"Hi @dkajtoch, please note that at HuggingFace we are not hosting this dataset: we are just using a script to download their data file and create a dataset from it.\r\n\r\nIt was the dataset owners decision to host their data at Zenodo. You can see this on their website: https://marcobaroni.org/composes/sick.html\r\n\r\nAnd yes, you are right: Zenodo is currently having some incidents and people are reporting problems from it.\r\n\r\nOn the other hand, we could contact the data owners and propose them to host their data at our Hugging Face Hub.\r\n\r\n@julien-c I guess so.\r\n",
"Thanks @albertvillanova. I know that the problem lies in the source data. I just wanted to point out that these kind of problems are unavoidable without having one place where data sources are cached. Websites may go down or data sources may move. Having a copy in Hugging Face Hub would be a great solution. ",
"Definitely, @dkajtoch! But we have to ask permission to the data owners. And many dataset licenses directly forbid data redistribution: in those cases we are not allowed to host their data on our Hub.",
"Ahhh good point! License is the problem :("
] | ## Describe the bug
Source download_url from zenodo.org does not respond.
`_DOWNLOAD_URL = "https://zenodo.org/record/2787612/files/SICK.zip?download=1"`
Other datasets also use zenodo.org to store data and they cannot be downloaded as well.
It would be better to actually use more reliable way to store original data like s3 bucket.
## Steps to reproduce the bug
```python
load_dataset("sick")
```
## Expected results
Dataset should be downloaded.
## Actual results
ConnectionError: Couldn't reach https://zenodo.org/record/2787612/files/SICK.zip?download=1 (ReadTimeout(ReadTimeoutError("HTTPSConnectionPool(host='zenodo.org', port=443): Read timed out. (read timeout=100)")))
## Environment info
- `datasets` version: 2.1.0
- Platform: Darwin-21.4.0-x86_64-i386-64bit
- Python version: 3.7.11
- PyArrow version: 7.0.0
- Pandas version: 1.3.5
| 4,182 |
https://github.com/huggingface/datasets/issues/4181 | Support streaming FLEURS dataset | [
"Yes, you just have to use `dl_manager.iter_archive` instead of `dl_manager.download_and_extract`.\r\n\r\nThat's because `download_and_extract` doesn't support TAR archives in streaming mode.",
"Tried to make it streamable, but I don't think it's really possible. @lhoestq @polinaeterna maybe you guys can check: \r\nhttps://huggingface.co/datasets/google/fleurs/commit/dcf80160cd77977490a8d32b370c027107f2407b \r\n\r\nreal quick. \r\n\r\nI think the problem is that we cannot ensure that the metadata file is found before the audio. Or is this possible somehow @lhoestq ? ",
"@patrickvonplaten I think the metadata file should be found first because the audio files are contained in a folder next to the metadata files (just as in common voice), so the metadata files should be \"on top of the list\" as they are closer to the root in the directories hierarchy ",
"@patrickvonplaten but apparently it doesn't... I don't really know why.",
"Yeah! Any ideas what could be the reason here? cc @lhoestq ?",
"The order of the files is determined when the TAR archive is created, depending on the commands the creator ran.\r\nIf the metadata file is not at the beginning of the file, that makes streaming completely inefficient. In this case the TAR archive needs to be recreated in an appropriate order.",
"Actually we could maybe just host the metadata file ourselves and then stream the audio data only. Don't think that this would be a problem for the FLEURS authors (I can ask them :-)) ",
"I made a PR to their repo to support streaming (by uploading the metadata file to the Hub). See:\r\n- https://huggingface.co/datasets/google/fleurs/discussions/4",
"I'm closing this issue as the PR above has been merged."
] | ## Dataset viewer issue for '*name of the dataset*'
https://huggingface.co/datasets/google/fleurs
```
Status code: 400
Exception: NotImplementedError
Message: Extraction protocol for TAR archives like 'https://storage.googleapis.com/xtreme_translations/FLEURS/af_za.tar.gz' is not implemented in streaming mode. Please use `dl_manager.iter_archive` instead.
```
Am I the one who added this dataset ? Yes
Can I fix this somehow in the script? @lhoestq @severo
| 4,181 |
https://github.com/huggingface/datasets/issues/4180 | Add some iteration method on a dataset column (specific for inference) | [
"Thanks for the suggestion ! I agree it would be nice to have something directly in `datasets` to do something as simple as that\r\n\r\ncc @albertvillanova @mariosasko @polinaeterna What do you think if we have something similar to pandas `Series` that wouldn't bring everything in memory when doing `dataset[\"audio\"]` ? Currently it returns a list with all the decoded audio data in memory.\r\n\r\nIt would be a breaking change though, since `isinstance(dataset[\"audio\"], list)` wouldn't work anymore, but we could implement a `Sequence` so that `dataset[\"audio\"][0]` still works and only loads one item in memory.\r\n\r\nYour alternative suggestion with `iterate` is also sensible, though maybe less satisfactory in terms of experience IMO",
"I agree that current behavior (decoding all audio file sin the dataset when accessing `dataset[\"audio\"]`) is not useful, IMHO. Indeed in our docs, we are constantly warning our collaborators not to do that.\r\n\r\nTherefore I upvote for a \"useful\" behavior of `dataset[\"audio\"]`. I don't think the breaking change is important in this case, as I guess no many people use it with its current behavior. Therefore, for me it seems reasonable to return a generator (instead of an in-memeory list) for \"special\" features, like Audio/Image.\r\n\r\n@lhoestq on the other hand I don't understand your proposal about Pandas-like... ",
"I recall I had the same idea while working on the `Image` feature, so I agree implementing something similar to `pd.Series` that lazily brings elements in memory would be beneficial.",
"@lhoestq @mariosasko Could you please give a link to that new feature of `pandas.Series`? As far as I remember since I worked with pandas for more than 6 years, there was no lazy in-memory feature; it was everything in-memory; that was the reason why other frameworks were created, like Vaex or Dask, e.g. ",
"Yea pandas doesn't do lazy loading. I was referring to pandas.Series to say that they have a dedicated class to represent a column ;)"
] | **Is your feature request related to a problem? Please describe.**
A clear and concise description of what the problem is.
Currently, `dataset["audio"]` will load EVERY element in the dataset in RAM, which can be quite big for an audio dataset.
Having an iterator (or sequence) type of object, would make inference with `transformers` 's `pipeline` easier to use and not so memory hungry.
**Describe the solution you'd like**
A clear and concise description of what you want to happen.
For a non breaking change:
```python
for audio in dataset.iterate("audio"):
# {"array": np.array(...), "sampling_rate":...}
```
For a breaking change solution (not necessary), changing the type of `dataset["audio"]` to a sequence type so that
```python
pipe = pipeline(model="...")
for out in pipe(dataset["audio"]):
# {"text":....}
```
could work
**Describe alternatives you've considered**
A clear and concise description of any alternative solutions or features you've considered.
```python
def iterate(dataset, key):
for item in dataset:
yield dataset[key]
for out in pipeline(iterate(dataset, "audio")):
# {"array": ...}
```
This works but requires the helper function which feels slightly clunky.
**Additional context**
Add any other context about the feature request here.
The context is actually to showcase better integration between `pipeline` and `datasets` in the Quicktour demo: https://github.com/huggingface/transformers/pull/16723/files
@lhoestq
| 4,180 |
https://github.com/huggingface/datasets/issues/4179 | Dataset librispeech_asr fails to load | [
"@patrickvonplaten Hi! I saw that you prepared this? :)",
"Another thing, but maybe this should be a separate issue: As I see from the code, it would try to use up to 16 simultaneous downloads? This is problematic for Librispeech or anything on OpenSLR. On [the homepage](https://www.openslr.org/), it says:\r\n\r\n> If you want to download things from this site, please download them one at a time, and please don't use any fancy software-- just download things from your browser or use 'wget'. We have a firewall rule to drop connections from hosts with more than 5 simultaneous connections, and certain types of download software may activate this rule.\r\n\r\nRelated: https://github.com/tensorflow/datasets/issues/3885",
"Hey @albertz,\r\n\r\nNice to see you here! It's been a while ;-) ",
"Sorry maybe the docs haven't been super clear here. By `split` we mean one of `train.500`, `train.360`, `train.100`, `validation`, `test`. For Librispeech, you'll have to specific a config (either `other` or `clean`) though:\r\n\r\n```py\r\ndatasets.load_dataset(\"librispeech_asr\", \"clean\")\r\n```\r\n\r\nshould work and give you all splits (being \"train\", \"test\", ...) for the clean config of the dataset.\r\n",
"If you need both `\"clean\"` and `\"other\"` I think you'll have to do concatenate them as follows: \r\n\r\n```py\r\nfrom datasets import concatenate_datasets, load_dataset\r\n\r\nother = load_dataset(\"librispeech_asr\", \"other\")\r\nclean = load_dataset(\"librispeech_asr\", \"clean\")\r\n\r\nlibrispeech = concatenate_datasets([other, clean])\r\n```\r\n\r\nSee https://huggingface.co/docs/datasets/v2.1.0/en/process#concatenate",
"Downloading one split would be:\r\n\r\n```py\r\nfrom datasets import load_dataset\r\n\r\nother = load_dataset(\"librispeech_asr\", \"other\", split=\"train.500\")\r\n```\r\n\r\n\r\n",
"cc @lhoestq FYI maybe the docs can be improved here",
"Ah thanks. But wouldn't it be easier/nicer (and more canonical) to just make it in a way that simply `load_dataset(\"librispeech_asr\")` works?",
"Pinging @lhoestq here, think this could make sense! Not sure however how the dictionary would then look like",
"Would it make sense to have `clean` as the default config ?\r\n\r\nAlso I think `load_dataset(\"librispeech_asr\")` should have raised you an error that says that you need to specify a config\r\n\r\nI also opened a PR to improve the doc: https://github.com/huggingface/datasets/pull/4183",
"> Would it make sense to have `clean` as the default config ?\r\n\r\nI think a user would expect that the default would give you the full dataset.\r\n\r\n> Also I think `load_dataset(\"librispeech_asr\")` should have raised you an error that says that you need to specify a config\r\n\r\nIt does raise an error, but this error confused me because I did not understand why I needed a config, or why I could not simply download the whole dataset, which is what people usually do with Librispeech.\r\n",
"+1 for @albertz. Also think lots of people download the whole dataset (`\"clean\"` + `\"other\"`) for Librispeech.\r\n\r\nThink there are also some people though who:\r\n- a) Don't have the memory to store the whole dataset\r\n- b) Just want to evaluate on one of the two configs",
"Ok ! Adding the \"all\" configuration would do the job then, thanks ! In the \"all\" configuration we can merge all the train.xxx splits into one \"train\" split, or keep them separate depending on what's the most practical to use (probably put everything in \"train\" no ?)",
"I'm not too familiar with how to work with HuggingFace datasets, but people often do some curriculum learning scheme, where they start with train.100, later go over to train.100 + train.360, and then later use the whole train (960h). It would be good if this is easily possible.\r\n",
"Hey @albertz, \r\n\r\nopened a PR here. Think by adding the \"subdataset\" class to each split \"train\", \"dev\", \"other\" as shown here: https://github.com/huggingface/datasets/pull/4184/files#r853272727 it should be easily possible (e.g. with the filter function https://huggingface.co/docs/datasets/v2.1.0/en/package_reference/main_classes#datasets.Dataset.filter )",
"But also since everything is cached one could also just do:\r\n\r\n```python\r\nload_dataset(\"librispeech\", \"clean\", \"train.100\")\r\nload_dataset(\"librispeech\", \"clean\", \"train.100+train.360\")\r\nload_dataset(\"librispeech\" \"all\", \"train\") \r\n```",
"Hi @patrickvonplaten ,\r\n\r\nload_dataset(\"librispeech_asr\", \"clean\", \"train.100\") actually downloads the whole dataset and not the 100 hr split, is this a bug?",
"Hmm, I don't really see how that's possible: https://github.com/huggingface/datasets/blob/d22e39a0693d4be7410cf9a5d41fd5aac22be3cc/datasets/librispeech_asr/librispeech_asr.py#L51\r\n\r\nNote that all datasets related to `\"clean\"` are downloaded, but only `\"train.100\"` should be used. \r\n\r\ncc @lhoestq @albertvillanova @mariosasko can we do anything against download dataset links that are not related to the \"split\" that one actually needs. E.g. why should the split `\"train.360\"` be downloaded if for the user executes the above command:\r\n\r\n```py\r\nload_dataset(\"librispeech_asr\", \"clean\", \"train.100\")\r\n```",
"@patrickvonplaten This problem is a bit harder than it may seem, and it has to do with how our scripts are structured - `_split_generators` downloads data for a split before its definition. There was an attempt to fix this in https://github.com/huggingface/datasets/pull/2249, but it wasn't flexible enough. Luckily, I have a plan of attack, and this issue is on our short-term roadmap, so I'll work on it soon.\r\n\r\nIn the meantime, one can use streaming or manually download a dataset script, remove unwanted splits and load a dataset via `load_dataset`.",
"> load_dataset(\"librispeech_asr\", \"clean\", \"train.100\") actually downloads the whole dataset and not the 100 hr split, is this a bug?\r\n\r\nSince this bug is still there and google led me here when I was searching for a solution, I am writing down how to quickly fix it (as suggested by @mariosasko) for whoever else is not familiar with how the HF Hub works.\r\n\r\nDownload the [librispeech_asr.py](https://huggingface.co/datasets/librispeech_asr/blob/main/librispeech_asr.py) script and remove the unwanted splits both from the [`_DL_URLS` dictionary](https://huggingface.co/datasets/librispeech_asr/blob/main/librispeech_asr.py#L47-L68) and from the [`_split_generators` function](https://huggingface.co/datasets/librispeech_asr/blob/main/librispeech_asr.py#L121-L241).\r\n[Here ](https://huggingface.co/datasets/andreagasparini/librispeech_test_only) I made an example with only the test sets.\r\n\r\nThen either save the script locally and load the dataset via \r\n```python\r\nload_dataset(\"${local_path}/librispeech_asr.py\")\r\n```\r\n\r\nor [create a new dataset repo on the hub](https://huggingface.co/new-dataset) named \"librispeech_asr\" and upload the script there, then you can just run\r\n```python\r\nload_dataset(\"${hugging_face_username}/librispeech_asr\")\r\n```",
"Fixed by https://github.com/huggingface/datasets/pull/4184"
] | ## Describe the bug
The dataset librispeech_asr (standard Librispeech) fails to load.
## Steps to reproduce the bug
```python
datasets.load_dataset("librispeech_asr")
```
## Expected results
It should download and prepare the whole dataset (all subsets).
In [the doc](https://huggingface.co/datasets/librispeech_asr), it says it has two configurations (clean and other).
However, the dataset doc says that not specifying `split` should just load the whole dataset, which is what I want.
Also, in case of this specific dataset, this is also the standard what the community uses. When you look at any publications with results on Librispeech, they always use the whole train dataset for training.
## Actual results
```
...
File "/home/az/.cache/huggingface/modules/datasets_modules/datasets/librispeech_asr/1f4602f6b5fed8d3ab3e3382783173f2e12d9877e98775e34d7780881175096c/librispeech_asr.py", line 119, in LibrispeechASR._split_generators
line: archive_path = dl_manager.download(_DL_URLS[self.config.name])
locals:
archive_path = <not found>
dl_manager = <local> <datasets.utils.download_manager.DownloadManager object at 0x7fc07b426160>
dl_manager.download = <local> <bound method DownloadManager.download of <datasets.utils.download_manager.DownloadManager object at 0x7fc07b426160>>
_DL_URLS = <global> {'clean': {'dev': 'http://www.openslr.org/resources/12/dev-clean.tar.gz', 'test': 'http://www.openslr.org/resources/12/test-clean.tar.gz', 'train.100': 'http://www.openslr.org/resources/12/train-clean-100.tar.gz', 'train.360': 'http://www.openslr.org/resources/12/train-clean-360.tar.gz'}, 'other'...
self = <local> <datasets_modules.datasets.librispeech_asr.1f4602f6b5fed8d3ab3e3382783173f2e12d9877e98775e34d7780881175096c.librispeech_asr.LibrispeechASR object at 0x7fc12a633310>
self.config = <local> BuilderConfig(name='default', version=0.0.0, data_dir='/home/az/i6/setups/2022-03-20--sis/work/i6_core/datasets/huggingface/DownloadAndPrepareHuggingFaceDatasetJob.TV6Nwm6dFReF/output/data_dir', data_files=None, description=None)
self.config.name = <local> 'default', len = 7
KeyError: 'default'
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.1.0
- Platform: Linux-5.4.0-107-generic-x86_64-with-glibc2.31
- Python version: 3.9.9
- PyArrow version: 6.0.1
- Pandas version: 1.4.2
| 4,179 |
https://github.com/huggingface/datasets/issues/4176 | Very slow between two operations | [] | Hello, in the processing stage, I use two operations. The first one : map + filter, is very fast and it uses the full cores, while the socond step is very slow and did not use full cores.
Also, there is a significant lag between them. Am I missing something ?
```
raw_datasets = raw_datasets.map(split_func,
batched=False,
num_proc=args.preprocessing_num_workers,
load_from_cache_file=not args.overwrite_cache,
desc = "running split para ==>")\
.filter(lambda example: example['text1']!='' and example['text2']!='',
num_proc=args.preprocessing_num_workers, desc="filtering ==>")
processed_datasets = raw_datasets.map(
preprocess_function,
batched=True,
num_proc=args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not args.overwrite_cache,
desc="Running tokenizer on dataset===>",
)
``` | 4,176 |
https://github.com/huggingface/datasets/issues/4169 | Timit_asr dataset cannot be previewed recently | [
"Thanks for reporting. The bug has already been detected, and we hope to fix it soon.",
"TIMIT is now a dataset that requires manual download, see #4145 \r\n\r\nTherefore it might take a bit more time to fix it",
"> TIMIT is now a dataset that requires manual download, see #4145\r\n> \r\n> Therefore it might take a bit more time to fix it\r\n\r\nThank you for your quickly response. Exactly, I also found the manual download issue in the morning. But when I used *list_datasets()* to check the available datasets, *'timit_asr'* is still in the list. So I am a little bit confused. If *'timit_asr'* need to be manually downloaded, does that mean we can **not** automatically download it **any more** in the future?",
"Yes exactly. If you try to load the dataset it will ask you to download it manually first, and to pass the downloaded and extracted data like `load_dataset(\"timir_asr\", data_dir=\"path/to/extracted/data\")`\r\n\r\nThe URL we were using was coming from a host that doesn't have the permission to redistribute the data, and the dataset owners (LDC) notified us about it.",
"I downloaded the timit_asr data and unzipped. But I can't run my code. Could you resolve this problem for me? Thanks\r\n\r\n import soundfile as sf\r\n import torch\r\n from datasets import load_dataset\r\n dataset = load_dataset(\"timit_asr\", data_dir=\"/Users/nguyenvannham/Documents/test_case/data\")\r\n \r\n \r\n Generating train split: 0 examples [00:00, ? examples/s]\r\n\r\nGenerating train split: 0 examples [00:00, ? examples/s]Traceback (most recent call last):\r\n\r\n File \"/opt/anaconda3/envs/audio/lib/python3.9/site-packages/datasets/builder.py\", line 1571, in _prepare_split_single\r\n for key, record in generator:\r\n\r\n File \"/Users/nguyenvannham/.cache/huggingface/modules/datasets_modules/datasets/timit_asr/43f9448dd5db58e95ee48a277f466481b151f112ea53e27f8173784da9254fb2/timit_asr.py\", line 138, in _generate_examples\r\n with txt_path.open(encoding=\"utf-8\") as op:\r\n\r\n File \"/opt/anaconda3/envs/audio/lib/python3.9/pathlib.py\", line 1252, in open\r\n return io.open(self, mode, buffering, encoding, errors, newline,\r\n\r\n File \"/opt/anaconda3/envs/audio/lib/python3.9/pathlib.py\", line 1120, in _opener\r\n return self._accessor.open(self, flags, mode)\r\n\r\nFileNotFoundError: [Errno 2] No such file or directory: '/Users/nguyenvannham/Documents/test_case/data/train/DR1/FCJF0/SA1.WAV.TXT'\r\n\r\n\r\nThe above exception was the direct cause of the following exception:\r\n\r\nTraceback (most recent call last):\r\n\r\n File \"/var/folders/t9/l8d3rwpn1k33_gjtqs732lzc0000gn/T/ipykernel_3891/1203313828.py\", line 1, in <module>\r\n dataset = load_dataset(\"timit_asr\", data_dir=\"/Users/nguyenvannham/Documents/test_case/data\")\r\n\r\n File \"/opt/anaconda3/envs/audio/lib/python3.9/site-packages/datasets/load.py\", line 1758, in load_dataset\r\n builder_instance.download_and_prepare(\r\n\r\n File \"/opt/anaconda3/envs/audio/lib/python3.9/site-packages/datasets/builder.py\", line 860, in download_and_prepare\r\n self._download_and_prepare(\r\n\r\n File \"/opt/anaconda3/envs/audio/lib/python3.9/site-packages/datasets/builder.py\", line 1612, in _download_and_prepare\r\n super()._download_and_prepare(\r\n\r\n File \"/opt/anaconda3/envs/audio/lib/python3.9/site-packages/datasets/builder.py\", line 953, in _download_and_prepare\r\n self._prepare_split(split_generator, **prepare_split_kwargs)\r\n\r\n File \"/opt/anaconda3/envs/audio/lib/python3.9/site-packages/datasets/builder.py\", line 1450, in _prepare_split\r\n for job_id, done, content in self._prepare_split_single(\r\n\r\n File \"/opt/anaconda3/envs/audio/lib/python3.9/site-packages/datasets/builder.py\", line 1607, in _prepare_split_single\r\n raise DatasetGenerationError(\"An error occurred while generating the dataset\") from e\r\n\r\nDatasetGenerationError: An error occurred while generating the dataset"
] | ## Dataset viewer issue for '*timit_asr*'
**Link:** *https://huggingface.co/datasets/timit_asr*
Issue: The timit-asr dataset cannot be previewed recently.
Am I the one who added this dataset ? Yes-No
No | 4,169 |
https://github.com/huggingface/datasets/issues/4163 | Optional Content Warning for Datasets | [
"Hi! You can use the `extra_gated_prompt` YAML field in a dataset card for displaying custom messages/warnings that the user must accept before gaining access to the actual dataset. This option also keeps the viewer hidden until the user agrees to terms. ",
"Hi @mariosasko, thanks for explaining how to add this feature. \r\n\r\nIf the current dataset yaml is:\r\n```\r\n---\r\nannotations_creators:\r\n- expert\r\nlanguage_creators:\r\n- expert-generated\r\nlanguages:\r\n- en\r\nlicense:\r\n- cc-by-4.0\r\nmultilinguality:\r\n- monolingual\r\npretty_name: HatemojiBuild\r\nsize_categories:\r\n- 1K<n<10K\r\nsource_datasets:\r\n- original\r\ntask_categories:\r\n- text-classification\r\ntask_ids:\r\n- hate-speech-detection\r\n---\r\n```\r\n\r\nCan you provide a minimal working example of how to added the gated prompt?\r\n\r\nThanks!",
"```\r\n---\r\nannotations_creators:\r\n- expert\r\nlanguage_creators:\r\n- expert-generated\r\nlanguages:\r\n- en\r\nlicense:\r\n- cc-by-4.0\r\nmultilinguality:\r\n- monolingual\r\npretty_name: HatemojiBuild\r\nsize_categories:\r\n- 1K<n<10K\r\nsource_datasets:\r\n- original\r\ntask_categories:\r\n- text-classification\r\ntask_ids:\r\n- hate-speech-detection\r\nextra_gated_prompt: \"This repository contains harmful content.\"\r\n---\r\n```\r\n\\+ enable `User Access requests` under the Settings pane.\r\n\r\nThere's a brief guide here https://discuss.huggingface.co/t/how-to-customize-the-user-access-requests-message/13953 , and you can see the field in action here, https://huggingface.co/datasets/mozilla-foundation/common_voice_7_0/blob/main/README.md (you need to agree the terms in the Dataset Card pane to be able to access the files pane, so this comes up 403 at first).\r\n\r\nAnd a working example here! https://huggingface.co/datasets/DDSC/dkhate :) Great to be able to mitigate harms in text.",
"-- is there a way to gate content anonymously, i.e. without registering which users access it?",
"+1 to @leondz's question. One scenario is if you don't want the dataset to be indexed by search engines or viewed in browser b/c of upstream conditions on data, but don't want to collect emails. Some ability to turn off the dataset viewer or add a gating mechanism without emails would be fantastic."
] | **Is your feature request related to a problem? Please describe.**
A clear and concise description of what the problem is.
We now have hate speech datasets on the hub, like this one: https://huggingface.co/datasets/HannahRoseKirk/HatemojiBuild
I'm wondering if there is an option to select a content warning message that appears before the dataset preview? Otherwise, people immediately see hate speech when clicking on this dataset.
**Describe the solution you'd like**
A clear and concise description of what you want to happen.
Implementation of a content warning message that separates users from the dataset preview until they click out of the warning.
**Describe alternatives you've considered**
A clear and concise description of any alternative solutions or features you've considered.
Possibly just a way to remove the dataset preview completely? I think I like the content warning option better, though.
**Additional context**
Add any other context about the feature request here.
| 4,163 |
https://github.com/huggingface/datasets/issues/4160 | RGBA images not showing | [
"Thanks for reporting. It's a known issue, and we hope to fix it soon.",
"Fixed, thanks!"
] | ## Dataset viewer issue for ceyda/smithsonian_butterflies_transparent
[**Link:** *link to the dataset viewer page*](https://huggingface.co/datasets/ceyda/smithsonian_butterflies_transparent)
Am I the one who added this dataset ? Yes
👉 More of a general issue of 'RGBA' png images not being supported
(the dataset itself is just for the huggan sprint and not that important, consider it just an example) | 4,160 |
https://github.com/huggingface/datasets/issues/4152 | ArrayND error in pyarrow 5 | [
"Where do we bump the required pyarrow version? Any inputs on how I fix this issue? ",
"We need to bump it in `setup.py` as well as update some CI job to use pyarrow 6 instead of 5 in `.circleci/config.yaml` and `.github/workflows/benchmarks.yaml`"
] | As found in https://github.com/huggingface/datasets/pull/3903, The ArrayND features fail on pyarrow 5:
```python
import pyarrow as pa
from datasets import Array2D
from datasets.table import cast_array_to_feature
arr = pa.array([[[0]]])
feature_type = Array2D(shape=(1, 1), dtype="int64")
cast_array_to_feature(arr, feature_type)
```
raises
```python
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
<ipython-input-8-04610f9fa78c> in <module>
----> 1 cast_array_to_feature(pa.array([[[0]]]), Array2D(shape=(1, 1), dtype="int32"))
~/Desktop/hf/datasets/src/datasets/table.py in wrapper(array, *args, **kwargs)
1672 return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
1673 else:
-> 1674 return func(array, *args, **kwargs)
1675
1676 return wrapper
~/Desktop/hf/datasets/src/datasets/table.py in cast_array_to_feature(array, feature, allow_number_to_str)
1806 return array_cast(array, get_nested_type(feature), allow_number_to_str=allow_number_to_str)
1807 elif not isinstance(feature, (Sequence, dict, list, tuple)):
-> 1808 return array_cast(array, feature(), allow_number_to_str=allow_number_to_str)
1809 raise TypeError(f"Couldn't cast array of type\n{array.type}\nto\n{feature}")
1810
~/Desktop/hf/datasets/src/datasets/table.py in wrapper(array, *args, **kwargs)
1672 return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
1673 else:
-> 1674 return func(array, *args, **kwargs)
1675
1676 return wrapper
~/Desktop/hf/datasets/src/datasets/table.py in array_cast(array, pa_type, allow_number_to_str)
1705 array = array.storage
1706 if isinstance(pa_type, pa.ExtensionType):
-> 1707 return pa_type.wrap_array(array)
1708 elif pa.types.is_struct(array.type):
1709 if pa.types.is_struct(pa_type) and (
AttributeError: 'Array2DExtensionType' object has no attribute 'wrap_array'
```
The thing is that `cast_array_to_feature` is called when writing an Arrow file, so creating an Arrow dataset using any ArrayND type currently fails.
`wrap_array` has been added in pyarrow 6, so we can either bump the required pyarrow version or fix this for pyarrow 5 | 4,152 |
https://github.com/huggingface/datasets/issues/4150 | Inconsistent splits generation for datasets without loading script (packaged dataset puts everything into a single split) | [] | ## Describe the bug
Splits for dataset loaders without scripts are prepared inconsistently. I think it might be confusing for users.
## Steps to reproduce the bug
* If you load a packaged datasets from Hub, it infers splits from directory structure / filenames (check out the data [here](https://huggingface.co/datasets/nateraw/test-imagefolder-dataset)):
```python
ds = load_dataset("nateraw/test-imagefolder-dataset")
print(ds)
### Output:
DatasetDict({
train: Dataset({
features: ['image', 'label'],
num_rows: 6
})
test: Dataset({
features: ['image', 'label'],
num_rows: 4
})
})
```
* If you do the same from locally stored data specifying only directory path you'll get the same:
```python
ds = load_dataset("/path/to/local/data/test-imagefolder-dataset")
print(ds)
### Output:
DatasetDict({
train: Dataset({
features: ['image', 'label'],
num_rows: 6
})
test: Dataset({
features: ['image', 'label'],
num_rows: 4
})
})
```
* However, if you explicitely specify package name (like `imagefolder`, `csv`, `json`), all the data is put into a single split:
```python
ds = load_dataset("imagefolder", data_dir="/path/to/local/data/test-imagefolder-dataset")
print(ds)
### Output:
DatasetDict({
train: Dataset({
features: ['image', 'label'],
num_rows: 10
})
})
```
## Expected results
For `load_dataset("imagefolder", data_dir="/path/to/local/data/test-imagefolder-dataset")` I expect the same output as of the two first options. | 4,150 |
https://github.com/huggingface/datasets/issues/4149 | load_dataset for winoground returning decoding error | [
"I thought I had fixed it with this after some helpful hints from @severo\r\n```python\r\nimport datasets \r\ntoken = 'hf_XXXXX'\r\ndataset = datasets.load_dataset(\r\n 'facebook/winoground', \r\n name='facebook--winoground', \r\n split='train', \r\n streaming=True,\r\n use_auth_token=token,\r\n)\r\n```\r\nbut I found out that wasn't the case\r\n```python\r\n[x for x in dataset]\r\n...\r\nClientResponseError: 401, message='Unauthorized', url=URL('https://huggingface.co/datasets/facebook/winoground/resolve/a86a60456fbbd242e9a744199071a6bd3e7fd9de/examples.jsonl')\r\n```",
"Hi ! This dataset structure (image + labels in a JSON file) is not supported yet, though we're adding support for this in in #4069 \r\n\r\nThe following structure will be supported soon:\r\n```\r\nmetadata.json\r\nimages/\r\n image0.png\r\n image1.png\r\n ...\r\n```\r\nWhere `metadata.json` is a JSON Lines file with labels or other metadata, and each line must have a \"file_name\" field with the name of the image file.\r\n\r\nFor the moment are only supported:\r\n- JSON files only\r\n- image files only\r\n\r\nSince this dataset is a mix of the two, at the moment it fails trying to read the images as JSON.\r\n\r\nTherefore to be able to load this dataset we need to wait for the new structure to be supported (very soon ^^), or add a dataset script in the repository that reads both the JSON and the images cc @TristanThrush \r\n",
"We'll also investigate the issue with the streaming download manager in https://github.com/huggingface/datasets/issues/4139 ;) thanks for reporting",
"Are there any updates on this?",
"In the meantime, anyone can always download the images.zip and examples.jsonl files directly from huggingface.co - let me know if anyone has issues with that.",
"I mirrored the files at https://huggingface.co/datasets/facebook/winoground in a folder on my local machine `winground`\r\nand when I tried\r\n```python\r\nimport datasets\r\nds = datasets.load_from_disk('./winoground')\r\n```\r\nI get the following error\r\n```python\r\n--------------------------------------------------------------------------\r\nFileNotFoundError Traceback (most recent call last)\r\nInput In [2], in <cell line: 1>()\r\n----> 1 ds = datasets.load_from_disk('./winoground')\r\n\r\nFile ~/.local/lib/python3.8/site-packages/datasets/load.py:1759, in load_from_disk(dataset_path, fs, keep_in_memory)\r\n 1757 return DatasetDict.load_from_disk(dataset_path, fs, keep_in_memory=keep_in_memory)\r\n 1758 else:\r\n-> 1759 raise FileNotFoundError(\r\n 1760 f\"Directory {dataset_path} is neither a dataset directory nor a dataset dict directory.\"\r\n 1761 )\r\n\r\nFileNotFoundError: Directory ./winoground is neither a dataset directory nor a dataset dict directory.\r\n```\r\nso still some work to be done on the backend imo.",
"Note that `load_from_disk` is the function that reloads an Arrow dataset saved with `my_dataset.save_to_disk`.\r\n\r\nOnce we do support images with metadata you'll be able to use `load_dataset(\"facebook/winoground\")` directly (or `load_dataset(\"./winoground\")` of you've cloned the winoground repository locally).",
"Apologies for the delay. I added a custom dataset loading script for winoground. It should work now, with an auth token:\r\n\r\n`examples = load_dataset('facebook/winoground', use_auth_token=<your auth token>)`\r\n\r\nLet me know if there are any issues",
"Adding the dataset loading script definitely didn't take as long as I thought it would 😅",
"killer"
] | ## Describe the bug
I am trying to use datasets to load winoground and I'm getting a JSON decoding error.
## Steps to reproduce the bug
```python
from datasets import load_dataset
token = 'hf_XXXXX' # my HF access token
datasets = load_dataset('facebook/winoground', use_auth_token=token)
```
## Expected results
I downloaded images.zip and examples.jsonl manually. I was expecting to have some trouble decoding json so I didn't use jsonlines but instead was able to get a complete set of 400 examples by doing
```python
import json
with open('examples.jsonl', 'r') as f:
examples = f.read().split('\n')
# Thinking this would error if the JSON is not utf-8 encoded
json_data = [json.loads(x) for x in examples]
print(json_data[-1])
```
and I see
```python
{'caption_0': 'someone is overdoing it',
'caption_1': 'someone is doing it over',
'collapsed_tag': 'Relation',
'id': 399,
'image_0': 'ex_399_img_0',
'image_1': 'ex_399_img_1',
'num_main_preds': 1,
'secondary_tag': 'Morpheme-Level',
'tag': 'Scope, Preposition'}
```
so I'm not sure what's going on here honestly. The file `examples.jsonl` doesn't have non-UTF-8 encoded text.
## Actual results
During the split operation after downloading, datasets encounters an error in the JSON ([trace](https://gist.github.com/odellus/e55d390ca203386bf551f38e0c63a46b) abbreviated for brevity).
```
datasets/packaged_modules/json/json.py:144 in Json._generate_tables(self, files)
...
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xff in position 0: invalid start byte
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.18.4
- Platform: Linux-5.13.0-39-generic-x86_64-with-glibc2.29
- Python version: 3.8.10
- PyArrow version: 7.0.0
| 4,149 |
https://github.com/huggingface/datasets/issues/4148 | fix confusing bleu metric example | [] | **Is your feature request related to a problem? Please describe.**
I would like to see the example in "Metric Card for BLEU" changed.
The 0th element in the predictions list is not closed in square brackets, and the 1st list is missing a comma.
The BLEU score are calculated correctly, but it is difficult to understand, so it would be helpful if you could correct this.
```
>> predictions = [
... ["hello", "there", "general", "kenobi", # <- no closing square bracket.
... ["foo", "bar" "foobar"] # <- no comma between "bar" and "foobar"
... ]
>>> references = [
... [["hello", "there", "general", "kenobi"]],
... [["foo", "bar", "foobar"]]
... ]
>>> bleu = datasets.load_metric("bleu")
>>> results = bleu.compute(predictions=predictions, references=references)
>>> print(results)
{'bleu': 0.6370964381207871, ...
```
**Describe the solution you'd like**
```
>> predictions = [
... ["hello", "there", "general", "kenobi", # <- no closing square bracket.
... ["foo", "bar" "foobar"] # <- no comma between "bar" and "foobar"
... ]
# and
>>> print(results)
{'bleu':1.0, ...
```
| 4,148 |
https://github.com/huggingface/datasets/issues/4146 | SAMSum dataset viewer not working | [
"https://huggingface.co/datasets/samsum\r\n\r\n```\r\nStatus code: 400\r\nException: ValueError\r\nMessage: Cannot seek streaming HTTP file\r\n```",
"Currently, only the datasets that can be streamed support the dataset viewer. Maybe @lhoestq @albertvillanova or @mariosasko could give more details about why the dataset cannot be streamed.",
"It looks like the host (https://arxiv.org) doesn't allow HTTP Range requests, which is what we use to stream data.\r\n\r\nThis can be fix if we host the data ourselves, which is ok since the dataset is under CC BY-NC-ND 4.0"
] | ## Dataset viewer issue for '*name of the dataset*'
**Link:** *link to the dataset viewer page*
*short description of the issue*
Am I the one who added this dataset ? Yes-No
| 4,146 |
https://github.com/huggingface/datasets/issues/4143 | Unable to download `Wikepedia` 20220301.en version | [
"Hi! We've recently updated the Wikipedia script, so these changes are only available on master and can be fetched as follows:\r\n```python\r\ndataset_wikipedia = load_dataset(\"wikipedia\", \"20220301.en\", revision=\"master\")\r\n```",
"Hi, how can I load the previous \"20200501.en\" version of wikipedia which had been downloaded to the default path? Thanks!",
"@JiaQiSJTU just reinstall the previous verision of the package, e.g. `!pip install -q datasets==1.0.0`"
] | ## Describe the bug
Unable to download `Wikepedia` dataset, 20220301.en version
## Steps to reproduce the bug
```python
!pip install apache_beam mwparserfromhell
dataset_wikipedia = load_dataset("wikipedia", "20220301.en")
```
## Actual results
```
ValueError: BuilderConfig 20220301.en not found.
Available: ['20200501.aa', '20200501.ab', '20200501.ace', '20200501.ady', '20200501.af', '20200501.ak', '20200501.als', '20200501.am', '20200501.an', '20200501.ang', '20200501.ar', '20200501.arc', '20200501.arz', '20200501.as', '20200501.ast', '20200501.atj', '20200501.av', '20200501.ay', '20200501.az', '20200501.azb', '20200501.ba', '20200501.bar', '20200501.bat-smg', '20200501.bcl', '20200501.be', '20200501.be-x-old', '20200501.bg', '20200501.bh', '20200501.bi', '20200501.bjn', '20200501.bm', '20200501.bn', '20200501.bo', '20200501.bpy', '20200501.br', '20200501.bs', '20200501.bug', '20200501.bxr', '20200501.ca', '20200501.cbk-zam', '20200501.cdo', '20200501.ce', '20200501.ceb', '20200501.ch', '20200501.cho', '20200501.chr', '20200501.chy', '20200501.ckb', '20200501.co', '20200501.cr', '20200501.crh', '20200501.cs', '20200501.csb', '20200501.cu', '20200501.cv', '20200501.cy', '20200501.da', '20200501.de', '20200501.din', '20200501.diq', '20200501.dsb', '20200501.dty', '20200501.dv', '20200501.dz', '20200501.ee', '20200501.el', '20200501.eml', '20200501.en', '20200501.eo', '20200501.es', '20200501.et', '20200501.eu', '20200501.ext', '20200501.fa', '20200501.ff', '20200501.fi', '20200501.fiu-vro', '20200501.fj', '20200501.fo', '20200501.fr', '20200501.frp', '20200501.frr', '20200501.fur', '20200501.fy', '20200501.ga', '20200501.gag', '20200501.gan', '20200501.gd', '20200501.gl', '20200501.glk', '20200501.gn', '20200501.gom', '20200501.gor', '20200501.got', '20200501.gu', '20200501.gv', '20200501.ha', '20200501.hak', '20200501.haw', '20200501.he', '20200501.hi', '20200501.hif', '20200501.ho', '20200501.hr', '20200501.hsb', '20200501.ht', '20200501.hu', '20200501.hy', '20200501.ia', '20200501.id', '20200501.ie', '20200501.ig', '20200501.ii', '20200501.ik', '20200501.ilo', '20200501.inh', '20200501.io', '20200501.is', '20200501.it', '20200501.iu', '20200501.ja', '20200501.jam', '20200501.jbo', '20200501.jv', '20200501.ka', '20200501.kaa', '20200501.kab', '20200501.kbd', '20200501.kbp', '20200501.kg', '20200501.ki', '20200501.kj', '20200501.kk', '20200501.kl', '20200501.km', '20200501.kn', '20200501.ko', '20200501.koi', '20200501.krc', '20200501.ks', '20200501.ksh', '20200501.ku', '20200501.kv', '20200501.kw', '20200501.ky', '20200501.la', '20200501.lad', '20200501.lb', '20200501.lbe', '20200501.lez', '20200501.lfn', '20200501.lg', '20200501.li', '20200501.lij', '20200501.lmo', '20200501.ln', '20200501.lo', '20200501.lrc', '20200501.lt', '20200501.ltg', '20200501.lv', '20200501.mai', '20200501.map-bms', '20200501.mdf', '20200501.mg', '20200501.mh', '20200501.mhr', '20200501.mi', '20200501.min', '20200501.mk', '20200501.ml', '20200501.mn', '20200501.mr', '20200501.mrj', '20200501.ms', '20200501.mt', '20200501.mus', '20200501.mwl', '20200501.my', '20200501.myv', '20200501.mzn', '20200501.na', '20200501.nah', '20200501.nap', '20200501.nds', '20200501.nds-nl', '20200501.ne', '20200501.new', '20200501.ng', '20200501.nl', '20200501.nn', '20200501.no', '20200501.nov', '20200501.nrm', '20200501.nso', '20200501.nv', '20200501.ny', '20200501.oc', '20200501.olo', '20200501.om', '20200501.or', '20200501.os', '20200501.pa', '20200501.pag', '20200501.pam', '20200501.pap', '20200501.pcd', '20200501.pdc', '20200501.pfl', '20200501.pi', '20200501.pih', '20200501.pl', '20200501.pms', '20200501.pnb', '20200501.pnt', '20200501.ps', '20200501.pt', '20200501.qu', '20200501.rm', '20200501.rmy', '20200501.rn', '20200501.ro', '20200501.roa-rup', '20200501.roa-tara', '20200501.ru', '20200501.rue', '20200501.rw', '20200501.sa', '20200501.sah', '20200501.sat', '20200501.sc', '20200501.scn', '20200501.sco', '20200501.sd', '20200501.se', '20200501.sg', '20200501.sh', '20200501.si', '20200501.simple', '20200501.sk', '20200501.sl', '20200501.sm', '20200501.sn', '20200501.so', '20200501.sq', '20200501.sr', '20200501.srn', '20200501.ss', '20200501.st', '20200501.stq', '20200501.su', '20200501.sv', '20200501.sw', '20200501.szl', '20200501.ta', '20200501.tcy', '20200501.te', '20200501.tet', '20200501.tg', '20200501.th', '20200501.ti', '20200501.tk', '20200501.tl', '20200501.tn', '20200501.to', '20200501.tpi', '20200501.tr', '20200501.ts', '20200501.tt', '20200501.tum', '20200501.tw', '20200501.ty', '20200501.tyv', '20200501.udm', '20200501.ug', '20200501.uk', '20200501.ur', '20200501.uz', '20200501.ve', '20200501.vec', '20200501.vep', '20200501.vi', '20200501.vls', '20200501.vo', '20200501.wa', '20200501.war', '20200501.wo', '20200501.wuu', '20200501.xal', '20200501.xh', '20200501.xmf', '20200501.yi', '20200501.yo', '20200501.za', '20200501.zea', '20200501.zh', '20200501.zh-classical', '20200501.zh-min-nan', '20200501.zh-yue', '20200501.zu']
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.0.0
- Platform: Ubuntu
- Python version: 3.6
- PyArrow version: 6.0.1 | 4,143 |
https://github.com/huggingface/datasets/issues/4142 | Add ObjectFolder 2.0 dataset | [
"Datasets are not tracked in this repository anymore."
] | ## Adding a Dataset
- **Name:** ObjectFolder 2.0
- **Description:** ObjectFolder 2.0 is a dataset of 1,000 objects in the form of implicit representations. It contains 1,000 Object Files each containing the complete multisensory profile for an object instance.
- **Paper:** [*link to the dataset paper if available*](https://arxiv.org/abs/2204.02389)
- **Data:** https://github.com/rhgao/ObjectFolder
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| 4,142 |
https://github.com/huggingface/datasets/issues/4141 | Why is the dataset not visible under the dataset preview section? | [] | ## Dataset viewer issue for '*name of the dataset*'
**Link:** *link to the dataset viewer page*
*short description of the issue*
Am I the one who added this dataset ? Yes-No
| 4,141 |
https://github.com/huggingface/datasets/issues/4140 | Error loading arxiv data set | [
"Hi! I think this error may be related to using an older version of the library. I was able to load the dataset without any issues using the latest version of `datasets`. Can you upgrade to the latest version of `datasets` and try again? :)",
"Hi! As @stevhliu suggested, to fix the issue, update the lib to the newest version with:\r\n```\r\npip install -U datasets\r\n```\r\nand download the dataset as follows:\r\n```python\r\nfrom datasets import load_dataset\r\ndset = load_dataset('scientific_papers', 'arxiv', download_mode=\"force_redownload\")\r\n```",
"Thanks for the quick response! It works now. The problem is that I used nlp. load_dataset instead of datasets. load_dataset."
] | ## Describe the bug
A clear and concise description of what the bug is.
I met the error below when loading arxiv dataset via `nlp.load_dataset('scientific_papers', 'arxiv',)`.
```
Traceback (most recent call last):
File "scripts/summarization.py", line 354, in <module>
main(args)
File "scripts/summarization.py", line 306, in main
model.hf_datasets = nlp.load_dataset('scientific_papers', 'arxiv')
File "/opt/conda/envs/longformer/lib/python3.7/site-packages/nlp/load.py", line 549, in load_dataset
download_config=download_config, download_mode=download_mode, ignore_verifications=ignore_verifications,
File "/opt/conda/envs/longformer/lib/python3.7/site-packages/nlp/builder.py", line 463, in download_and_prepare
dl_manager=dl_manager, verify_infos=verify_infos, **download_and_prepare_kwargs
File "/opt/conda/envs/longformer/lib/python3.7/site-packages/nlp/builder.py", line 522, in _download_and_prepare
self.info.download_checksums, dl_manager.get_recorded_sizes_checksums(), "dataset source files"
File "/opt/conda/envs/longformer/lib/python3.7/site-packages/nlp/utils/info_utils.py", line 38, in verify_checksums
raise NonMatchingChecksumError(error_msg + str(bad_urls))
nlp.utils.info_utils.NonMatchingChecksumError: Checksums didn't match for dataset source files:
['https://drive.google.com/uc?id=1b3rmCSIoh6VhD4HKWjI4HOW-cSwcwbeC&export=download', 'https://drive.google.com/uc?id=1lvsqvsFi3W-pE1SqNZI0s8NR9rC1tsja&export=download']
```
I then tried to ignore verification steps by `ignore_verifications=True` and there is another error.
```
Traceback (most recent call last):
File "/opt/conda/envs/longformer/lib/python3.7/site-packages/nlp/builder.py", line 537, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/opt/conda/envs/longformer/lib/python3.7/site-packages/nlp/builder.py", line 810, in _prepare_split
for key, record in utils.tqdm(generator, unit=" examples", total=split_info.num_examples, leave=False):
File "/opt/conda/envs/longformer/lib/python3.7/site-packages/tqdm/std.py", line 1195, in __iter__
for obj in iterable:
File "/opt/conda/envs/longformer/lib/python3.7/site-packages/nlp/datasets/scientific_papers/9e4f2cfe3d8494e9f34a84ce49c3214605b4b52a3d8eb199104430d04c52cc12/scientific_papers.py", line 108, in _generate_examples
with open(path, encoding="utf-8") as f:
NotADirectoryError: [Errno 20] Not a directory: '/home/username/.cache/huggingface/datasets/downloads/c0deae7af7d9c87f25dfadf621f7126f708d7dcac6d353c7564883084a000076/arxiv-dataset/train.txt'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "scripts/summarization.py", line 354, in <module>
main(args)
File "scripts/summarization.py", line 306, in main
model.hf_datasets = nlp.load_dataset('scientific_papers', 'arxiv', ignore_verifications=True)
File "/opt/conda/envs/longformer/lib/python3.7/site-packages/nlp/load.py", line 549, in load_dataset
download_config=download_config, download_mode=download_mode, ignore_verifications=ignore_verifications,
File "/opt/conda/envs/longformer/lib/python3.7/site-packages/nlp/builder.py", line 463, in download_and_prepare
dl_manager=dl_manager, verify_infos=verify_infos, **download_and_prepare_kwargs
File "/opt/conda/envs/longformer/lib/python3.7/site-packages/nlp/builder.py", line 539, in _download_and_prepare
raise OSError("Cannot find data file. " + (self.manual_download_instructions or ""))
OSError: Cannot find data file.
```
## Steps to reproduce the bug
```python
# Sample code to reproduce the bug
```
## Expected results
A clear and concise description of the expected results.
## Actual results
Specify the actual results or traceback.
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version:
- Platform:
- Python version:
- PyArrow version:
| 4,140 |
https://github.com/huggingface/datasets/issues/4139 | Dataset viewer issue for Winoground | [
"related (same dataset): https://github.com/huggingface/datasets/issues/4149. But the issue is different. Looking at it",
"I thought this issue was related to the error I was seeing, but upon consideration I'd think the dataset viewer would return a 500 (unable to create the split like me) or a 404 (unable to load split b/c it was never created) error if it was having the issue I was seeing in #4149. 401 message makes it look like dataset viewer isn't passing through the identity of the user who has signed the licensing agreement when making the request to GET [examples.jsonl](https://huggingface.co/datasets/facebook/winoground/resolve/a86a60456fbbd242e9a744199071a6bd3e7fd9de/examples.jsonl).",
"Pinging @SBrandeis, as it seems related to gated datasets and access tokens.",
"To replicate:\r\n\r\n```python\r\n>>> import datasets\r\n>>> dataset= datasets.load_dataset('facebook/winoground', name='facebook--winoground', split='train', use_auth_token=\"hf_app_...\", streaming=True)\r\n>>> next(iter(dataset))\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/iterable_dataset.py\", line 497, in __iter__\r\n for key, example in self._iter():\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/iterable_dataset.py\", line 494, in _iter\r\n yield from ex_iterable\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/iterable_dataset.py\", line 87, in __iter__\r\n yield from self.generate_examples_fn(**self.kwargs)\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/iterable_dataset.py\", line 439, in wrapper\r\n for key, table in generate_tables_fn(**kwargs):\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/packaged_modules/json/json.py\", line 85, in _generate_tables\r\n for file_idx, file in enumerate(files):\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/utils/streaming_download_manager.py\", line 679, in __iter__\r\n yield from self.generator(*self.args, **self.kwargs)\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/utils/streaming_download_manager.py\", line 731, in _iter_from_urlpaths\r\n for dirpath, _, filenames in xwalk(urlpath, use_auth_token=use_auth_token):\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/utils/streaming_download_manager.py\", line 623, in xwalk\r\n for dirpath, dirnames, filenames in fs.walk(main_hop):\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/fsspec/spec.py\", line 372, in walk\r\n listing = self.ls(path, detail=True, **kwargs)\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/fsspec/asyn.py\", line 85, in wrapper\r\n return sync(self.loop, func, *args, **kwargs)\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/fsspec/asyn.py\", line 65, in sync\r\n raise return_result\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/fsspec/asyn.py\", line 25, in _runner\r\n result[0] = await coro\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/fsspec/implementations/http.py\", line 196, in _ls\r\n out = await self._ls_real(url, detail=detail, **kwargs)\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/fsspec/implementations/http.py\", line 150, in _ls_real\r\n self._raise_not_found_for_status(r, url)\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/fsspec/implementations/http.py\", line 208, in _raise_not_found_for_status\r\n response.raise_for_status()\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/aiohttp/client_reqrep.py\", line 1004, in raise_for_status\r\n raise ClientResponseError(\r\naiohttp.client_exceptions.ClientResponseError: 401, message='Unauthorized', url=URL('https://huggingface.co/datasets/facebook/winoground/resolve/a86a60456fbbd242e9a744199071a6bd3e7fd9de/examples.jsonl')\r\n```\r\n\r\n*edited to fix `use_token` -> `use_auth_token`, thx @odellus*",
"~~Using your command to replicate and changing `use_token` to `use_auth_token` fixes the problem I was seeing in #4149.~~\r\nNevermind it gave me an iterator to a method returning the same 401s. Changing `use_token` to `use_auth_token` does not fix the issue.",
"After investigation with @severo , we found a potential culprit: https://github.com/huggingface/datasets/blob/3cd0a009a43f9f174056d70bfa2ca32216181926/src/datasets/utils/streaming_download_manager.py#L610-L624\r\n\r\nThe streaming manager does not seem to pass `use_auth_token` to `fsspec` when streaming and not iterating content of a zip archive\r\n\r\ncc @albertvillanova @lhoestq ",
"I was able to reproduce it on a private dataset, let me work on a fix",
"Hey @lhoestq, Thanks for working on a fix! Any plans to merge #4173 into master? ",
"Thanks for the heads up, I still need to fix some tests that are failing in the CI before merging ;)",
"The fix has been merged, we'll do a new release soon, and update the dataset viewer",
"Fixed, thanks!\r\n<img width=\"1119\" alt=\"Capture d’écran 2022-06-21 à 18 41 09\" src=\"https://user-images.githubusercontent.com/1676121/174853571-afb0749c-4178-4c89-ab40-bb162a449788.png\">\r\n"
] | ## Dataset viewer issue for 'Winoground'
**Link:** [*link to the dataset viewer page*](https://huggingface.co/datasets/facebook/winoground/viewer/facebook--winoground/train)
*short description of the issue*
Getting 401, message='Unauthorized'
The dataset is subject to authorization, but I can access the files from the interface, so I assume I'm granted to access it. I'd assume the permission somehow doesn't propagate to the dataset viewer tool.
Am I the one who added this dataset ? No
| 4,139 |
https://github.com/huggingface/datasets/issues/4138 | Incorrect Russian filenames encoding after extraction by datasets.DownloadManager.download_and_extract() | [
"To reproduce:\r\n\r\n```python\r\n>>> import datasets\r\n>>> datasets.get_dataset_split_names('MalakhovIlya/RuREBus', config_name='raw_txt')\r\nTraceback (most recent call last):\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/inspect.py\", line 280, in get_dataset_config_info\r\n for split_generator in builder._split_generators(\r\n File \"/home/slesage/.cache/huggingface/modules/datasets_modules/datasets/MalakhovIlya--RuREBus/21046f5f1a0cf91187d68c30918d78d934ec7113ec435e146776d4f28f12c4ed/RuREBus.py\", line 101, in _split_generators\r\n decode_file_names(folder)\r\n File \"/home/slesage/.cache/huggingface/modules/datasets_modules/datasets/MalakhovIlya--RuREBus/21046f5f1a0cf91187d68c30918d78d934ec7113ec435e146776d4f28f12c4ed/RuREBus.py\", line 26, in decode_file_names\r\n for root, dirs, files in os.walk(folder, topdown=False):\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/streaming.py\", line 66, in wrapper\r\n return function(*args, use_auth_token=use_auth_token, **kwargs)\r\nTypeError: xwalk() got an unexpected keyword argument 'topdown'\r\n\r\nThe above exception was the direct cause of the following exception:\r\n\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/inspect.py\", line 323, in get_dataset_split_names\r\n info = get_dataset_config_info(\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/inspect.py\", line 285, in get_dataset_config_info\r\n raise SplitsNotFoundError(\"The split names could not be parsed from the dataset config.\") from err\r\ndatasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.\r\n```\r\n\r\nIt's not related to the dataset viewer. Maybe @albertvillanova or @lhoestq could help more on this issue.",
"Hi! This issue stems from the fact that `xwalk`, which is a streamable version of `os.walk`, doesn't support the `topdown` param due to `fsspec`'s `walk` also not supporting it, so fixing this issue could be tricky. \r\n\r\n@MalakhovIlyaPavlovich You can avoid the error by tweaking your data processing and not using this param. (and `Path.rename`, which also cannot be streamed) ",
"@mariosasko thank you for your reply. I couldn't reproduce error showed by @severo either on Ubuntu 20.04.3 LTS, Windows 10 and Google Colab environments. But trying to avoid using os.walk(topdown=False) and Path.rename(), In _split_generators I replaced\r\n```\r\ndef decode_file_names(folder):\r\n for root, dirs, files in os.walk(folder, topdown=False):\r\n root = Path(root)\r\n for file in files:\r\n old_name = root / Path(file)\r\n new_name = root / Path(\r\n file.encode('cp437').decode('cp866'))\r\n old_name.rename(new_name)\r\n for dir in dirs:\r\n old_name = root / Path(dir)\r\n new_name = root / Path(dir.encode('cp437').decode('cp866'))\r\n old_name.rename(new_name)\r\n\r\nfolder = dl_manager.download_and_extract(self._RAW_TXT_URLS)['raw_txt']\r\ndecode_file_names(folder)\r\n```\r\nby\r\n```\r\ndef extract(zip_file_path):\r\n p = Path(zip_file_path)\r\n dest_dir = str(p.parent / 'extracted' / p.stem)\r\n os.makedirs(dest_dir, exist_ok=True)\r\n with zipfile.ZipFile(zip_file_path) as archive:\r\n for file_info in tqdm(archive.infolist(), desc='Extracting'):\r\n filename = file_info.filename.encode('cp437').decode('cp866')\r\n target = os.path.join(dest_dir, *filename.split('/'))\r\n os.makedirs(os.path.dirname(target), exist_ok=True)\r\n if not file_info.is_dir():\r\n with archive.open(file_info) as source, open(target, 'wb') as dest:\r\n shutil.copyfileobj(source, dest)\r\n return dest_dir\r\n\r\nzip_file = dl_manager.download(self._RAW_TXT_URLS)['raw_txt']\r\nif not is_url(zip_file):\r\n folder = extract(zip_file)\r\nelse:\r\n folder = None\r\n```\r\nand now everything works well except data viewer for \"raw_txt\" subset: dataset preview on hub shows \"No data.\". As far as I understand dl_manager.download returns original URL when we are calling datasets.get_dataset_split_names and my suspicions are that dataset viewer can do smth similar. I couldn't find information about how it works. I would be very grateful, if you could tell me how to fix this)",
"This is what I get when I try to stream the `raw_txt` subset:\r\n```python\r\n>>> dset = load_dataset(\"MalakhovIlya/RuREBus\", \"raw_txt\", split=\"raw_txt\", streaming=True)\r\n>>> next(iter(dset))\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\nStopIteration\r\n```\r\nSo there is a bug in your script.",
"streaming=True helped me to find solution. I fixed\r\n```\r\ndef extract(zip_file_path):\r\n p = Path(zip_file_path)\r\n dest_dir = str(p.parent / 'extracted' / p.stem)\r\n os.makedirs(dest_dir, exist_ok=True)\r\n with zipfile.ZipFile(zip_file_path) as archive:\r\n for file_info in tqdm(archive.infolist(), desc='Extracting'):\r\n filename = file_info.filename.encode('cp437').decode('cp866')\r\n target = os.path.join(dest_dir, *filename.split('/'))\r\n os.makedirs(os.path.dirname(target), exist_ok=True)\r\n if not file_info.is_dir():\r\n with archive.open(file_info) as source, open(target, 'wb') as dest:\r\n shutil.copyfileobj(source, dest)\r\n return dest_dir\r\n\r\nzip_file = dl_manager.download(self._RAW_TXT_URLS)['raw_txt']\r\nfolder = extract(zip_file)\r\n```\r\nby \r\n```\r\nfolder = dl_manager.download_and_extract(self._RAW_TXT_URLS)['raw_txt']\r\npath = os.path.join(folder, 'MED_txt/unparsed_txt')\r\nfor root, dirs, files in os.walk(path):\r\n decoded_root_name = Path(root).name.encode('cp437').decode('cp866')\r\n```\r\n@mariosasko thank you for your help :)"
] | ## Dataset viewer issue for 'MalakhovIlya/RuREBus'
**Link:** https://huggingface.co/datasets/MalakhovIlya/RuREBus
**Description**
Using os.walk(topdown=False) in DatasetBuilder causes following error:
Status code: 400
Exception: TypeError
Message: xwalk() got an unexpected keyword argument 'topdown'
Couldn't find where "xwalk" come from. How can I fix this?
Am I the one who added this dataset ? Yes
| 4,138 |
https://github.com/huggingface/datasets/issues/4134 | ELI5 supporting documents | [
"Hi ! Please post your question on the [forum](https://discuss.huggingface.co/), more people will be able to help you there ;)"
] | if i am using dense search to create supporting documents for eli5 how much time it will take bcz i read somewhere that it takes about 18 hrs?? | 4,134 |
https://github.com/huggingface/datasets/issues/4133 | HANS dataset preview broken | [
"The dataset cannot be loaded, be it in normal or streaming mode.\r\n\r\n```python\r\n>>> import datasets\r\n>>> dataset=datasets.load_dataset(\"hans\", split=\"train\", streaming=True)\r\n>>> next(iter(dataset))\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/iterable_dataset.py\", line 497, in __iter__\r\n for key, example in self._iter():\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/iterable_dataset.py\", line 494, in _iter\r\n yield from ex_iterable\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/iterable_dataset.py\", line 87, in __iter__\r\n yield from self.generate_examples_fn(**self.kwargs)\r\n File \"/home/slesage/.cache/huggingface/modules/datasets_modules/datasets/hans/1bbcb735c482acd54f2e118074b59cfd2bf5f7a5a285d4d540d1e632216672ac/hans.py\", line 121, in _generate_examples\r\n for idx, line in enumerate(open(filepath, \"rb\")):\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/fsspec/spec.py\", line 1595, in __next__\r\n out = self.readline()\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/fsspec/spec.py\", line 1592, in readline\r\n return self.readuntil(b\"\\n\")\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/fsspec/spec.py\", line 1581, in readuntil\r\n self.seek(start + found + len(char))\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/fsspec/implementations/http.py\", line 676, in seek\r\n raise ValueError(\"Cannot seek streaming HTTP file\")\r\nValueError: Cannot seek streaming HTTP file\r\n>>> dataset=datasets.load_dataset(\"hans\", split=\"train\", streaming=False)\r\nDownloading and preparing dataset hans/plain_text (download: 29.51 MiB, generated: 30.34 MiB, post-processed: Unknown size, total: 59.85 MiB) to /home/slesage/.cache/huggingface/datasets/hans/plain_text/1.0.0/1bbcb735c482acd54f2e118074b59cfd2bf5f7a5a285d4d540d1e632216672ac...\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/load.py\", line 1687, in load_dataset\r\n builder_instance.download_and_prepare(\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/builder.py\", line 605, in download_and_prepare\r\n self._download_and_prepare(\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/builder.py\", line 1104, in _download_and_prepare\r\n super()._download_and_prepare(dl_manager, verify_infos, check_duplicate_keys=verify_infos)\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/builder.py\", line 694, in _download_and_prepare\r\n self._prepare_split(split_generator, **prepare_split_kwargs)\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/datasets/builder.py\", line 1087, in _prepare_split\r\n for key, record in logging.tqdm(\r\n File \"/home/slesage/hf/datasets-preview-backend/.venv/lib/python3.9/site-packages/tqdm/std.py\", line 1180, in __iter__\r\n for obj in iterable:\r\n File \"/home/slesage/.cache/huggingface/modules/datasets_modules/datasets/hans/1bbcb735c482acd54f2e118074b59cfd2bf5f7a5a285d4d540d1e632216672ac/hans.py\", line 121, in _generate_examples\r\n for idx, line in enumerate(open(filepath, \"rb\")):\r\nValueError: readline of closed file\r\n```\r\n\r\n",
"Hi! I've opened a PR that should make this dataset stremable. You can test it as follows:\r\n```python\r\nfrom datasets import load_dataset\r\ndset = load_dataset(\"hans\", split=\"train\", streaming=True, revision=\"49decd29839c792ecc24ac88f861cbdec30c1c40\")\r\n```\r\n\r\n@severo The current script doesn't throw an error in normal mode (only in streaming mode) on my local machine or in Colab. Can you update your installation of `datasets` and see if that fixes the issue?",
"Thanks for this. It works well, thanks! The dataset viewer is using https://github.com/huggingface/datasets/releases/tag/2.0.0, I'm eager to upgrade to 2.0.1 😉"
] | ## Dataset viewer issue for '*hans*'
**Link:** [https://huggingface.co/datasets/hans](https://huggingface.co/datasets/hans)
HANS dataset preview is broken with error 400
Am I the one who added this dataset ? No
| 4,133 |
https://github.com/huggingface/datasets/issues/4129 | dataset metadata for reproducibility | [
"+1 on this idea. This could be powerful for helping better track datasets used for model training and help with automatic model card creation. \r\n\r\nOne possible way of doing this would be to store some/most/all the arguments passed to `load_dataset` if a hub id is passed. i.e. store the Hub ID, configuration, etc. \r\n\r\ncc @tomaarsen"
] | When pulling a dataset from the hub, it would be useful to have some metadata about the specific dataset and version that is used. The metadata could then be passed to the `Trainer` which could then be saved to a model card. This is useful for people who run many experiments on different versions (commits/branches) of the same dataset.
The dataset could have a list of “source datasets” metadata and ignore what happens to them before arriving in the Trainer (i.e. ignore mapping, filtering, etc.).
Here is a basic representation (made by @lhoestq )
```python
>>> from datasets import load_dataset
>>>
>>> my_dataset = load_dataset(...)["train"]
>>> my_dataset = my_dataset.map(...)
>>>
>>> my_dataset.sources
[HFHubDataset(repo_id=..., revision=..., arguments={...})]
``` | 4,129 |
https://github.com/huggingface/datasets/issues/4126 | dataset viewer issue for common_voice | [
"Yes, it's a known issue, and we expect to fix it soon.",
"Fixed.\r\n\r\n<img width=\"1393\" alt=\"Capture d’écran 2022-04-25 à 15 42 05\" src=\"https://user-images.githubusercontent.com/1676121/165101176-d729d85b-efff-45a8-bad1-b69223edba5f.png\">\r\n"
] | ## Dataset viewer issue for 'common_voice'
**Link:** https://huggingface.co/datasets/common_voice
Server Error
Status code: 400
Exception: TypeError
Message: __init__() got an unexpected keyword argument 'audio_column'
Am I the one who added this dataset ? No
| 4,126 |
https://github.com/huggingface/datasets/issues/4124 | Image decoding often fails when transforming Image datasets | [
"A quick hack I have found is that we can call the image first before running the transforms and it makes sure the image is decoded before being passed on.\r\n\r\nFor this I just needed to add `example['img'] = example['img']` to the top of my `generate_flipped_data` function, defined above, so that image decode in invoked.\r\n\r\nAfter this minor change this function works:\r\n```python\r\ndef generate_flipped_data(example, p=0.5):\r\n \"\"\"\r\n A Dataset mapping functions that transforms some of the image up-side-down.\r\n If the probability value (p) is 0.5 approximately half the images will be flipped upside-down\r\n Args:\r\n example: An example from the dataset containing a Python dictionary with \"img\" and \"is_flipped\" key-value pair\r\n p: probability of flipping the image up-side-down, Default 0.5\r\n\r\n Returns:\r\n example: A Dataset object\r\n\r\n \"\"\"\r\n example['img'] = example['img'] # <<< This is the only change\r\n if rng.random() > p: # the flip the image and set is_flipped column to 1\r\n example['img'] = example['img'].transpose(\r\n 1) # ImageOps.flip(example['img']) #example['img'].transpose(Image.FLIP_TOP_BOTTOM)\r\n example['is_flipped'] = 1\r\n\r\n return example\r\n```",
"Hi @RafayAK, thanks for reporting.\r\n\r\nCurrent implementation of the Image feature performs the decoding only if the \"img\" field is accessed by the mapped function.\r\n\r\nIn your original `generate_flipped_data` function:\r\n- it only accesses the \"img\" field (and thus performs decoding) if `rng.random() > p`;\r\n- on the other hand, for the cases where `rng.random() <= p`, the \"img\" field is not accessed and thus no decoding is performed for those examples\r\n\r\nBy adding the code line `example['img'] = example['img']`, you make sure the \"img\" field is accessed in all cases, and the decoding is done for all examples.\r\n\r\nAlso note that there is a little bug in your implementation: `p` is not the probability of flipping, but the probability of not-flipping; the larger is `p`, the smaller is the probability of flipping.\r\n\r\nSome refactoring (fixing also `p`):\r\n```python\r\ndef generate_flipped_data(example, p=0.5):\r\n \"\"\"\r\n A Dataset mapping functions that transforms some of the image up-side-down.\r\n If the probability value (p) is 0.5 approximately half the images will be flipped upside-down.\r\n\r\n Args:\r\n example: An example from the dataset containing a Python dictionary with \"img\" and \"is_flipped\" key-value pair\r\n p: probability of flipping the image up-side-down, Default 0.5\r\n\r\n Returns:\r\n example: A Dataset object\r\n\r\n \"\"\"\r\n do_flip = rng.random() < p # Note the \"<\" sign here instead of \">\"\r\n example['img'] = example['img'].transpose(1) if do_flip else example['img'] # Note \"img\" is always accessed\r\n example['is_flipped'] = 1 if do_flip else 0\r\n return example",
"@albertvillanova Thanks for letting me know this is intended behavior. The docs are severely lacking on this, if I hadn't posted this here I would have never found out how I'm actually supposed to modify images in a Dataset object.",
"@albertvillanova Secondly if you check the error message it shows that around 1999 images were successfully created, I'm pretty sure some of them were also flipped during the process. Back to my main contention, sometimes the decoding takes place other times it fails. \r\n\r\nI suppose to run `map` on any dataset all the examples should be invoked even if on some of them we end up doing nothing, is that right?",
"Hi @RafayAK! I've opened a PR with the fix, which adds a fallback to reattempt casting to PyArrow format with a more robust (but more expensive) procedure if the first attempt fails. Feel free to test it by installing `datasets` from the PR branch with the following command:\r\n```\r\npip install git+https://github.com/huggingface/datasets.git@fix-4124\r\n```",
"@mariosasko I'll try this right away and report back.",
"@mariosasko Thanks a lot for looking into this, now the `map` function at least behaves as one would expect a function to behave. \r\n\r\nLooking forward to exploring Hugging Face more and even contributing 😃.\r\n\r\n```bash\r\n $ conda list | grep datasets\r\ndatasets 2.0.1.dev0 pypi_0 pypi\r\n\r\n```\r\n\r\n```python\r\ndef preprocess_data(dataset):\r\n \"\"\"\r\n Helper funtion to pre-process HuggingFace Cifar-100 Dataset to remove fine_label and coarse_label columns and\r\n add is_flipped column\r\n Args:\r\n dataset: HuggingFace CIFAR-100 Dataset Object\r\n\r\n Returns:\r\n new_dataset: A Dataset object with \"img\" and \"is_flipped\" columns only\r\n\r\n \"\"\"\r\n # remove fine_label and coarse_label columns\r\n new_dataset = dataset.remove_columns(['fine_label', 'coarse_label'])\r\n # add the column for is_flipped\r\n new_dataset = new_dataset.add_column(name=\"is_flipped\", column=np.zeros((len(new_dataset)), dtype=np.uint8))\r\n\r\n return new_dataset\r\n\r\n\r\ndef generate_flipped_data(example, p=0.5):\r\n \"\"\"\r\n A Dataset mapping functions that transforms some of the image up-side-down.\r\n If the probability value (p) is 0.5 approximately half the images will be flipped upside-down\r\n Args:\r\n example: An example from the dataset containing a Python dictionary with \"img\" and \"is_flipped\" key-value pair\r\n p: probability of flipping the image up-side-down, Default 0.5\r\n\r\n Returns:\r\n example: A Dataset object\r\n\r\n \"\"\"\r\n # example['img'] = example['img']\r\n if rng.random() > p: # the flip the image and set is_flipped column to 1\r\n example['img'] = example['img'].transpose(\r\n 1) # ImageOps.flip(example['img']) #example['img'].transpose(Image.FLIP_TOP_BOTTOM)\r\n example['is_flipped'] = 1\r\n\r\n return example\r\n\r\nmy_test = preprocess_data(test_dataset)\r\nmy_test = my_test.map(generate_flipped_data)\r\n```\r\n\r\nThe output now show the function was applied successfully:\r\n``` bash\r\n/home/rafay/anaconda3/envs/pytorch_new/bin/python /home/rafay/Documents/you_only_live_once/upside_down_detector/create_dataset.py\r\nDownloading builder script: 5.61kB [00:00, 3.16MB/s] \r\nDownloading metadata: 4.21kB [00:00, 2.56MB/s] \r\nReusing dataset cifar100 (/home/rafay/.cache/huggingface/datasets/cifar100/cifar100/1.0.0/f365c8b725c23e8f0f8d725c3641234d9331cd2f62919d1381d1baa5b3ba3142)\r\nReusing dataset cifar100 (/home/rafay/.cache/huggingface/datasets/cifar100/cifar100/1.0.0/f365c8b725c23e8f0f8d725c3641234d9331cd2f62919d1381d1baa5b3ba3142)\r\n100%|██████████| 10000/10000 [00:01<00:00, 5149.15ex/s]\r\n```\r\n"
] | ## Describe the bug
When transforming/modifying images in an image dataset using the `map` function the PIL images often fail to decode in time for the image transforms, causing errors.
Using a debugger it is easy to see what the problem is, the Image decode invocation does not take place and the resulting image passed around is still raw bytes:
```
[{'bytes': b'\x89PNG\r\n\x1a\n\x00\x00\x00\rIHDR\x00\x00\x00 \x00\x00\x00 \x08\x02\x00\x00\x00\xfc\x18\xed\xa3\x00\x00\x08\x02IDATx\x9cEVIs[\xc7\x11\xeemf\xde\x82\x8d\x80\x08\x89"\xb5V\\\xb6\x94(\xe5\x9f\x90\xca5\x7f$\xa7T\xe5\x9f&9\xd9\x8a\\.\xdb\xa4$J\xa4\x00\x02x\xc0{\xb3t\xe7\x00\xca\x99\xd3\\f\xba\xba\xbf\xa5?|\xfa\xf4\xa2\xeb\xba\xedv\xa3f^\xf8\xd5\x0bY\xb6\x10\xb3\xaaDq\xcd\x83\x87\xdf5\xf3gZ\x1a\x04\x0f\xa0fp\xfa\xe0\xd4\x07?\x9dN\xc4\xb1\x99\xfd\xf2\xcb/\x97\x97\x97H\xa2\xaaf\x16\x82\xaf\xeb\xca{\xbf\xd9l.\xdf\x7f\xfa\xcb_\xff&\x88\x08\x00\x80H\xc0\x80@.;\x0f\x8c@#v\xe3\xe5\xfc\xd1\x9f\xee6q\xbf\xdf\xa6\x14\'\x93\xf1\xc3\xe5\xe3\xd1x\x14c\x8c1\xa5\x1c\x9dsM\xd3\xb4\xed\x08\x89SJ)\xa5\xedv\xbb^\xafNO\x97D\x84Hf ....
```
## Steps to reproduce the bug
```python
from datasets import load_dataset, Dataset
import numpy as np
# seeded NumPy random number generator for reprodducinble results.
rng = np.random.default_rng(seed=0)
test_dataset = load_dataset('cifar100', split="test")
def preprocess_data(dataset):
"""
Helper function to pre-process HuggingFace Cifar-100 Dataset to remove fine_label and coarse_label columns and
add is_flipped column
Args:
dataset: HuggingFace CIFAR-100 Dataset Object
Returns:
new_dataset: A Dataset object with "img" and "is_flipped" columns only
"""
# remove fine_label and coarse_label columns
new_dataset = dataset.remove_columns(['fine_label', 'coarse_label'])
# add the column for is_flipped
new_dataset = new_dataset.add_column(name="is_flipped", column=np.zeros((len(new_dataset)), dtype=np.uint8))
return new_dataset
def generate_flipped_data(example, p=0.5):
"""
A Dataset mapping function that transforms some of the images up-side-down.
If the probability value (p) is 0.5 approximately half the images will be flipped upside-down
Args:
example: An example from the dataset containing a Python dictionary with "img" and "is_flipped" key-value pair
p: the probability of flipping the image up-side-down, Default 0.5
Returns:
example: A Dataset object
"""
# example['img'] = example['img']
if rng.random() > p: # the flip the image and set is_flipped column to 1
example['img'] = example['img'].transpose(
1) # ImageOps.flip(example['img']) #example['img'].transpose(Image.FLIP_TOP_BOTTOM)
example['is_flipped'] = 1
return example
my_test = preprocess_data(test_dataset)
my_test = my_test.map(generate_flipped_data)
```
## Expected results
The dataset should be transformed without problems.
## Actual results
```
/home/rafay/anaconda3/envs/pytorch_new/bin/python /home/rafay/Documents/you_only_live_once/upside_down_detector/create_dataset.py
Reusing dataset cifar100 (/home/rafay/.cache/huggingface/datasets/cifar100/cifar100/1.0.0/f365c8b725c23e8f0f8d725c3641234d9331cd2f62919d1381d1baa5b3ba3142)
Reusing dataset cifar100 (/home/rafay/.cache/huggingface/datasets/cifar100/cifar100/1.0.0/f365c8b725c23e8f0f8d725c3641234d9331cd2f62919d1381d1baa5b3ba3142)
20%|█▉ | 1999/10000 [00:00<00:01, 5560.44ex/s]
Traceback (most recent call last):
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 2326, in _map_single
writer.write(example)
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_writer.py", line 441, in write
self.write_examples_on_file()
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_writer.py", line 399, in write_examples_on_file
self.write_batch(batch_examples=batch_examples)
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_writer.py", line 492, in write_batch
arrays.append(pa.array(typed_sequence))
File "pyarrow/array.pxi", line 230, in pyarrow.lib.array
File "pyarrow/array.pxi", line 110, in pyarrow.lib._handle_arrow_array_protocol
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_writer.py", line 185, in __arrow_array__
out = pa.array(cast_to_python_objects(data, only_1d_for_numpy=True))
File "pyarrow/array.pxi", line 316, in pyarrow.lib.array
File "pyarrow/array.pxi", line 39, in pyarrow.lib._sequence_to_array
File "pyarrow/error.pxi", line 143, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 99, in pyarrow.lib.check_status
pyarrow.lib.ArrowInvalid: Could not convert <PIL.Image.Image image mode=RGB size=32x32 at 0x7F56AEE61DE0> with type Image: did not recognize Python value type when inferring an Arrow data type
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/home/rafay/Documents/you_only_live_once/upside_down_detector/create_dataset.py", line 55, in <module>
my_test = my_test.map(generate_flipped_data)
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 1953, in map
return self._map_single(
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 519, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 486, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/fingerprint.py", line 458, in wrapper
out = func(self, *args, **kwargs)
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 2360, in _map_single
writer.finalize()
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_writer.py", line 522, in finalize
self.write_examples_on_file()
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_writer.py", line 399, in write_examples_on_file
self.write_batch(batch_examples=batch_examples)
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_writer.py", line 492, in write_batch
arrays.append(pa.array(typed_sequence))
File "pyarrow/array.pxi", line 230, in pyarrow.lib.array
File "pyarrow/array.pxi", line 110, in pyarrow.lib._handle_arrow_array_protocol
File "/home/rafay/anaconda3/envs/pytorch_new/lib/python3.10/site-packages/datasets/arrow_writer.py", line 185, in __arrow_array__
out = pa.array(cast_to_python_objects(data, only_1d_for_numpy=True))
File "pyarrow/array.pxi", line 316, in pyarrow.lib.array
File "pyarrow/array.pxi", line 39, in pyarrow.lib._sequence_to_array
File "pyarrow/error.pxi", line 143, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 99, in pyarrow.lib.check_status
pyarrow.lib.ArrowInvalid: Could not convert <PIL.Image.Image image mode=RGB size=32x32 at 0x7F56AEE61DE0> with type Image: did not recognize Python value type when inferring an Arrow data type
Process finished with exit code 1
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.0.0
- Platform: Linux(Fedora 35)
- Python version: 3.10
- PyArrow version: 7.0.0
| 4,124 |
https://github.com/huggingface/datasets/issues/4123 | Building C4 takes forever | [
"Hi @StellaAthena, thanks for reporting.\r\n\r\nPlease note, that our `datasets` library performs several operations in order to load a dataset, among them:\r\n- it downloads all the required files: for C4 \"en\", 378.69 GB of JSON GZIPped files\r\n- it parses their content to generate the dataset\r\n- it caches the dataset in an Arrow file: for C4 \"en\", this file size is 1.87 TB\r\n- it memory-maps the Arrow file\r\n\r\nIf it suits your use case, you might load this dataset in streaming mode:\r\n- no Arrow file is generated\r\n- you can iterate over elements immediately (no need to wait to download all the entire files)\r\n\r\n```python\r\nIn [45]: from datasets import load_dataset\r\n ...: ds = load_dataset(\"c4\", \"en\", split=\"train\", streaming=True)\r\n ...: for item in ds:\r\n ...: print(item)\r\n ...: break\r\n ...: \r\n{'text': 'Beginners BBQ Class Taking Place in Missoula!\\nDo you want to get better at making delicious BBQ? You will have the opportunity, put this on your calendar now. Thursday, September 22nd join World Class BBQ Champion, Tony Balay from Lonestar Smoke Rangers. He will be teaching a beginner level class for everyone who wants to get better with their culinary skills.\\nHe will teach you everything you need to know to compete in a KCBS BBQ competition, including techniques, recipes, timelines, meat selection and trimming, plus smoker and fire information.\\nThe cost to be in the class is $35 per person, and for spectators it is free. Included in the cost will be either a t-shirt or apron and you will be tasting samples of each meat that is prepared.', 'timestamp': '2019-04-25T12:57:54Z', 'url': 'https://klyq.com/beginners-bbq-class-taking-place-in-missoula/'}\r\n```\r\nI hope this is useful for your use case."
] | ## Describe the bug
C4-en is a 300 GB dataset. However, when I try to download it through the hub it takes over _six hours_ to generate the train/test split from the downloaded files. This is an absurd amount of time and an unnecessary waste of resources.
## Steps to reproduce the bug
```python
c4 = datasets.load("c4", "en")
```
## Expected results
I would like to be able to download pre-split data.
## Environment info
- `datasets` version: 2.0.0
- Platform: Linux-5.13.0-35-generic-x86_64-with-glibc2.29
- Python version: 3.8.10
- PyArrow version: 7.0.0
- Pandas version: 1.4.1
| 4,123 |
https://github.com/huggingface/datasets/issues/4122 | medical_dialog zh has very slow _generate_examples | [
"Hi @nbroad1881, thanks for reporting.\r\n\r\nLet me have a look to try to improve its performance. ",
"Thanks @nbroad1881 for reporting! I don't recall it taking so long. I will also have a look at this. \r\n@albertvillanova please let me know if I am doing something unnecessary or time consuming.",
"Hi @nbroad1881 and @vrindaprabhu,\r\n\r\nAs a workaround for the performance of the parsing of the raw data files (this could be addressed in a subsequent PR), I have found that there are also processed data files, that do not require parsing. I have added these as new configurations `processed.en` and `processed.zh`:\r\n```python\r\nds = load_dataset(\"medical_dialog\", \"processed.zh\")\r\n```"
] | ## Describe the bug
After downloading the files from Google Drive, `load_dataset("medical_dialog", "zh", data_dir="./")` takes an unreasonable amount of time. Generating the train/test split for 33% of the dataset takes over 4.5 hours.
## Steps to reproduce the bug
The easiest way I've found to download files from Google Drive is to use `gdown` and use Google Colab because the download speeds will be very high due to the fact that they are both in Google Cloud.
```python
file_ids = [
"1AnKxGEuzjeQsDHHqL3NqI_aplq2hVL_E",
"1tt7weAT1SZknzRFyLXOT2fizceUUVRXX",
"1A64VBbsQ_z8wZ2LDox586JIyyO6mIwWc",
"1AKntx-ECnrxjB07B6BlVZcFRS4YPTB-J",
"1xUk8AAua_x27bHUr-vNoAuhEAjTxOvsu",
"1ezKTfe7BgqVN5o-8Vdtr9iAF0IueCSjP",
"1tA7bSOxR1RRNqZst8cShzhuNHnayUf7c",
"1pA3bCFA5nZDhsQutqsJcH3d712giFb0S",
"1pTLFMdN1A3ro-KYghk4w4sMz6aGaMOdU",
"1dUSnG0nUPq9TEQyHd6ZWvaxO0OpxVjXD",
"1UfCH05nuWiIPbDZxQzHHGAHyMh8dmPQH",
]
for i in file_ids:
url = f"https://drive.google.com/uc?id={i}"
!gdown $url
from datasets import load_dataset
ds = load_dataset("medical_dialog", "zh", data_dir="./")
```
## Expected results
Faster load time
## Actual results
`Generating train split: 33%: 625519/1921127 [4:31:03<31:39:20, 11.37 examples/s]`
## Environment info
- `datasets` version: 2.0.0
- Platform: Linux-5.4.144+-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.13
- PyArrow version: 6.0.1
- Pandas version: 1.3.5
@vrindaprabhu , could you take a look at this since you implemented it? I think the `_generate_examples` function might need to be rewritten | 4,122 |
https://github.com/huggingface/datasets/issues/4121 | datasets.load_metric can not load a local metirc | [
"Hello, could you tell me how this issue can be fixed? I'm coming across the same issue."
] | ## Describe the bug
No matter how I hard try to tell load_metric that I want to load a local metric file, it still continues to fetch things on the Internet. And unfortunately it says 'ConnectionError: Couldn't reach'. However I can download this file without connectionerror and tell load_metric its local directory. And it comes back where it begins...
## Steps to reproduce the bug
```python
metric = load_metric(path=r'C:\Users\Gare\PycharmProjects\Gare\blue\bleu.py')
ConnectionError: Couldn't reach https://github.com/tensorflow/nmt/raw/master/nmt/scripts/bleu.py
metric = load_metric(path='bleu')
ConnectionError: Couldn't reach https://raw.githubusercontent.com/huggingface/datasets/1.12.1/metrics/bleu/bleu.py
metric = load_metric(path='./blue/bleu.py')
ConnectionError: Couldn't reach https://github.com/tensorflow/nmt/raw/master/nmt/scripts/bleu.py
```
## Expected results
I do read the docs [here](https://huggingface.co/docs/datasets/package_reference/loading_methods#datasets.load_metric). There are no other parameters that help function to distinguish from local and online file but path. As what I code above, it should load from local.
## Actual results
> metric = load_metric(path=r'C:\Users\Gare\PycharmProjects\Gare\blue\bleu.py')
> ~\AppData\Local\Temp\ipykernel_19636\1855752034.py in <module>
----> 1 metric = load_metric(path=r'C:\Users\Gare\PycharmProjects\Gare\blue\bleu.py')
D:\Program Files\Anaconda\envs\Gare\lib\site-packages\datasets\load.py in load_metric(path, config_name, process_id, num_process, cache_dir, experiment_id, keep_in_memory, download_config, download_mode, script_version, **metric_init_kwargs)
817 if data_files is None and data_dir is not None:
818 data_files = os.path.join(data_dir, "**")
--> 819
820 self.name = name
821 self.revision = revision
D:\Program Files\Anaconda\envs\Gare\lib\site-packages\datasets\load.py in prepare_module(path, script_version, download_config, download_mode, dataset, force_local_path, dynamic_modules_path, return_resolved_file_path, return_associated_base_path, data_files, **download_kwargs)
639 self,
640 path: str,
--> 641 download_config: Optional[DownloadConfig] = None,
642 download_mode: Optional[DownloadMode] = None,
643 dynamic_modules_path: Optional[str] = None,
D:\Program Files\Anaconda\envs\Gare\lib\site-packages\datasets\utils\file_utils.py in cached_path(url_or_filename, download_config, **download_kwargs)
297 token = hf_api.HfFolder.get_token()
298 if token:
--> 299 headers["authorization"] = f"Bearer {token}"
300 return headers
301
D:\Program Files\Anaconda\envs\Gare\lib\site-packages\datasets\utils\file_utils.py in get_from_cache(url, cache_dir, force_download, proxies, etag_timeout, resume_download, user_agent, local_files_only, use_etag, max_retries, use_auth_token)
604 def _resumable_file_manager():
605 with open(incomplete_path, "a+b") as f:
--> 606 yield f
607
608 temp_file_manager = _resumable_file_manager
ConnectionError: Couldn't reach https://github.com/tensorflow/nmt/raw/master/nmt/scripts/bleu.py
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.0.0
- Platform: Windows-10-10.0.22000-SP0
- Python version: 3.7.13
- PyArrow version: 7.0.0
- Pandas version: 1.3.4
Any advice would be appreciated. | 4,121 |
https://github.com/huggingface/datasets/issues/4120 | Representing dictionaries (json) objects as features | [] | In the process of adding a new dataset to the hub, I stumbled upon the inability to represent dictionaries that contain different key names, unknown in advance (and may differ between samples), original asked in the [forum](https://discuss.huggingface.co/t/representing-nested-dictionary-with-different-keys/16442).
For instance:
```
sample1 = {"nps": {
"a": {"id": 0, "text": "text1"},
"b": {"id": 1, "text": "text2"},
}}
sample2 = {"nps": {
"a": {"id": 0, "text": "text1"},
"b": {"id": 1, "text": "text2"},
"c": {"id": 2, "text": "text3"},
}}
sample3 = {"nps": {
"a": {"id": 0, "text": "text1"},
"b": {"id": 1, "text": "text2"},
"c": {"id": 2, "text": "text3"},
"d": {"id": 3, "text": "text4"},
}}
```
the `nps` field cannot be represented as a Feature while maintaining its original structure.
@lhoestq suggested to add JSON as a new feature type, which will solve this problem.
It seems like an alternative solution would be to change the original data format, which isn't an optimal solution in my case. Moreover, JSON is a common structure, that will likely to be useful in future datasets as well. | 4,120 |
https://github.com/huggingface/datasets/issues/4118 | Failing CI tests on Windows | [] | ## Describe the bug
Our CI Windows tests are failing from yesterday: https://app.circleci.com/pipelines/github/huggingface/datasets/11092/workflows/9cfdb1dd-0fec-4fe0-8122-5f533192ebdc/jobs/67414
| 4,118 |
https://github.com/huggingface/datasets/issues/4117 | AttributeError: module 'huggingface_hub' has no attribute 'hf_api' | [
"Hi @arymbe, thanks for reporting.\r\n\r\nUnfortunately, I'm not able to reproduce your problem.\r\n\r\nCould you please write the complete stack trace? That way we will be able to see which package originates the exception.",
"Hello, thank you for your fast replied. this is the complete error that I got\r\n\r\n---------------------------------------------------------------------------\r\n\r\nAttributeError Traceback (most recent call last)\r\n\r\n---------------------------------------------------------------------------\r\n\r\nAttributeError Traceback (most recent call last)\r\n\r\nInput In [27], in <module>\r\n----> 1 from datasets import load_dataset\r\n\r\nvenv/lib/python3.8/site-packages/datasets/__init__.py:39, in <module>\r\n 37 from .arrow_dataset import Dataset, concatenate_datasets\r\n 38 from .arrow_reader import ReadInstruction\r\n---> 39 from .builder import ArrowBasedBuilder, BeamBasedBuilder, BuilderConfig, DatasetBuilder, GeneratorBasedBuilder\r\n 40 from .combine import interleave_datasets\r\n 41 from .dataset_dict import DatasetDict, IterableDatasetDict\r\n\r\nvenv/lib/python3.8/site-packages/datasets/builder.py:40, in <module>\r\n 32 from .arrow_reader import (\r\n 33 HF_GCP_BASE_URL,\r\n 34 ArrowReader,\r\n (...)\r\n 37 ReadInstruction,\r\n 38 )\r\n 39 from .arrow_writer import ArrowWriter, BeamWriter\r\n---> 40 from .data_files import DataFilesDict, sanitize_patterns\r\n 41 from .dataset_dict import DatasetDict, IterableDatasetDict\r\n 42 from .features import Features\r\n\r\nvenv/lib/python3.8/site-packages/datasets/data_files.py:297, in <module>\r\n 292 except FileNotFoundError:\r\n 293 raise FileNotFoundError(f\"The directory at {base_path} doesn't contain any data file\") from None\r\n 296 def _resolve_single_pattern_in_dataset_repository(\r\n--> 297 dataset_info: huggingface_hub.hf_api.DatasetInfo,\r\n 298 pattern: str,\r\n 299 allowed_extensions: Optional[list] = None,\r\n 300 ) -> List[PurePath]:\r\n 301 data_files_ignore = FILES_TO_IGNORE\r\n 302 fs = HfFileSystem(repo_info=dataset_info)\r\n\r\nAttributeError: module 'huggingface_hub' has no attribute 'hf_api'",
"This is weird... It is long ago that the package `huggingface_hub` has a submodule called `hf_api`.\r\n\r\nMaybe you have a problem with your installed `huggingface_hub`...\r\n\r\nCould you please try to update it?\r\n```shell\r\npip install -U huggingface_hub\r\n```",
"Yap, I've updated several times. Then, I've tried numeral combination of datasets and huggingface_hub versions. However, I think your point is right that there is a problem with my huggingface_hub installation. I'll try another way to find the solution. I'll update it later when I get the solution. Thank you :)",
"I'm sorry I can't reproduce your problem.\r\n\r\nMaybe you could try to create a new Python virtual environment and install all dependencies there from scratch. You can use either:\r\n- Python venv: https://docs.python.org/3/library/venv.html\r\n- or conda venv (if you are using conda): https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html",
"Facing the same issue.\r\n\r\nResponse from `pip show datasets`\r\n```\r\nName: datasets\r\nVersion: 1.15.1\r\nSummary: HuggingFace community-driven open-source library of datasets\r\nHome-page: https://github.com/huggingface/datasets\r\nAuthor: HuggingFace Inc.\r\nAuthor-email: [email protected]\r\nLicense: Apache 2.0\r\nLocation: /usr/local/lib/python3.8/dist-packages\r\nRequires: aiohttp, dill, fsspec, huggingface-hub, multiprocess, numpy, packaging, pandas, pyarrow, requests, tqdm, xxhash\r\nRequired-by: lm-eval\r\n```\r\n\r\nResponse from `pip show huggingface_hub`\r\n\r\n```\r\nName: huggingface-hub\r\nVersion: 0.8.1\r\nSummary: Client library to download and publish models, datasets and other repos on the huggingface.co hub\r\nHome-page: https://github.com/huggingface/huggingface_hub\r\nAuthor: Hugging Face, Inc.\r\nAuthor-email: [email protected]\r\nLicense: Apache\r\nLocation: /usr/local/lib/python3.8/dist-packages\r\nRequires: filelock, packaging, pyyaml, requests, tqdm, typing-extensions\r\nRequired-by: datasets\r\n```\r\n\r\nresponse from `datasets-cli env`\r\n\r\n```\r\nTraceback (most recent call last):\r\n File \"/usr/local/bin/datasets-cli\", line 5, in <module>\r\n from datasets.commands.datasets_cli import main\r\n File \"/usr/local/lib/python3.8/dist-packages/datasets/__init__.py\", line 37, in <module>\r\n from .builder import ArrowBasedBuilder, BeamBasedBuilder, BuilderConfig, DatasetBuilder, GeneratorBasedBuilder\r\n File \"/usr/local/lib/python3.8/dist-packages/datasets/builder.py\", line 44, in <module>\r\n from .data_files import DataFilesDict, _sanitize_patterns\r\n File \"/usr/local/lib/python3.8/dist-packages/datasets/data_files.py\", line 120, in <module>\r\n dataset_info: huggingface_hub.hf_api.DatasetInfo,\r\n File \"/usr/local/lib/python3.8/dist-packages/huggingface_hub/__init__.py\", line 105, in __getattr__\r\n raise AttributeError(f\"No {package_name} attribute {name}\")\r\nAttributeError: No huggingface_hub attribute hf_api\r\n```",
"A workaround: \r\nI changed lines around Line 125 in `__init__.py` of `huggingface_hub` to something like\r\n```\r\n__getattr__, __dir__, __all__ = _attach(\r\n __name__,\r\n submodules=['hf_api'],\r\n```\r\nand it works ( which gives `datasets` direct access to `huggingface_hub.hf_api` ).",
"I was getting the same issue. After trying a few versions, following combination worked for me.\r\ndataset==2.3.2\r\nhuggingface_hub==0.7.0\r\n\r\nIn another environment, I just installed latest repos from pip through `pip install -U transformers datasets tokenizers evaluate`, resulting in following versions. This also worked. Hope it helps someone. \r\n\r\ndatasets-2.3.2 evaluate-0.1.2 huggingface-hub-0.8.1 responses-0.18.0 tokenizers-0.12.1 transformers-4.20.1",
"For layoutlm_v3 finetune\r\ndatasets-2.3.2 evaluate-0.1.2 huggingface-hub-0.8.1 responses-0.18.0 tokenizers-0.12.1 transformers-4.12.5",
"(For layoutlmv3 fine-tuning) In my case, modifying `requirements.txt` as below worked.\r\n\r\n- python = 3.7\r\n\r\n```\r\ndatasets==2.3.2\r\nevaluate==0.1.2\r\nhuggingface-hub==0.8.1\r\nresponse==0.5.0\r\ntokenizers==0.10.1\r\ntransformers==4.12.5\r\nseqeval==1.2.2\r\ndeepspeed==0.5.7\r\ntensorboard==2.7.0\r\nseqeval==1.2.2\r\nsentencepiece\r\ntimm==0.4.12\r\nPillow\r\neinops\r\ntextdistance\r\nshapely\r\n```",
"> For layoutlm_v3 finetune datasets-2.3.2 evaluate-0.1.2 huggingface-hub-0.8.1 responses-0.18.0 tokenizers-0.12.1 transformers-4.12.5\r\n\r\nGOOD!! Thanks!",
"I encountered the same issue where the problem is the absence of the 'scipy' library.\r\nTo solve this open your terminal or command prompt and run the following command to install 'scipy': pip install scipy .\r\nRestart the kernel and rerun the cell and it will work.\r\n",
"> I was getting the same issue. After trying a few versions, following combination worked for me. dataset==2.3.2 huggingface_hub==0.7.0\r\n> \r\n> In another environment, I just installed latest repos from pip through `pip install -U transformers datasets tokenizers evaluate`, resulting in following versions. This also worked. Hope it helps someone.\r\n> \r\n> datasets-2.3.2 evaluate-0.1.2 huggingface-hub-0.8.1 responses-0.18.0 tokenizers-0.12.1 transformers-4.20.1\r\n\r\n\r\n\r\n> I was getting the same issue. After trying a few versions, following combination worked for me. dataset==2.3.2 huggingface_hub==0.7.0\r\n> \r\n> In another environment, I just installed latest repos from pip through `pip install -U transformers datasets tokenizers evaluate`, resulting in following versions. This also worked. Hope it helps someone.\r\n> \r\n> datasets-2.3.2 evaluate-0.1.2 huggingface-hub-0.8.1 responses-0.18.0 tokenizers-0.12.1 transformers-4.20.1\r\n\r\nI face with the same issue. After using your approach I solve the issue.Thank you very much.\r\n`pip install -U datasets`"
] | ## Describe the bug
Could you help me please. I got this following error.
AttributeError: module 'huggingface_hub' has no attribute 'hf_api'
## Steps to reproduce the bug
when I imported the datasets
# Sample code to reproduce the bug
from datasets import list_datasets, load_dataset, list_metrics, load_metric
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.0.0
- Platform: macOS-12.3-x86_64-i386-64bit
- Python version: 3.8.9
- PyArrow version: 7.0.0
- Pandas version: 1.3.5
- Huggingface-hub: 0.5.0
- Transformers: 4.18.0
Thank you in advance. | 4,117 |
https://github.com/huggingface/datasets/issues/4115 | ImageFolder add option to ignore some folders like '.ipynb_checkpoints' | [
"Maybe it would be nice to ignore private dirs like this one (ones starting with `.`) by default. \r\n\r\nCC @mariosasko ",
"Maybe we can add a `ignore_hidden_files` flag to the builder configs of our packaged loaders (to be consistent across all of them), wdyt @lhoestq @albertvillanova? ",
"I think they should always ignore them actually ! Not sure if adding a flag would be helpful",
"@lhoestq But what if the user explicitly requests those files via regex?\r\n\r\n`glob.glob` ignores hidden files (files starting with \".\") by default unless they are explicitly requested, but fsspec's `glob` doesn't follow this behavior, which is probably a bug, so maybe we can raise an issue or open a PR in their repo?",
"> @lhoestq But what if the user explicitly requests those files via regex?\r\n\r\nUsually hidden files are meant to be ignored. If they are data files, they must be placed outside a hidden directory in the first place right ? I think it's more sensible to explain this than adding a flag.\r\n\r\n> glob.glob ignores hidden files (files starting with \".\") by default unless they are explicitly requested, but fsspec's glob doesn't follow this behavior, which is probably a bug, so maybe we can raise an issue or open a PR in their repo?\r\n\r\nAfter globbing using `fsspec`, we already ignore files that start with a `.` in `_resolve_single_pattern_locally` and `_resolve_single_pattern_in_dataset_repository`, I guess we can just account for parent directories as well ?\r\n\r\nWe could open an issue on `fsspec` but I think they won't change this since it's an important breaking change for them."
] | **Is your feature request related to a problem? Please describe.**
I sometimes like to peek at the dataset images from jupyterlab. thus '.ipynb_checkpoints' folder appears where my dataset is and (just realized) leads to accidental duplicate image additions. I think this is an easy enough thing to miss especially if the dataset is very large.
**Describe the solution you'd like**
maybe have an option `ignore` or something .gitignore style
`dataset = load_dataset("imagefolder", data_dir="./data/original", ignore="regex?")`
**Describe alternatives you've considered**
Could filter out manually
| 4,115 |
https://github.com/huggingface/datasets/issues/4114 | Allow downloading just some columns of a dataset | [
"In the general case you can’t always reduce the quantity of data to download, since you can’t parse CSV or JSON data without downloading the whole files right ? ^^ However we could explore this case-by-case I guess",
"Actually for csv pandas has `usecols` which allows loading a subset of columns in a more efficient way afaik, but yes, you're right this might be more complex than I thought.",
"Bumping the visibility of this :) Is there a recommended way of doing this?",
"Passing `columns=[...]` to `load_dataset()` in streaming mode does work if the dataset is in Parquet format, but for other formats it's either not possible or not implemented",
"I tried using the `columns=['bambara']` on this dataset `oza75/bambara-tts` which is in parquet, but it does not work. This feature is really useful because sometimes you don't want to download the whole dataset but just a few columns.",
"It doesn't work for the dataset with `parquet` format. Are we missing something?",
"It only works for `streaming=True`. When not streaming it does download the full files locally before reading the data",
"Hi @lhoestq, I have an audio dataset of 250GB on the huggingface hub in parquet format. I only wanted to load the text column. It is taking a lot of time. It seems like it is downloading audio as well even in streaming mode. "
] | **Is your feature request related to a problem? Please describe.**
Some people are interested in doing label analysis of a CV dataset without downloading all the images. Downloading the whole dataset does not always makes sense for this kind of use case
**Describe the solution you'd like**
Be able to just download some columns of a dataset, such as doing
```python
load_dataset("huggan/wikiart",columns=["artist", "genre"])
```
Although this might make things a bit complicated in terms of local caching of datasets. | 4,114 |
https://github.com/huggingface/datasets/issues/4113 | Multiprocessing with FileLock fails in python 3.9 | [
"Closing this one because it must be used this way actually:\r\n```python\r\ndef main():\r\n with FileLock(\"tmp.lock\"):\r\n with Pool(2) as pool:\r\n pool.map(run, range(2))\r\n\r\nif __name__ == \"__main__\":\r\n main()\r\n```"
] | On python 3.9, this code hangs:
```python
from multiprocessing import Pool
from filelock import FileLock
def run(i):
print(f"got the lock in multi process [{i}]")
with FileLock("tmp.lock"):
with Pool(2) as pool:
pool.map(run, range(2))
```
This is because the subprocesses try to acquire the lock from the main process for some reason. This is not the case in older versions of python.
This can cause many issues in python 3.9. In particular, we use multiprocessing to fetch data files when you load a dataset (as long as there are >16 data files). Therefore `imagefolder` hangs, and I expect any dataset that needs to download >16 files to hang as well.
Let's see if we can fix this and have a CI that runs on 3.9.
cc @mariosasko @julien-c | 4,113 |
https://github.com/huggingface/datasets/issues/4112 | ImageFolder with Grayscale images dataset | [
"Hi! Replacing:\r\n```python\r\ntransformed_dataset = dataset.with_transform(transforms)\r\ntransformed_dataset.set_format(type=\"torch\", device=\"cuda\")\r\n```\r\n\r\nwith:\r\n```python\r\ndef transform_func(examples):\r\n examples[\"image\"] = [transforms(img).to(\"cuda\") for img in examples[\"image\"]]\r\n return examples\r\n\r\ntransformed_dataset = dataset.with_transform(transform_func)\r\n```\r\nshould fix the issue. `datasets` doesn't support chaining of transforms (you can think of `set_format`/`with_format` as a predefined transform func for `set_transform`/`with_transforms`), so the last transform (in your case, `set_format`) takes precedence over the previous ones (in your case `with_format`). And the PyTorch formatter is not supported by the Image feature, hence the error (adding support for that is on our short-term roadmap).",
"Ok thanks a lot for the code snippet!\r\n\r\nI love the way `datasets` is easy to use but it made it really long to pre-process all the images (400.000 in my case) before training anything. `ImageFolder` from pytorch is faster in my case but force me to have the images on my local machine.\r\n\r\nI don't know how to speed up the process without switching to `ImageFolder` :smile: ",
"You can pass `ignore_verifications=True` in `load_dataset` to skip checksum verification, which takes a lot of time if the number of files is large. We will consider making this the default behavior."
] | Hi, I'm facing a problem with a grayscale images dataset I have uploaded [here](https://huggingface.co/datasets/ChainYo/rvl-cdip) (RVL-CDIP)
I'm getting an error while I want to use images for training a model with PyTorch DataLoader. Here is the full traceback:
```bash
AttributeError: Caught AttributeError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/torch/utils/data/_utils/worker.py", line 287, in _worker_loop
data = fetcher.fetch(index)
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 49, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 49, in <listcomp>
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 1765, in __getitem__
return self._getitem(
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 1750, in _getitem
formatted_output = format_table(
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/formatting/formatting.py", line 532, in format_table
return formatter(pa_table, query_type=query_type)
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/formatting/formatting.py", line 281, in __call__
return self.format_row(pa_table)
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/formatting/torch_formatter.py", line 58, in format_row
return self.recursive_tensorize(row)
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/formatting/torch_formatter.py", line 54, in recursive_tensorize
return map_nested(self._recursive_tensorize, data_struct, map_list=False)
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/utils/py_utils.py", line 314, in map_nested
mapped = [
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/utils/py_utils.py", line 315, in <listcomp>
_single_map_nested((function, obj, types, None, True, None))
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/utils/py_utils.py", line 267, in _single_map_nested
return {k: _single_map_nested((function, v, types, None, True, None)) for k, v in pbar}
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/utils/py_utils.py", line 267, in <dictcomp>
return {k: _single_map_nested((function, v, types, None, True, None)) for k, v in pbar}
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/utils/py_utils.py", line 251, in _single_map_nested
return function(data_struct)
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/formatting/torch_formatter.py", line 51, in _recursive_tensorize
return self._tensorize(data_struct)
File "/home/chainyo/miniconda3/envs/gan-bird/lib/python3.8/site-packages/datasets/formatting/torch_formatter.py", line 38, in _tensorize
if np.issubdtype(value.dtype, np.integer):
AttributeError: 'bytes' object has no attribute 'dtype'
```
I don't really understand why the image is still a bytes object while I used transformations on it. Here the code I used to upload the dataset (and it worked well):
```python
train_dataset = load_dataset("imagefolder", data_dir="data/train")
train_dataset = train_dataset["train"]
test_dataset = load_dataset("imagefolder", data_dir="data/test")
test_dataset = test_dataset["train"]
val_dataset = load_dataset("imagefolder", data_dir="data/val")
val_dataset = val_dataset["train"]
dataset = DatasetDict({
"train": train_dataset,
"val": val_dataset,
"test": test_dataset
})
dataset.push_to_hub("ChainYo/rvl-cdip")
```
Now here is the code I am using to get the dataset and prepare it for training:
```python
img_size = 512
batch_size = 128
normalize = [(0.5), (0.5)]
data_dir = "ChainYo/rvl-cdip"
dataset = load_dataset(data_dir, split="train")
transforms = transforms.Compose([
transforms.Resize(img_size),
transforms.CenterCrop(img_size),
transforms.ToTensor(),
transforms.Normalize(*normalize)
])
transformed_dataset = dataset.with_transform(transforms)
transformed_dataset.set_format(type="torch", device="cuda")
train_dataloader = torch.utils.data.DataLoader(
transformed_dataset, batch_size=batch_size, shuffle=True, num_workers=4, pin_memory=True
)
```
But this get me the error above. I don't understand why it's doing this kind of weird thing?
Do I need to map something on the dataset? Something like this:
```python
labels = dataset.features["label"].names
num_labels = dataset.features["label"].num_classes
def preprocess_data(examples):
images = [ex.convert("RGB") for ex in examples["image"]]
labels = [ex for ex in examples["label"]]
return {"images": images, "labels": labels}
features = Features({
"images": Image(decode=True, id=None),
"labels": ClassLabel(num_classes=num_labels, names=labels)
})
decoded_dataset = dataset.map(preprocess_data, remove_columns=dataset.column_names, features=features, batched=True, batch_size=100)
```
| 4,112 |
https://github.com/huggingface/datasets/issues/4107 | Unable to view the dataset and loading the same dataset throws the error - ArrowInvalid: Exceeded maximum rows | [
"Thanks for reporting. I'm looking at it",
" It's not related to the dataset viewer in itself. I can replicate the error with:\r\n\r\n```\r\n>>> import datasets as ds\r\n>>> d = ds.load_dataset('Pavithree/explainLikeImFive')\r\nUsing custom data configuration Pavithree--explainLikeImFive-b68b6d8112cd8a51\r\nDownloading and preparing dataset json/Pavithree--explainLikeImFive to /home/slesage/.cache/huggingface/datasets/json/Pavithree--explainLikeImFive-b68b6d8112cd8a51/0.0.0/ac0ca5f5289a6cf108e706efcf040422dbbfa8e658dee6a819f20d76bb84d26b...\r\nDownloading data: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 305M/305M [00:03<00:00, 98.6MB/s]\r\nDownloading data: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 17.9M/17.9M [00:00<00:00, 75.7MB/s]\r\nDownloading data: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 11.9M/11.9M [00:00<00:00, 70.6MB/s]\r\nDownloading data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:05<00:00, 1.92s/it]\r\nExtracting data files: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00, 1948.42it/s]\r\nFailed to read file '/home/slesage/.cache/huggingface/datasets/downloads/5fee9c8819754df277aee6f252e4db6897d785231c21938407b8862ca871d246' with error <class 'pyarrow.lib.ArrowInvalid'>: Exceeded maximum rows\r\nTraceback (most recent call last):\r\n File \"/home/slesage/hf/datasets/src/datasets/packaged_modules/json/json.py\", line 144, in _generate_tables\r\n dataset = json.load(f)\r\n File \"/home/slesage/.pyenv/versions/3.8.11/lib/python3.8/json/__init__.py\", line 293, in load\r\n return loads(fp.read(),\r\n File \"/home/slesage/.pyenv/versions/3.8.11/lib/python3.8/json/__init__.py\", line 357, in loads\r\n return _default_decoder.decode(s)\r\n File \"/home/slesage/.pyenv/versions/3.8.11/lib/python3.8/json/decoder.py\", line 340, in decode\r\n raise JSONDecodeError(\"Extra data\", s, end)\r\njson.decoder.JSONDecodeError: Extra data: line 1 column 916 (char 915)\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/home/slesage/hf/datasets/src/datasets/load.py\", line 1691, in load_dataset\r\n builder_instance.download_and_prepare(\r\n File \"/home/slesage/hf/datasets/src/datasets/builder.py\", line 605, in download_and_prepare\r\n self._download_and_prepare(\r\n File \"/home/slesage/hf/datasets/src/datasets/builder.py\", line 694, in _download_and_prepare\r\n self._prepare_split(split_generator, **prepare_split_kwargs)\r\n File \"/home/slesage/hf/datasets/src/datasets/builder.py\", line 1151, in _prepare_split\r\n for key, table in logging.tqdm(\r\n File \"/home/slesage/.pyenv/versions/datasets/lib/python3.8/site-packages/tqdm/std.py\", line 1168, in __iter__\r\n for obj in iterable:\r\n File \"/home/slesage/hf/datasets/src/datasets/packaged_modules/json/json.py\", line 146, in _generate_tables\r\n raise e\r\n File \"/home/slesage/hf/datasets/src/datasets/packaged_modules/json/json.py\", line 122, in _generate_tables\r\n pa_table = paj.read_json(\r\n File \"pyarrow/_json.pyx\", line 246, in pyarrow._json.read_json\r\n File \"pyarrow/error.pxi\", line 143, in pyarrow.lib.pyarrow_internal_check_status\r\n File \"pyarrow/error.pxi\", line 99, in pyarrow.lib.check_status\r\npyarrow.lib.ArrowInvalid: Exceeded maximum rows\r\n```\r\n\r\ncc @lhoestq @albertvillanova @mariosasko ",
"It seems that train.json is not a valid JSON Lines file: it has several JSON objects in the first line (the 915th character in the first line starts a new object, and there's no \"\\n\")\r\n\r\nYou need to have one JSON object per line",
"I'm closing this issue.\r\n\r\n@Pavithree, please, feel free to re-open it if fixing the JSON file does not solve it.",
"Thank you! that fixes the issue."
] | ## Dataset viewer issue - -ArrowInvalid: Exceeded maximum rows
**Link:** *https://huggingface.co/datasets/Pavithree/explainLikeImFive*
*This is the subset of original eli5 dataset https://huggingface.co/datasets/vblagoje/lfqa. I just filtered the data samples which belongs to one particular subreddit thread. However, the dataset preview for train split returns the below mentioned error:
Status code: 400
Exception: ArrowInvalid
Message: Exceeded maximum rows
When I try to load the same dataset it returns ArrowInvalid: Exceeded maximum rows error*
Am I the one who added this dataset ? Yes
| 4,107 |
https://github.com/huggingface/datasets/issues/4105 | push to hub fails with huggingface-hub 0.5.0 | [
"Hi ! Indeed there was a breaking change in `huggingface_hub` 0.5.0 in `HfApi.create_repo`, which is called here in `datasets` by passing the org name in both the `repo_id` and the `organization` arguments:\r\n\r\nhttps://github.com/huggingface/datasets/blob/2230f7f7d7fbaf102cff356f5a8f3bd1561bea43/src/datasets/arrow_dataset.py#L3363-L3369\r\n\r\nI think we should fix that in `huggingface_hub`, will keep you posted. In the meantime please use `huggingface_hub` 0.4.0",
"I'll be sending a fix for this later today on the `huggingface_hub` side.\r\n\r\nThe error would be converted to a `FutureWarning` if `datasets` uses kwargs instead of positional, for example here: \r\n\r\nhttps://github.com/huggingface/datasets/blob/2230f7f7d7fbaf102cff356f5a8f3bd1561bea43/src/datasets/arrow_dataset.py#L3363-L3369\r\n\r\nto be:\r\n\r\n``` python\r\n api.create_repo(\r\n name=dataset_name,\r\n token=token,\r\n repo_type=\"dataset\",\r\n organization=organization,\r\n private=private,\r\n )\r\n```\r\n\r\nBut `name` and `organization` are deprecated in `huggingface_hub=0.5`, and people should pass `repo_id='org/name` instead. Note that `repo_id` was introduced in 0.5 and if `datasets` wants to support older `huggingface_hub` versions (which I encourage it to do), there needs to be a helper function to do that. It can be something like:\r\n\r\n\r\n```python\r\ndef create_repo(\r\n client,\r\n name: str,\r\n token: Optional[str] = None,\r\n organization: Optional[str] = None,\r\n private: Optional[bool] = None,\r\n repo_type: Optional[str] = None,\r\n exist_ok: Optional[bool] = False,\r\n space_sdk: Optional[str] = None,\r\n) -> str:\r\n try:\r\n return client.create_repo(\r\n repo_id=f\"{organization}/{name}\",\r\n token=token,\r\n private=private,\r\n repo_type=repo_type,\r\n exist_ok=exist_ok,\r\n space_sdk=space_sdk,\r\n )\r\n except TypeError:\r\n return client.create_repo(\r\n name=name,\r\n organization=organization,\r\n token=token,\r\n private=private,\r\n repo_type=repo_type,\r\n exist_ok=exist_ok,\r\n space_sdk=space_sdk,\r\n )\r\n```\r\n\r\nin a `utils/_fixes.py` kinda file and and be used internally.\r\n\r\nI'll be sending a patch to `huggingface_hub` to convert the error reported in this issue to a `FutureWarning`.",
"PR with the hotfix on the `huggingface_hub` side: https://github.com/huggingface/huggingface_hub/pull/822",
"We can definitely change `push_to_hub` to use `repo_id` in `datasets` and require `huggingface_hub>=0.5.0`.\r\n\r\nLet me open a PR :)",
"`huggingface_hub` 0.5.1 just got released with a fix, feel free to update `huggingface_hub` ;)"
] | ## Describe the bug
`ds.push_to_hub` is failing when updating a dataset in the form "org_id/repo_id"
## Steps to reproduce the bug
```python
from datasets import load_dataset
ds = load_dataset("rubrix/news_test")
ds.push_to_hub("<your-user>/news_test", token="<your-token>")
```
## Expected results
The dataset is successfully uploaded
## Actual results
An error validation is raised:
```bash
if repo_id and (name or organization):
> raise ValueError(
"Only pass `repo_id` and leave deprecated `name` and "
"`organization` to be None."
E ValueError: Only pass `repo_id` and leave deprecated `name` and `organization` to be None.
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.18.1
- `huggingface-hub`: 0.5
- Platform: macOS
- Python version: 3.8.12
- PyArrow version: 6.0.0
cc @adrinjalali
| 4,105 |
https://github.com/huggingface/datasets/issues/4104 | Add time series data - stock market | [
"Can I use instructions present in below link for time series dataset as well? \r\nhttps://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md ",
"cc'ing @kashif and @NielsRogge for visibility!",
"@INF800 happy to add this dataset! I will try to set a PR by the end of the day... if you can kindly point me to the dataset? Also, note we have a bunch of time series datasets checked in e.g. `electricity_load_diagrams` or `monash_tsf`, and ideally this dataset could also be in a similar format. ",
"Thankyou. This is how raw data looks like before cleaning for an individual stocks:\r\n\r\n1. https://github.com/INF800/marktech/tree/raw-data/f/data/raw\r\n2. https://github.com/INF800/marktech/tree/raw-data/t/data/raw\r\n3. https://github.com/INF800/marktech/tree/raw-data/rdfn/data/raw\r\n4. https://github.com/INF800/marktech/tree/raw-data/irbt/data/raw\r\n5. https://github.com/INF800/marktech/tree/raw-data/hll/data/raw\r\n6. https://github.com/INF800/marktech/tree/raw-data/infy/data/raw\r\n7. https://github.com/INF800/marktech/tree/raw-data/reli/data/raw\r\n8. https://github.com/INF800/marktech/tree/raw-data/hdbk/data/raw\r\n\r\n> Scraping is automated using GitHub Actions. So, everyday we will see a new file added in the above links.\r\n\r\nI can rewrite the cleaning scripts to make sure it fits HF dataset standards. (P.S I am very much new to HF dataset)\r\n\r\nThe data set above can be converted into univariate regression / multivariate regression / sequence to sequence generation dataset etc. So, do we have some kind of transformation modules that will read the dataset as some type of dataset (`GenericTimeData`) and convert it to other possible dataset relating to a specific ML task. **By having this kind of transformation module, I only have to add data once** and use transformation module whenever necessary\r\n\r\nAdditionally, having some kind of versioning for the dataset will be really helpful because it will keep on updating - especially time series datasets ",
"thanks @INF800 I'll have a look. I believe it should be possible to incorporate this into the time-series format.",
"Referencing https://github.com/qingsongedu/time-series-transformers-review",
"@INF800 yes I am aware of the review repository and paper which is more or less a collection of abstracts etc. I am working on a unified library of implementations of these papers together with datasets to be then able to compare/contrast and build upon the research etc. but I am not ready to share them publicly just yet.\r\n\r\nIn any case regarding your dataset at the moment its seems from looking at the csv files, its mixture of textual and numerical data, sometimes in the same column etc. As you know, for time series models we would need just numeric data so I would need your help in disambiguating the dataset you have collected and also perhaps starting with just numerical data to start with... \r\n\r\nDo you think you can make a version with just numerical data?",
"> @INF800 yes I am aware of the review repository and paper which is more or less a collection of abstracts etc. I am working on a unified library of implementations of these papers together with datasets to be then able to compare/contrast and build upon the research etc. but I am not ready to share them publicly just yet.\r\n> \r\n> In any case regarding your dataset at the moment its seems from looking at the csv files, its mixture of textual and numerical data, sometimes in the same column etc. As you know, for time series models we would need just numeric data so I would need your help in disambiguating the dataset you have collected and also perhaps starting with just numerical data to start with...\r\n> \r\n> Do you think you can make a version with just numerical data?\r\n\r\nWill share the numeric data and conversion script within end of this week. \r\n\r\nI am on a business trip currently - it is in my desktop.",
"thanks @INF800 [email protected] should work",
"It should be in your inbox!\r\n\r\nOn Sun, 21 Jul, 2024, 9:44 pm Kashif Rasul, ***@***.***>\r\nwrote:\r\n\r\n> thanks @INF800 <https://github.com/INF800> ***@***.*** should\r\n> work\r\n>\r\n> —\r\n> Reply to this email directly, view it on GitHub\r\n> <https://github.com/huggingface/datasets/issues/4104#issuecomment-2241701256>,\r\n> or unsubscribe\r\n> <https://github.com/notifications/unsubscribe-auth/AK4GSXLHCOGNTU5ERJ6M3ITZNPM6TAVCNFSM6AAAAABLG65FLKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENBRG4YDCMRVGY>\r\n> .\r\n> You are receiving this because you were mentioned.Message ID:\r\n> ***@***.***>\r\n>\r\n"
] | ## Adding a Time Series Dataset
- **Name:** 2min ticker data for stock market
- **Description:** 8 stocks' data collected for 1month post ukraine-russia war. 4 NSE stocks and 4 NASDAQ stocks. Along with technical indicators (additional features) as shown in below image
- **Data:** Collected by myself from investing.com
- **Motivation:** Test applicability of transformer based model on stock market / time series problem
 | 4,104 |
https://github.com/huggingface/datasets/issues/4101 | How can I download only the train and test split for full numbers using load_dataset()? | [
"Hi! Can you please specify the full name of the dataset? IIRC `full_numbers` is one of the configs of the `svhn` dataset, and its generation is slow due to data being stored in binary Matlab files. Even if you specify a specific split, `datasets` downloads all of them, but we plan to fix that soon and only download the requested split.\r\n\r\nIf you are in a hurry, download the `svhn` script [here](`https://huggingface.co/datasets/svhn/blob/main/svhn.py`), remove [this code](https://huggingface.co/datasets/svhn/blob/main/svhn.py#L155-L162), and run:\r\n```python\r\nfrom datasets import load_dataset\r\ndset = load_dataset(\"path/to/your/local/script.py\", \"full_numbers\")\r\n```\r\n\r\nAnd to make loading easier in Colab, you can create a dataset repo on the Hub and upload the script there. Or push the script to Google Drive and mount the drive in Colab."
] | How can I download only the train and test split for full numbers using load_dataset()?
I do not need the extra split and it will take 40 mins just to download in Colab. I have very short time in hand. Please help. | 4,101 |
https://github.com/huggingface/datasets/issues/4099 | UnicodeDecodeError: 'ascii' codec can't decode byte 0xe5 in position 213: ordinal not in range(128) | [
"Hi @andreybond, thanks for reporting.\r\n\r\nUnfortunately, I'm not able to able to reproduce your issue:\r\n```python\r\nIn [4]: from datasets import load_dataset\r\n ...: datasets = load_dataset(\"nielsr/XFUN\", \"xfun.ja\")\r\n\r\nIn [5]: datasets\r\nOut[5]: \r\nDatasetDict({\r\n train: Dataset({\r\n features: ['id', 'input_ids', 'bbox', 'labels', 'image', 'entities', 'relations'],\r\n num_rows: 194\r\n })\r\n validation: Dataset({\r\n features: ['id', 'input_ids', 'bbox', 'labels', 'image', 'entities', 'relations'],\r\n num_rows: 71\r\n })\r\n})\r\n```\r\n\r\nThe only reason I can imagine this issue may arise is if your default encoding is not \"UTF-8\" (and it is ASCII instead). This is usually the case on Windows machines; but you say your environment is a Linux machine. Maybe you change your machine default encoding?\r\n\r\nCould you please check this?\r\n```python\r\nIn [6]: import sys\r\n\r\nIn [7]: sys.getdefaultencoding()\r\nOut[7]: 'utf-8'\r\n```",
"I opened a PR in the original dataset loading script:\r\n- microsoft/unilm#677\r\n\r\nand fixed the corresponding dataset script on the Hub:\r\n- https://huggingface.co/datasets/nielsr/XFUN/commit/73ba5e026621e05fb756ae0f267eb49971f70ebd",
"import sys\r\nsys.getdefaultencoding()\r\n\r\nreturned: 'utf-8'\r\n\r\n---------------------\r\n\r\nI've just cloned master branch - your fix works! Thank you!"
] | ## Describe the bug
Error "UnicodeDecodeError: 'ascii' codec can't decode byte 0xe5 in position 213: ordinal not in range(128)" is thrown when downloading dataset.
## Steps to reproduce the bug
```python
from datasets import load_dataset
datasets = load_dataset("nielsr/XFUN", "xfun.ja")
```
## Expected results
Dataset should be downloaded without exceptions
## Actual results
Stack trace (for the second-time execution):
Downloading and preparing dataset xfun/xfun.ja to /root/.cache/huggingface/datasets/nielsr___xfun/xfun.ja/0.0.0/e06e948b673d1be9a390a83c05c10e49438bf03dd85ae9a4fe06f8747a724477...
Downloading data files: 100%
2/2 [00:00<00:00, 88.48it/s]
Extracting data files: 100%
2/2 [00:00<00:00, 79.60it/s]
UnicodeDecodeErrorTraceback (most recent call last)
<ipython-input-31-79c26bd1109c> in <module>
1 from datasets import load_dataset
2
----> 3 datasets = load_dataset("nielsr/XFUN", "xfun.ja")
/usr/local/lib/python3.6/dist-packages/datasets/load.py in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, ignore_verifications, keep_in_memory, save_infos, revision, use_auth_token, task, streaming, **config_kwargs)
/usr/local/lib/python3.6/dist-packages/datasets/builder.py in download_and_prepare(self, download_config, download_mode, ignore_verifications, try_from_hf_gcs, dl_manager, base_path, use_auth_token, **download_and_prepare_kwargs)
604 )
605
--> 606 # By default, return all splits
607 if split is None:
608 split = {s: s for s in self.info.splits}
/usr/local/lib/python3.6/dist-packages/datasets/builder.py in _download_and_prepare(self, dl_manager, verify_infos)
/usr/local/lib/python3.6/dist-packages/datasets/builder.py in _download_and_prepare(self, dl_manager, verify_infos, **prepare_split_kwargs)
692 Args:
693 split: `datasets.Split` which subset of the data to read.
--> 694
695 Returns:
696 `Dataset`
/usr/local/lib/python3.6/dist-packages/datasets/builder.py in _prepare_split(self, split_generator, check_duplicate_keys)
/usr/local/lib/python3.6/dist-packages/tqdm/notebook.py in __iter__(self)
252 if not self.disable:
253 self.display(check_delay=False)
--> 254
255 def __iter__(self):
256 try:
/usr/local/lib/python3.6/dist-packages/tqdm/std.py in __iter__(self)
1183 for obj in iterable:
1184 yield obj
-> 1185 return
1186
1187 mininterval = self.mininterval
~/.cache/huggingface/modules/datasets_modules/datasets/nielsr--XFUN/e06e948b673d1be9a390a83c05c10e49438bf03dd85ae9a4fe06f8747a724477/XFUN.py in _generate_examples(self, filepaths)
140 logger.info("Generating examples from = %s", filepath)
141 with open(filepath[0], "r") as f:
--> 142 data = json.load(f)
143
144 for doc in data["documents"]:
/usr/lib/python3.6/json/__init__.py in load(fp, cls, object_hook, parse_float, parse_int, parse_constant, object_pairs_hook, **kw)
294
295 """
--> 296 return loads(fp.read(),
297 cls=cls, object_hook=object_hook,
298 parse_float=parse_float, parse_int=parse_int,
/usr/lib/python3.6/encodings/ascii.py in decode(self, input, final)
24 class IncrementalDecoder(codecs.IncrementalDecoder):
25 def decode(self, input, final=False):
---> 26 return codecs.ascii_decode(input, self.errors)[0]
27
28 class StreamWriter(Codec,codecs.StreamWriter):
UnicodeDecodeError: 'ascii' codec can't decode byte 0xe5 in position 213: ordinal not in range(128)
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.0.0 (but reproduced with many previous versions)
- Platform: Docker: Linux da5b74136d6b 5.3.0-1031-azure #32~18.04.1-Ubuntu SMP Mon Jun 22 15:27:23 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux ; Base docker image is : huggingface/transformers-pytorch-cpu
- Python version: 3.6.9
- PyArrow version: 6.0.1
| 4,099 |
https://github.com/huggingface/datasets/issues/4096 | Add support for streaming Zarr stores for hosted datasets | [
"Hi @jacobbieker, thanks for your request and study of possible alternatives.\r\n\r\nWe are very interested in finding a way to make `datasets` useful to you.\r\n\r\nLooking at the Zarr docs, I saw that among its storage alternatives, there is the ZIP file format: https://zarr.readthedocs.io/en/stable/api/storage.html#zarr.storage.ZipStore\r\n\r\nThis might be convenient for many reasons:\r\n- On the one hand, we avoid the Git issue with huge number of small files: chunks files are compressed into a single ZIP file\r\n- On the other hand, the ZIP file format is specially suited for streaming data because it allows random access to its component files (i.e. it supports random access to its chunks)\r\n\r\nAnyway, I think that a Python loading script will be necessary: you need to implement additional logic to select certain chunks (based on date or other criteria).\r\n\r\nPlease, let me know if this makes sense to you.",
"Ah okay, I missed the option of zip files for zarr, I'll try that with our repos and see if it works! Thanks a lot!",
"Hi @jacobbieker, does the Zarr ZipStore work for your use case?",
"Hi,\r\n\r\nYes, it seems to! I got it working for https://huggingface.co/datasets/openclimatefix/mrms thanks for the help! ",
"On behalf of the Zarr developers, let me say THANK YOU for working to support Zarr on HF! 🙏 Zarr is a 100% open-source and community driven project (fiscally sponsored by NumFocus). We see it as an ideal format for ML training datasets, particularly in scientific domains.\r\n\r\nI think the solution of zipping the Zarr store is a reasonable way to balance the constraints of Git LFS with the structure of Zarr.\r\n\r\nIt would be amazing to get something on the [Hugging Face Datasets Docs](https://huggingface.co/docs/datasets/index) about how to best work with Zarr. Let me know if there's a way I could help with that effort.",
"Also just noting here that I was able to lazily open @jacobbieker's dataset over the internet from HF hub 🚀 !\r\n\r\n```python\r\nimport xarray as xr\r\nurl = \"https://huggingface.co/datasets/openclimatefix/mrms/resolve/main/data/2016_001.zarr.zip\"\r\nzip_url = 'zip:///::' + url\r\nds = xr.open_dataset(zip_url, engine='zarr', chunks={})\r\n```\r\n\r\n<img width=\"740\" alt=\"image\" src=\"https://user-images.githubusercontent.com/1197350/164508663-bc75cdc0-734d-44f4-9562-2877ecfdf433.png\">\r\n",
"However, I wasn't able to get streaming working using the Datasets api:\r\n\r\n```python\r\nfrom datasets import load_dataset\r\nds = load_dataset(\"openclimatefix/mrms\", streaming=True, split='train')\r\nitem = next(iter(ds))\r\n```\r\n\r\n<details>\r\n<summary>FileNotFoundError traceback</summary>\r\n\r\n```\r\nNo config specified, defaulting to: mrms/2021\r\nzip://::https://huggingface.co/datasets/openclimatefix/mrms/resolve/main/data/2016_001.zarr.zip\r\ndata/2016_001.zarr.zip\r\nzip://2016_001.zarr.zip::https://huggingface.co/datasets/openclimatefix/mrms/resolve/main/data/2016_001.zarr.zip\r\n---------------------------------------------------------------------------\r\nFileNotFoundError Traceback (most recent call last)\r\nInput In [1], in <cell line: 3>()\r\n 1 from datasets import load_dataset\r\n 2 ds = load_dataset(\"openclimatefix/mrms\", streaming=True, split='train')\r\n----> 3 item = next(iter(ds))\r\n\r\nFile /opt/miniconda3/envs/hugginface/lib/python3.9/site-packages/datasets/iterable_dataset.py:497, in IterableDataset.__iter__(self)\r\n 496 def __iter__(self):\r\n--> 497 for key, example in self._iter():\r\n 498 if self.features:\r\n 499 # we encode the example for ClassLabel feature types for example\r\n 500 encoded_example = self.features.encode_example(example)\r\n\r\nFile /opt/miniconda3/envs/hugginface/lib/python3.9/site-packages/datasets/iterable_dataset.py:494, in IterableDataset._iter(self)\r\n 492 else:\r\n 493 ex_iterable = self._ex_iterable\r\n--> 494 yield from ex_iterable\r\n\r\nFile /opt/miniconda3/envs/hugginface/lib/python3.9/site-packages/datasets/iterable_dataset.py:87, in ExamplesIterable.__iter__(self)\r\n 86 def __iter__(self):\r\n---> 87 yield from self.generate_examples_fn(**self.kwargs)\r\n\r\nFile ~/.cache/huggingface/modules/datasets_modules/datasets/openclimatefix--mrms/2a6f697014d7eb3caf586ca137d47ca38785ae2fe36248611b021f8248b59936/mrms.py:150, in MRMS._generate_examples(self, filepath, split)\r\n 147 filepath = \"[https://huggingface.co/datasets/openclimatefix/mrms/resolve/main/data/2016_001.zarr.zip](https://huggingface.co/datasets/openclimatefix/mrms/resolve/main/data/2016_001.zarr.zip%3C/span%3E%3Cspan) style=\"color:rgb(175,0,0)\">\"\r\n 148 # TODO: This method handles input defined in _split_generators to yield (key, example) tuples from the dataset.\r\n 149 # The `key` is for legacy reasons (tfds) and is not important in itself, but must be unique for each example.\r\n--> 150 with zarr.storage.FSStore(fsspec.open(\"zip::\" + filepath, mode='r'), mode='r') as store:\r\n 151 data = xr.open_zarr(store)\r\n 152 for key, row in enumerate(data[\"time\"].values):\r\n\r\nFile /opt/miniconda3/envs/hugginface/lib/python3.9/site-packages/zarr/storage.py:1120, in FSStore.__init__(self, url, normalize_keys, key_separator, mode, exceptions, dimension_separator, **storage_options)\r\n 1117 import fsspec\r\n 1118 self.normalize_keys = normalize_keys\r\n-> 1120 protocol, _ = fsspec.core.split_protocol(url)\r\n 1121 # set auto_mkdir to True for local file system\r\n 1122 if protocol in (None, \"file\") and not storage_options.get(\"auto_mkdir\"):\r\n\r\nFile /opt/miniconda3/envs/hugginface/lib/python3.9/site-packages/fsspec/core.py:514, in split_protocol(urlpath)\r\n 512 def split_protocol(urlpath):\r\n 513 \"\"\"Return protocol, path pair\"\"\"\r\n--> 514 urlpath = stringify_path(urlpath)\r\n 515 if \"://\" in urlpath:\r\n 516 protocol, path = urlpath.split(\"://\", 1)\r\n\r\nFile /opt/miniconda3/envs/hugginface/lib/python3.9/site-packages/fsspec/utils.py:315, in stringify_path(filepath)\r\n 313 return filepath\r\n 314 elif hasattr(filepath, \"__fspath__\"):\r\n--> 315 return filepath.__fspath__()\r\n 316 elif isinstance(filepath, pathlib.Path):\r\n 317 return str(filepath)\r\n\r\nFile /opt/miniconda3/envs/hugginface/lib/python3.9/site-packages/fsspec/core.py:98, in OpenFile.__fspath__(self)\r\n 96 def __fspath__(self):\r\n 97 # may raise if cannot be resolved to local file\r\n---> 98 return self.open().__fspath__()\r\n\r\nFile /opt/miniconda3/envs/hugginface/lib/python3.9/site-packages/fsspec/core.py:140, in OpenFile.open(self)\r\n 132 def open(self):\r\n 133 \"\"\"Materialise this as a real open file without context\r\n 134 \r\n 135 The file should be explicitly closed to avoid enclosed file\r\n (...)\r\n 138 been deleted; but a with-context is better style.\r\n 139 \"\"\"\r\n--> 140 out = self.__enter__()\r\n 141 closer = out.close\r\n 142 fobjects = self.fobjects.copy()[:-1]\r\n\r\nFile /opt/miniconda3/envs/hugginface/lib/python3.9/site-packages/fsspec/core.py:103, in OpenFile.__enter__(self)\r\n 100 def __enter__(self):\r\n 101 mode = self.mode.replace(\"t\", \"\").replace(\"b\", \"\") + \"b\"\r\n--> 103 f = self.fs.open(self.path, mode=mode)\r\n 105 self.fobjects = [f]\r\n 107 if self.compression is not None:\r\n\r\nFile /opt/miniconda3/envs/hugginface/lib/python3.9/site-packages/fsspec/spec.py:1009, in AbstractFileSystem.open(self, path, mode, block_size, cache_options, compression, **kwargs)\r\n 1007 else:\r\n 1008 ac = kwargs.pop(\"autocommit\", not self._intrans)\r\n-> 1009 f = self._open(\r\n 1010 path,\r\n 1011 mode=mode,\r\n 1012 block_size=block_size,\r\n 1013 autocommit=ac,\r\n 1014 cache_options=cache_options,\r\n 1015 **kwargs,\r\n 1016 )\r\n 1017 if compression is not None:\r\n 1018 from fsspec.compression import compr\r\n\r\nFile /opt/miniconda3/envs/hugginface/lib/python3.9/site-packages/fsspec/implementations/zip.py:96, in ZipFileSystem._open(self, path, mode, block_size, autocommit, cache_options, **kwargs)\r\n 94 if mode != \"rb\":\r\n 95 raise NotImplementedError\r\n---> 96 info = self.info(path)\r\n 97 out = self.zip.open(path, \"r\")\r\n 98 out.size = info[\"size\"]\r\n\r\nFile /opt/miniconda3/envs/hugginface/lib/python3.9/site-packages/fsspec/archive.py:42, in AbstractArchiveFileSystem.info(self, path, **kwargs)\r\n 40 return self.dir_cache[path + \"/\"]\r\n 41 else:\r\n---> 42 raise FileNotFoundError(path)\r\n\r\nFileNotFoundError:\r\n```\r\n\r\n</details>\r\n\r\nIs this a bug? Or am I just doing it wrong...",
"I'm still messing around with that dataset, so the data might have moved. I currently have each year of MRMS precipitation rate data as it's own zarr, but as they are quite large (on order of 100GB each) I'm working to split them into single days, and as such they are still being moved around, I was just trying to get a proof of concept working originally. ",
"I've mostly finished rearranging the data now and uploading some more, so this works now:\r\n```python\r\nimport datasets\r\nds = datasets.load_dataset(\"openclimatefix/mrms\", streaming=True, split=\"train\")\r\nitem = next(iter(ds))\r\nprint(item.keys())\r\nprint(item[\"timestamp\"])\r\n```\r\n\r\nThe MRMS data now goes most of 2016-2022, with quite a few gaps I'm working on filling in",
"Hi @albertvillanova, I noticed there is now the [HFFileSystem](https://huggingface.co/docs/huggingface_hub/main/en/guides/hf_file_system), where the docs show an example of writing a Zarr store directly to the Hub, and no mention of having too many files. Is there still the restriction on lots of files in `datasets`? It would be more convenient to be able to have the geospatial data in one large Zarr store, rather than in multiple smaller ones, but happy to continue using zipped Zarrs if thats the recommended way.",
"Hi @jacobbieker.\r\n\r\nThanks for coming back to this pending issue. \r\n\r\nIn fact, we are now using the `fsspec` API in our `HFFileSystem`, which was not the case when you created this issue.\r\nOn the other hand, I am not sure of the current limitations, both in terms of the number of files or performance when loading.\r\n- If I remember correctly, I think there is a limit in the maximum number of files per directory: 10k\r\n\r\nI think it would be best to try a POC again and discuss any issues that arise and whether we can fix them on our end (both `datasets` and the Hub).\r\nWe would really like to support the Zarr format 100% and that the Hub is really convenient for your use case. So do not hesitate to report any problem: you can ping me on the Hub as @albertvillanova"
] | **Is your feature request related to a problem? Please describe.**
Lots of geospatial data is stored in the Zarr format. This format works well for n-dimensional data and coordinates, and can have good compression. Unfortunately, HF datasets doesn't support streaming in data in Zarr format as far as I can tell. Zarr stores are designed to be easily streamed in from cloud storage, especially with xarray and fsspec. Since geospatial data tends to be very large, and on the order of TBs of data or 10's of TBs of data for a single dataset, it can be difficult to store the dataset locally for users. Just adding Zarr stores with HF git doesn't work well (see https://github.com/huggingface/datasets/issues/3823) as Zarr splits the data into lots of small chunks for fast loading, and that doesn't work well with git. I've somewhat gotten around that issue by tarring each Zarr store and uploading them as a single file, which seems to be working (see https://huggingface.co/datasets/openclimatefix/gfs-reforecast for example data files, although the script isn't written yet). This does mean that streaming doesn't quite work though. On the other hand, in https://huggingface.co/datasets/openclimatefix/eumetsat_uk_hrv we stream in a Zarr store from a public GCP bucket quite easily.
**Describe the solution you'd like**
A way to upload Zarr stores for hosted datasets so that we can stream it with xarray and fsspec.
**Describe alternatives you've considered**
Tarring each Zarr store individually and just extracting them in the dataset script -> Downside this is a lot of data that probably doesn't fit locally for a lot of potential users.
Pre-prepare examples in a format like Parquet -> Would use a lot more storage, and a lot less flexibility, in the eumetsat_uk_hrv, we use the one Zarr store for multiple different configurations.
| 4,096 |
https://github.com/huggingface/datasets/issues/4094 | Helo Mayfrends | [] | ## Adding a Dataset
- **Name:** *name of the dataset*
- **Description:** *short description of the dataset (or link to social media or blog post)*
- **Paper:** *link to the dataset paper if available*
- **Data:** *link to the Github repository or current dataset location*
- **Motivation:** *what are some good reasons to have this dataset*
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| 4,094 |
https://github.com/huggingface/datasets/issues/4093 | elena-soare/crawled-ecommerce: missing dataset | [
"It's a bug! Thanks for reporting, I'm looking at it.",
"By the way, the error on our part is due to the huge size of every row (~90MB). The dataset viewer does not support such big dataset rows for the moment.\r\nAnyway, we're working to give a hint about this in the dataset viewer.",
"Fixed. See https://huggingface.co/datasets/elena-soare/crawled-ecommerce/viewer/elena-soare--crawled-ecommerce/train.\r\n\r\n<img width=\"1552\" alt=\"Capture d’écran 2022-04-12 à 11 23 51\" src=\"https://user-images.githubusercontent.com/1676121/162929722-2e2b80e2-154a-4b61-87bd-e341bd6c46e6.png\">\r\n\r\nThanks for reporting!"
] | elena-soare/crawled-ecommerce
**Link:** *link to the dataset viewer page*
*short description of the issue*
Am I the one who added this dataset ? Yes-No
| 4,093 |
https://github.com/huggingface/datasets/issues/4091 | Build a Dataset One Example at a Time Without Loading All Data Into Memory | [
"Hi! Yes, the problem with `add_item` is that it keeps examples in memory, so you are left with these options:\r\n* writing a dataset loading script in which you iterate over `custom_example_dict_streamer` and yield the examples (in `_generate examples`)\r\n* storing the data in a JSON/CSV/Parquet/TXT file and using `Dataset.from_{format}`\r\n* using `add_item` + `save_to_disk` on smaller chunks: \r\n ```python\r\n from datasets import Dataset, concatenate_datasets\r\n MAX_SAMPLES_IN_MEMORY = 1000\r\n samples_in_dset = 0\r\n dset = Dataset.from_dict({\"col1\": [], \"col2\": []}) # empty dataset\r\n path_to_save_dir = \"path/to/save/dir\"\r\n num_chunks = 0\r\n for example_dict in custom_example_dict_streamer(\"/path/to/raw/data\"):\r\n dset = dset.add_item(example_dict)\r\n samples_in_dset += 1\r\n if samples_in_dset == MAX_SAMPLES_IN_MEMORY:\r\n samples_in_dset = 0\r\n dset.save_to_disk(f\"{path_to_save_dir}{num_chunks}\")\r\n num_chunks =+ 1\r\n dset = Dataset.from_dict({\"col1\": [], \"col2\": []}) # empty dataset\r\n if samples_in_dset > 0:\r\n dset.save_to_disk(f\"{path_to_save_dir}{num_chunks}\")\r\n num_chunks =+ 1\r\n loaded_dsets = [] # memory-mapped\r\n for chunk_num in range(num_chunks):\r\n dset = Dataset.load_from_disk(f\"{path_to_save_dir}{chunk_num}\") \r\n loaded_dsets.append(dset)\r\n final_dset = concatenate_datasets(dset)\r\n ```\r\n If you still have issues with this approach, you can try to delete unused datasets with `gc.collect()` to free some memory. ",
"This is really elegant, thank you @mariosasko! I will try this."
] | **Is your feature request related to a problem? Please describe.**
I have a very large dataset stored on disk in a custom format. I have some custom code that reads one data example at a time and yields it in the form of a dictionary. I want to construct a `Dataset` with all examples, and then save it to disk. I later want to load the saved `Dataset` and use it like any other HuggingFace dataset, get splits, wrap it in a PyTorch `DataLoader`, etc. **Crucially, I do not ever want to materialize all the data in memory while building the dataset.**
**Describe the solution you'd like**
I would like to be able to do something like the following. Notice how each example is read and then immediately added to the dataset. We do not store all the data in memory when constructing the `Dataset`. If it helps, I will know the schema of my dataset before hand.
```
# Initialize an empty Dataset, possibly from a known schema.
dataset = Dataset()
# Read in examples one by one using a custom data streamer.
for example_dict in custom_example_dict_streamer("/path/to/raw/data"):
# Add this example to the dict but do not store it in memory.
dataset.add_item(example_dict)
# Save the final dataset to disk as an Arrow-backed dataset.
dataset.save_to_disk("/path/to/dataset")
...
# I'd like to be able to later `load_from_disk` and use the loaded Dataset
# just like any other memory-mapped pyarrow-backed HuggingFace dataset...
loaded_dataset = Dataset.load_from_disk("/path/to/dataset")
loaded_dataset.set_format(type="torch", columnns=["foo", "bar", "baz"])
dataloader = torch.utils.data.DataLoader(loaded_dataset, batch_size=16)
...
```
**Describe alternatives you've considered**
I initially tried to read all the data into memory, construct a Pandas DataFrame and then call `Dataset.from_pandas`. This would not work as it requires storing all the data in memory. It seems that there is an `add_item` method already -- I tried to implement something like the desired API written above, but I've not been able to initialize an empty `Dataset` (this seems to require several layers of constructing `datasets.table.Table` which requires constructing a `pyarrow.lib.Table`, etc). I also considered writing my data to multiple sharded CSV files or JSON files and then using `from_csv` or `from_json`. I'd prefer not to do this because (1) I'd prefer to avoid the intermediate step of creating these temp CSV/JSON files and (2) I'm not sure if `from_csv` and `from_json` use memory-mapping.
Do you have any suggestions on how I'd be able to achieve this use case? Does something already exist to support this? Thank you very much in advance! | 4,091 |
https://github.com/huggingface/datasets/issues/4086 | Dataset viewer issue for McGill-NLP/feedbackQA | [
"Hi @cslizc, thanks for reporting.\r\n\r\nI have just forced the refresh of the corresponding cache and the preview is working now.",
"thank you so much"
] | ## Dataset viewer issue for '*McGill-NLP/feedbackQA*'
**Link:** *[link to the dataset viewer page](https://huggingface.co/datasets/McGill-NLP/feedbackQA)*
*short description of the issue*
The dataset can be loaded correctly with `load_dataset` but the preview doesn't work. Error message:
```
Status code: 400
Exception: Status400Error
Message: Not found. Maybe the cache is missing, or maybe the dataset does not exist.
```
Am I the one who added this dataset ? Yes
| 4,086 |
https://github.com/huggingface/datasets/issues/4085 | datasets.set_progress_bar_enabled(False) not working in datasets v2 | [
"Now, I can't find any reference to set_progress_bar_enabled in the code.\r\n\r\nI think it have been deleted",
"Hi @virilo,\r\n\r\nPlease note that since `datasets` version 2.0.0, we have aligned with `transformers` the management of the progress bar (among other things):\r\n- #3897\r\n\r\nNow, you should update your code to use `datasets.logging.disable_progress_bar`.\r\n\r\nYou have more info in our docs: [Logging methods](https://huggingface.co/docs/datasets/package_reference/logging_methods)",
"One important thing for beginner like me is: from datasets.utils.logging import disable_progress_bar\r\nDo not forget the 'utils' or you will waste a long time like me...."
] | ## Describe the bug
datasets.set_progress_bar_enabled(False) not working in datasets v2
## Steps to reproduce the bug
```python
datasets.set_progress_bar_enabled(False)
```
## Expected results
datasets not using any progress bar
## Actual results
AttributeError: module 'datasets' has no attribute 'set_progress_bar_enabled
## Environment info
datasets version 2
| 4,085 |
https://github.com/huggingface/datasets/issues/4084 | Errors in `Train with Datasets` Tensorflow code section on Huggingface.co | [
"Hi @blackhat-coder, thanks for reporting.\r\n\r\nPlease note that the `transformers` library updated their data collators API last year (version 4.10.0):\r\n- huggingface/transformers#13105\r\n\r\nnow requiring to pass `return_tensors` argument at Data Collator instantiation.\r\n\r\nAnd therefore, we also updated in the `datasets` library documentation all the examples using `transformers` data collators.\r\n\r\nIf you would like to follow our examples, please update your installed `transformers` version:\r\n```\r\npip install -U transformers\r\n```"
] | ## Describe the bug
Hi
### Error 1
Running the Tensforlow code on [Huggingface](https://huggingface.co/docs/datasets/use_dataset) gives a TypeError: __init__() got an unexpected keyword argument 'return_tensors'
### Error 2
`DataCollatorWithPadding` isn't imported
## Steps to reproduce the bug
```python
import tensorflow as tf
from datasets import load_dataset
from transformers import AutoTokenizer
dataset = load_dataset('glue', 'mrpc', split='train')
tokenizer = AutoTokenizer.from_pretrained('bert-base-cased')
dataset = dataset.map(lambda e: tokenizer(e['sentence1'], truncation=True, padding='max_length'), batched=True)
data_collator = DataCollatorWithPadding(tokenizer=tokenizer, return_tensors="tf")
train_dataset = dataset["train"].to_tf_dataset(
columns=['input_ids', 'token_type_ids', 'attention_mask', 'label'],
shuffle=True,
batch_size=16,
collate_fn=data_collator,
)
```
This is the same code on Huggingface.co
## Actual results
TypeError: __init__() got an unexpected keyword argument 'return_tensors'
## Environment info
- `datasets` version: 2.0.0
- Platform: Windows-10-10.0.19044-SP0
- Python version: 3.9.7
- PyArrow version: 6.0.0
- Pandas version: 1.4.1
> | 4,084 |
https://github.com/huggingface/datasets/issues/4080 | NonMatchingChecksumError for downloading conll2012_ontonotesv5 dataset | [
"Hi @richarddwang,\r\n\r\n\r\nIndeed, we have recently updated the loading script of that dataset (and fixed that bug as well):\r\n- #4002\r\n\r\nThat fix will be available in our next `datasets` library release. In the meantime, you can incorporate that fix by:\r\n- installing `datasets` from our GitHub repo:\r\n```bash\r\npip install git+https://github.com/huggingface/datasets#egg=datasets\r\n```\r\n- forcing the data files to be redownloaded\r\n```python\r\nds = load_dataset('conll2012_ontonotesv5', 'english_v4', split=\"test\", download_mode=\"force_redownload\")\r\n```\r\n\r\nFeel free to re-open this issue if the problem persists. \r\n\r\nDuplicate of:\r\n- #4031"
] | ## Steps to reproduce the bug
```python
datasets.load_dataset("conll2012_ontonotesv5", "english_v12")
```
## Actual results
```
Downloading builder script: 32.2kB [00:00, 9.72MB/s]
Downloading metadata: 20.0kB [00:00, 10.4MB/s]
Downloading and preparing dataset conll2012_ontonotesv5/english_v12 (download: 174.83 MiB, generated: 204.29 MiB, post-processed: Unknown size
, total: 379.12 MiB) to ...
Traceback (most recent call last): [315/390]
File "/home/yisiang/lgtn/conll2012/run.py", line 86, in <module>
train()
File "/home/yisiang/lgtn/conll2012/run.py", line 65, in train
trainer.fit(model, datamodule=dm)
File "/home/yisiang/miniconda3/envs/ai/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 740, in fit
self._call_and_handle_interrupt(
File "/home/yisiang/miniconda3/envs/ai/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 685, in _call_and_handle_inte
rrupt
return trainer_fn(*args, **kwargs)
File "/home/yisiang/miniconda3/envs/ai/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 777, in _fit_impl
self._run(model, ckpt_path=ckpt_path)
File "/home/yisiang/miniconda3/envs/ai/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 1131, in _run
self._data_connector.prepare_data()
File "/home/yisiang/miniconda3/envs/ai/lib/python3.9/site-packages/pytorch_lightning/trainer/connectors/data_connector.py", line 154, in pre
pare_data
self.trainer.datamodule.prepare_data()
File "/home/yisiang/miniconda3/envs/ai/lib/python3.9/site-packages/pytorch_lightning/core/datamodule.py", line 474, in wrapped_fn
fn(*args, **kwargs)
File "/home/yisiang/lgtn/_abstract_task/data.py", line 43, in prepare_data
raw_dsets = datasets.load_dataset(**load_dataset_kwargs)
File "/home/yisiang/miniconda3/envs/ai/lib/python3.9/site-packages/datasets/load.py", line 1687, in load_dataset
builder_instance.download_and_prepare(
File "/home/yisiang/miniconda3/envs/ai/lib/python3.9/site-packages/datasets/builder.py", line 605, in download_and_prepare
self._download_and_prepare(
File "/home/yisiang/miniconda3/envs/ai/lib/python3.9/site-packages/datasets/builder.py", line 1104, in _download_and_prepare
super()._download_and_prepare(dl_manager, verify_infos, check_duplicate_keys=verify_infos)
File "/home/yisiang/miniconda3/envs/ai/lib/python3.9/site-packages/datasets/builder.py", line 676, in _download_and_prepare
verify_checksums(
File "/home/yisiang/miniconda3/envs/ai/lib/python3.9/site-packages/datasets/utils/info_utils.py", line 40, in verify_checksums
raise NonMatchingChecksumError(error_msg + str(bad_urls))
datasets.utils.info_utils.NonMatchingChecksumError: Checksums didn't match for dataset source files:
['https://md-datasets-cache-zipfiles-prod.s3.eu-west-1.amazonaws.com/zmycy7t9h9-1.zip']
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.0.0 | 4,080 |
https://github.com/huggingface/datasets/issues/4077 | ArrowInvalid: Parquet magic bytes not found in footer. Either the file is corrupted or this is not a parquet file. | [] | ## Describe the bug
When uploading a relatively large image dataset of > 1GB, reloading doesn't work for me, even though pushing to the hub went just fine.
Basically, I do:
```
from datasets import load_dataset
dataset = load_dataset("imagefolder", data_files="path_to_my_files")
dataset.push_to_hub("dataset_name") # works fine, no errors
reloaded_dataset = load_dataset("dataset_name")
```
and it returns:
```
/usr/local/lib/python3.7/dist-packages/pyarrow/error.pxi in pyarrow.lib.check_status()
ArrowInvalid: Parquet magic bytes not found in footer. Either the file is corrupted or this is not a parquet file.
```
I created a Colab notebook to reproduce my error: https://colab.research.google.com/drive/141LJCcM2XyqprPY83nIQ-Zk3BbxWeahq?usp=sharing
| 4,077 |
https://github.com/huggingface/datasets/issues/4075 | Add CCAgT dataset | [
"Awesome ! Let us know if you have questions or if we can help ;) I'm assigning you\r\n\r\nPS: if possible, please try to not use Google Drive links in your dataset script, since Google Drive has download quotas and is not always reliable.",
"HI, I was waiting to come out in the second version to do the implementation.\r\n\r\n- Paper: https://dx.doi.org/10.2139/ssrn.4126881\r\n- Data: [Data mendelay](http://doi.org/10.17632/wg4bpm33hj.2)",
"Nice ! 🚀 ",
"The link of CCAgT dataset is: https://huggingface.co/datasets/lapix/CCAgT"
] | ## Adding a Dataset
- **Name:** CCAgT dataset: Images of Cervical Cells with AgNOR Stain Technique
- **Description:** The dataset contains 2540 images (1600x1200 where each pixel is 0.111μm×0.111μm) from three different slides, having at least one nucleus per image. These images are from fields belonging to a sample cervical slide, colored with silver-stained, a method known as Argyrophilic Nucleolar Organizer Regions (AgNOR).
- **Paper:** https://doi.org/10.1109/cbms49503.2020.00110
- **Data:** https://arquivos.ufsc.br/d/373be2177a33426a9e6c/ or https://drive.google.com/drive/u/4/folders/1TBpYCv6S1ydASLauSzcsvO7Wc5O-WUw0
- **Motivation:** This is a unique dataset (because of the stain), for a major health problem, cervical cancer, with real data.
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
Hi, this is a public version of the dataset that I have been working on, soon we will have another version of this dataset. But until this new version goes out, I thought I would add this dataset here, if it makes sense for the repository. You can assign the task to me if possible | 4,075 |
https://github.com/huggingface/datasets/issues/4074 | Error in google/xtreme_s dataset card | [
"Hi @wranai, thanks for reporting.\r\n\r\nPlease note that the information about language families and groups is taken form the original paper: [XTREME-S: Evaluating Cross-lingual Speech Representations](https://arxiv.org/abs/2203.10752).\r\n\r\nIf that information is wrong, feel free to contact the paper's authors to suggest that correction.\r\n\r\nJust note that Hungarian language (contrary to their geographically surrounding neighbor languages) belongs to the Uralic (languages) family, together with (among others) Finnish, Estonian, some other languages in northern regions of Scandinavia..."
] | **Link:** https://huggingface.co/datasets/google/xtreme_s
Not a big deal but Hungarian is considered an Eastern European language, together with Serbian, Slovak, Slovenian (all correctly categorized; Slovenia is mostly to the West of Hungary, by the way).
| 4,074 |
https://github.com/huggingface/datasets/issues/4071 | Loading issue for xuyeliu/notebookCDG dataset | [
"Hi @Jun-jie-Huang,\r\n\r\nAs the error message says, \".pkl\" data files are not supported.\r\n\r\nIf you would like to share your dataset on the Hub, you would need:\r\n- either to create a Python loading script, that loads the data in any format\r\n- or to transform your data files to one of the supported formats (listed in the error message above: CSV, JSON, Parquet, TXT,...)\r\n\r\nYou can find the details in our docs: \r\n- How to share a dataset: https://huggingface.co/docs/datasets/share\r\n- How to create a dataset loading script: https://huggingface.co/docs/datasets/dataset_script\r\n\r\nFeel free to re-open this issue and ping us if you need further assistance."
] | ## Dataset viewer issue for '*xuyeliu/notebookCDG*'
**Link:** *[link to the dataset viewer page](https://huggingface.co/datasets/xuyeliu/notebookCDG)*
*Couldn't load the xuyeliu/notebookCDG with provided scripts: *
```
from datasets import load_dataset
dataset = load_dataset("xuyeliu/notebookCDG/dataset_notebook.pkl")
```
I get an error message as follows:
FileNotFoundError: Couldn't find a dataset script at /home/code_documentation/code/xuyeliu/notebookCDG/notebookCDG.py or any data file in the same directory. Couldn't find 'xuyeliu/notebookCDG' on the Hugging Face Hub either: FileNotFoundError: Unable to resolve any data file that matches ['**train*'] in dataset repository xuyeliu/notebookCDG with any supported extension ['csv', 'tsv', 'json', 'jsonl', 'parquet', 'txt', 'blp', 'bmp', 'dib', 'bufr', 'cur', 'pcx', 'dcx', 'dds', 'ps', 'eps', 'fit', 'fits', 'fli', 'flc', 'ftc', 'ftu', 'gbr', 'gif', 'grib', 'h5', 'hdf', 'png', 'apng', 'jp2', 'j2k', 'jpc', 'jpf', 'jpx', 'j2c', 'icns', 'ico', 'im', 'iim', 'tif', 'tiff', 'jfif', 'jpe', 'jpg', 'jpeg', 'mpg', 'mpeg', 'msp', 'pcd', 'pxr', 'pbm', 'pgm', 'ppm', 'pnm', 'psd', 'bw', 'rgb', 'rgba', 'sgi', 'ras', 'tga', 'icb', 'vda', 'vst', 'webp', 'wmf', 'emf', 'xbm', 'xpm', 'zip']
Am I the one who added this dataset ? No
| 4,071 |
https://github.com/huggingface/datasets/issues/4062 | Loading mozilla-foundation/common_voice_7_0 dataset failed | [
"Hi @aapot, thanks for reporting.\r\n\r\nWe are investigating the cause of this issue. We will keep you informed. ",
"When making HTTP request from code line:\r\n```\r\nresponse = requests.get(f\"{_API_URL}/bucket/dataset/{path}/{use_cdn}\", timeout=10.0).json()\r\n```\r\nit cannot be decoded to JSON because it raises a 404 Not Found error.\r\n\r\nThe request is fixed if removing the `/{use_cdn}` from the URL.\r\n\r\nMaybe there was a change in the Common Voice API?\r\n\r\nCC: @anton-l @patrickvonplaten @polinaeterna ",
"We have contacted by email the data owners of the Common Voice dataset.",
"Hotfix: https://huggingface.co/datasets/mozilla-foundation/common_voice_7_0/commit/17b237961e4f7f84a2a0aea645abe5428a9d568e",
"I have also made the hotfix for all the rest of Common Voice script versions: 8.0, 6.1, 6.0,..., 1.0",
"Hey, is there anything new?\r\nI could not load the dataset.",
"cc @lhoestq @polinaeterna ",
"Hi @ngoquanghuy99! The dataset should load fine if you go through the following steps:\r\n\r\n1. Go to https://huggingface.co/datasets/mozilla-foundation/common_voice_9_0 and click \"Access repository\" if you see a message about sharing your contact information with Mozilla Foundation at the top of the page. If you've already done that then skip to step 2.\r\n2. Run the command `huggingface-cli login` in your terminal or notebook to authenticate your machine.\r\n3. Load the dataset with `use_auth_token=True`:\r\n```python\r\nfrom datasets import load_dataset\r\n\r\ndataset = load_dataset(\"mozilla-foundation/common_voice_9_0\", \"ab\", use_auth_token=True)\r\n```",
"Thanks @anton-l \r\nI could load the dataset now, but in another way.\r\nThanks anyways!",
"> Thanks @anton-l I could load the dataset now, but in another way. Thanks anyways!\r\n\r\nCan you share the \"another way\" please?"
] | ## Describe the bug
I wanted to load `mozilla-foundation/common_voice_7_0` dataset with `fi` language and `test` split from datasets on Colab/Kaggle notebook, but I am getting an error `JSONDecodeError: [Errno Expecting value] Not Found: 0` while loading it. The bug seems to affect other languages and splits too than just the `fi` and `test` split.
## Steps to reproduce the bug
```python
from datasets import load_dataset
dataset = load_dataset("mozilla-foundation/common_voice_7_0", "fi", split="test", use_auth_token="YOUR TOKEN")
```
## Expected results
load `mozilla-foundation/common_voice_7_0` dataset succesfully
## Actual results
```
JSONDecodeError Traceback (most recent call last)
/opt/conda/lib/python3.7/site-packages/requests/models.py in json(self, **kwargs)
909 try:
--> 910 return complexjson.loads(self.text, **kwargs)
911 except JSONDecodeError as e:
/opt/conda/lib/python3.7/site-packages/simplejson/__init__.py in loads(s, encoding, cls, object_hook, parse_float, parse_int, parse_constant, object_pairs_hook, use_decimal, **kw)
524 and not use_decimal and not kw):
--> 525 return _default_decoder.decode(s)
526 if cls is None:
/opt/conda/lib/python3.7/site-packages/simplejson/decoder.py in decode(self, s, _w, _PY3)
369 s = str(s, self.encoding)
--> 370 obj, end = self.raw_decode(s)
371 end = _w(s, end).end()
/opt/conda/lib/python3.7/site-packages/simplejson/decoder.py in raw_decode(self, s, idx, _w, _PY3)
399 idx += 3
--> 400 return self.scan_once(s, idx=_w(s, idx).end())
JSONDecodeError: Expecting value: line 1 column 1 (char 0)
During handling of the above exception, another exception occurred:
JSONDecodeError Traceback (most recent call last)
/tmp/ipykernel_358/370980805.py in <module>
1 # load Common Voice 7.0 dataset from Huggingface with Finnish "test" split
----> 2 test_dataset = load_dataset("mozilla-foundation/common_voice_7_0", "fi", split="test", use_auth_token=True)
/opt/conda/lib/python3.7/site-packages/datasets/load.py in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, ignore_verifications, keep_in_memory, save_infos, revision, use_auth_token, task, streaming, **config_kwargs)
1690 ignore_verifications=ignore_verifications,
1691 try_from_hf_gcs=try_from_hf_gcs,
-> 1692 use_auth_token=use_auth_token,
1693 )
1694
/opt/conda/lib/python3.7/site-packages/datasets/builder.py in download_and_prepare(self, download_config, download_mode, ignore_verifications, try_from_hf_gcs, dl_manager, base_path, use_auth_token, **download_and_prepare_kwargs)
604 if not downloaded_from_gcs:
605 self._download_and_prepare(
--> 606 dl_manager=dl_manager, verify_infos=verify_infos, **download_and_prepare_kwargs
607 )
608 # Sync info
/opt/conda/lib/python3.7/site-packages/datasets/builder.py in _download_and_prepare(self, dl_manager, verify_infos)
1102
1103 def _download_and_prepare(self, dl_manager, verify_infos):
-> 1104 super()._download_and_prepare(dl_manager, verify_infos, check_duplicate_keys=verify_infos)
1105
1106 def _get_examples_iterable_for_split(self, split_generator: SplitGenerator) -> ExamplesIterable:
/opt/conda/lib/python3.7/site-packages/datasets/builder.py in _download_and_prepare(self, dl_manager, verify_infos, **prepare_split_kwargs)
670 split_dict = SplitDict(dataset_name=self.name)
671 split_generators_kwargs = self._make_split_generators_kwargs(prepare_split_kwargs)
--> 672 split_generators = self._split_generators(dl_manager, **split_generators_kwargs)
673
674 # Checksums verification
~/.cache/huggingface/modules/datasets_modules/datasets/mozilla-foundation--common_voice_7_0/fe20cac47c166e25b1f096ab661832e3da7cf298ed4a91dcaa1343ad972d175b/common_voice_7_0.py in _split_generators(self, dl_manager)
151
152 self._log_download(self.config.name, bundle_version, hf_auth_token)
--> 153 archive = dl_manager.download(self._get_bundle_url(self.config.name, bundle_url_template))
154
155 if self.config.version < datasets.Version("5.0.0"):
~/.cache/huggingface/modules/datasets_modules/datasets/mozilla-foundation--common_voice_7_0/fe20cac47c166e25b1f096ab661832e3da7cf298ed4a91dcaa1343ad972d175b/common_voice_7_0.py in _get_bundle_url(self, locale, url_template)
130 path = urllib.parse.quote(path.encode("utf-8"), safe="~()*!.'")
131 use_cdn = self.config.size_bytes < 20 * 1024 * 1024 * 1024
--> 132 response = requests.get(f"{_API_URL}/bucket/dataset/{path}/{use_cdn}", timeout=10.0).json()
133 return response["url"]
134
/opt/conda/lib/python3.7/site-packages/requests/models.py in json(self, **kwargs)
915 raise RequestsJSONDecodeError(e.message)
916 else:
--> 917 raise RequestsJSONDecodeError(e.msg, e.doc, e.pos)
918
919 @property
JSONDecodeError: [Errno Expecting value] Not Found: 0
```
## Environment info
- `datasets` version: 2.0.0
- Platform: Linux-5.10.90+-x86_64-with-debian-bullseye-sid
- Python version: 3.7.12
- PyArrow version: 5.0.0
- Pandas version: 1.3.5
| 4,062 |
https://github.com/huggingface/datasets/issues/4061 | Loading cnn_dailymail dataset failed | [
"Hi @Arij-Aladel, thanks for reporting.\r\n\r\nThis issue was already reported \r\n- #3784\r\n\r\nand its root cause is a change in the Google Drive service. See:\r\n- #3786 \r\n\r\nWe have already fixed it in our 2.0.0 release. See:\r\n- #3787 \r\n\r\nPlease, update your `datasets` version:\r\n```\r\npip install -U datasets\r\n```\r\nand retry loading the dataset by forcing its redownload:\r\n```python\r\ndataset = load_dataset(\"cnn_dailymail\", \"3.0.0\", download_mode=\"force_redownload\")\r\n```"
] | ## Describe the bug
I wanted to load cnn_dailymail dataset from huggingface datasets on jupyter lab, but I am getting an error ` NotADirectoryError:[Errno20] Not a directory ` while loading it.
## Steps to reproduce the bug
```python
from datasets import load_dataset
dataset = load_dataset('cnn_dailymail', '3.0.0')
```
## Expected results
load `cnn_dailymail` dataset succesfully
## Actual results
failed to load and get error
> NotADirectoryError: [Errno 20] Not a directory
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` 1.8.0:
- Platform: Ubuntu-20.04
- Python version: 3.9.10
- PyArrow version: 3.0.0
| 4,061 |
https://github.com/huggingface/datasets/issues/4057 | `load_dataset` consumes too much memory for audio + tar archives | [
"Hi ! Could it be because you need to free the memory used by `tarfile` by emptying the tar `members` by any chance ?\r\n```python\r\n yield key, {\"audio\": {\"path\": audio_name, \"bytes\": audio_file_obj.read()}}\r\n audio_tarfile.members = [] # free memory\r\n key += 1\r\n```\r\n\r\nand then you can set `DEFAULT_WRITER_BATCH_SIZE` to whatever value makes more sense for your dataset.\r\n\r\nLet me know if the issue persists (which could happen, given that you managed to run your generator without RAM issues and using os.walk didn't solve the issue)",
"Thanks for your reply! Tried it but the issue persists. ",
"I also run out of memory when loading `mozilla-foundation/common_voice_8_0` that also uses `tarfile` via `dl_manager.iter_archive`. There seems to be some data files that stay in memory somewhere\r\n\r\nI don't have the issue with other compression formats like gzipped files",
"I'm facing a similar memory leak issue when loading cv8. As you said @lhoestq \r\n\r\n`load_dataset(\"mozilla-foundation/common_voice_8_0\", \"en\", use_auth_token=True, writer_batch_size=1)`\r\n\r\nThis issue is happening on a 32GB RAM machine. \r\n\r\nAny updates on how to fix this?",
"I've run a memory profiler to see where's the leak comes from:\r\n\r\n\r\n\r\n... it seems that it's related to the tarfile lib buffer reader. But I don't know why it's only happening on the huggingface script",
"I have the same problem when loading video into numpy. \r\n```\r\nyield id,{ \r\n \"video\": imageio.v3.imread(video_path),\r\n \"label\": int(label)\r\n}\r\n```\r\nSince video files are heavy, it can only processes a dozen samples before OOM.",
"For video datasets I think you can just define the max number of video that can stay in memory by adding this class attribute to your dataset builer:\r\n```py\r\nDEFAULT_WRITER_BATCH_SIZE = 8 # only 8 videos at a time in memory before flushing the dataset writer\r\n```",
"same thing happens for me with `load_dataset(\"mozilla-foundation/common_voice_8_0\", \"en\", use_auth_token=True, writer_batch_size=1)` on azure ml. seems to fill up `tmp` and not release that memory until OOM",
"I'll add that I'm encountering the same issue with\r\n`load_dataset('wikipedia', 'ceb', runner='DirectRunner', split='train')`.\r\nSame for `'es'` in place of `'ceb'`.",
"> I'll add that I'm encountering the same issue with\r\n> load_dataset('wikipedia', 'ceb', runner='DirectRunner', split='train').\r\n> Same for 'es' in place of 'ceb'.\r\n\r\nThis is because the Apache Beam `DirectRunner` runs with the full data in memory unfortunately. Optimizing the `DirectRunner` is not in the scope of the `datasets` library, but rather in the Apache Beam project I believe. If you have memory issues with the `DirectRunner`, please consider switching to a machine with more RAM, or to distributed processing runtimes like Spark, Flink or DataFlow. There is a bit of documentation here: https://huggingface.co/docs/datasets/beam",
"> > I'll add that I'm encountering the same issue with\r\n> > `load_dataset('wikipedia', 'ceb', runner='DirectRunner', split='train')`.\r\n> > Same for `'es'` in place of `'ceb'`.\r\n> \r\n> This is because the Apache Beam `DirectRunner` runs with the full data in memory unfortunately. Optimizing the `DirectRunner` is not in the scope of the `datasets` library, but rather in the Apache Beam project I believe. If you have memory issues with the `DirectRunner`, please consider switching to a machine with more RAM, or to distributed processing runtimes like Spark, Flink or DataFlow. There is a bit of documentation here: https://huggingface.co/docs/datasets/beam\r\n\r\nFair enough, but this line of code crashed an AWS instance with 1024GB of RAM! I have also tried with `Runner='Flink'` on an environment with 51GB of RAM, which also failed.\r\n\r\nApache Beam has tons of open tickets already - is it worth submitting one to them over this?",
"> Fair enough, but this line of code crashed an AWS instance with 1024GB of RAM!\r\n\r\nWhat, wikipedia is not even bigger than 20GB\r\n\r\ncc @albertvillanova",
"> > Fair enough, but this line of code crashed an AWS instance with 1024GB of RAM!\r\n> \r\n> What, wikipedia is not even bigger than 20GB\r\n> \r\n> cc @albertvillanova\r\n\r\nLuckily, on Colab you can watch the call stack at the bottom of the screen - much of the time and space complexity seems to come from `_parse_and_clean_wikicode()` rather than the actual download process. As far as I can tell, the script is loading the full dataset and then cleaning it all at once, which is consuming a lot of memory.",
"I think we are mixing many different bugs in this Issue page:\r\n- TAR archive with audio files\r\n- video file\r\n- distributed parsing of Wikipedia using Apache Beam\r\n\r\n@dan-the-meme-man may I ask you to open a separate Issue for your problem? Then I will address it. It is important to fix it because we are currently working on a Datasets enhancement to be able to provide all Wikipedias already preprocessed.\r\n\r\nOn the other hand, I think we could keep this Issue page for the original problem: TAR archive with audio files. That is not fixed yet either.",
"Is there an update on the TAR archive issue with audio files? Happy to lend a hand in fixing this :)",
"I found the issue with Common Voice 8 and opened a PR to fix it: https://huggingface.co/datasets/mozilla-foundation/common_voice_8_0/discussions/2\r\n\r\nBasically the `metadata` dict that contains the transcripts per audio file was continuously getting filled with bytes from `f.read()` because of this code:\r\n```python\r\nresult = metadata[path]\r\nresult[\"audio\"] = {\"path\": path, \"bytes\": f.read()}\r\n```\r\ncopying the result with `result = dict(metadata[path])` fixes it: the bytes are no longer added to `metadata`\r\n\r\nI also opened PRs to the other CV datasets",
"Amazing, that's a great find! Thanks @lhoestq!",
"I'm closing this one for now, but feel free to reopen if you encounter other memory issues with audio datasets"
] |
## Description
`load_dataset` consumes more and more memory until it's killed, even though it's made with a generator. I'm adding a loading script for a new dataset, made up of ~15s audio coming from a tar file. Tried setting `DEFAULT_WRITER_BATCH_SIZE = 1` as per the discussion in #741 but the problem persists.
## Steps to reproduce the bug
Here's my implementation of `_generate_examples`:
```python
class MyDatasetBuilder(datasets.GeneratorBasedBuilder):
DEFAULT_WRITER_BATCH_SIZE = 1
...
def _split_generators(self, dl_manager):
archive_path = dl_manager.download(_DL_URLS[self.config.name])
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN,
gen_kwargs={
"audio_tarfile_path": archive_path["audio_tarfile"]
},
),
]
def _generate_examples(self, audio_tarfile_path):
key = 0
with tarfile.open(audio_tarfile_path, mode="r|") as audio_tarfile:
for audio_tarinfo in audio_tarfile:
audio_name = audio_tarinfo.name
audio_file_obj = audio_tarfile.extractfile(audio_tarinfo)
yield key, {"audio": {"path": audio_name, "bytes": audio_file_obj.read()}}
key += 1
```
I then try to load via `ds = load_dataset('./datasets/my_new_dataset', writer_batch_size=1)`, and memory usage grows until all 8GB of my machine are taken and process is killed (`Killed`). Also tried an untarred version of this using `os.walk` but the same happened.
I created a script to confirm that one can safely go through such a generator, which runs just fine with memory <500MB at all times.
```python
import tarfile
def generate_examples():
audio_tarfile = tarfile.open("audios.tar", mode="r|")
key = 0
for audio_tarinfo in audio_tarfile:
audio_name = audio_tarinfo.name
audio_file_obj = audio_tarfile.extractfile(audio_tarinfo)
yield key, {"audio": {"path": audio_name, "bytes": audio_file_obj.read()}}
key += 1
if __name__ == "__main__":
examples = generate_examples()
for example in examples:
pass
```
## Expected results
Memory consumption should be similar to the non-huggingface script.
## Actual results
Process is killed after consuming too much memory.
## Environment info
- `datasets` version: 2.0.1.dev0
- Platform: Linux-4.19.0-20-cloud-amd64-x86_64-with-debian-10.12
- Python version: 3.7.12
- PyArrow version: 6.0.1
- Pandas version: 1.3.5 | 4,057 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.