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/2860 | Cannot download TOTTO dataset | [
"Hola @mrm8488, thanks for reporting.\r\n\r\nApparently, the data source host changed their URL one week ago: https://github.com/google-research-datasets/ToTTo/commit/cebeb430ec2a97747e704d16a9354f7d9073ff8f\r\n\r\nI'm fixing it."
] | Error: Couldn't find file at https://storage.googleapis.com/totto/totto_data.zip
`datasets version: 1.11.0`
# How to reproduce:
```py
from datasets import load_dataset
dataset = load_dataset('totto')
```
| 2,860 |
https://github.com/huggingface/datasets/issues/2859 | Loading allenai/c4 in streaming mode does too many HEAD requests | [
"https://github.com/huggingface/datasets/blob/6c766f9115d686182d76b1b937cb27e099c45d68/src/datasets/builder.py#L179-L186",
"Thanks a lot!!!"
] | This does 60,000+ HEAD requests to get all the ETags of all the data files:
```python
from datasets import load_dataset
load_dataset("allenai/c4", streaming=True)
```
It makes loading the dataset completely impractical.
The ETags are used to compute the config id (it must depend on the data files being used).
Instead of using the ETags, we could simply use the commit hash of the dataset repository on the hub, as well and the glob pattern used to resolve the files (here it's `*` by default, to load all the files of the repository) | 2,859 |
https://github.com/huggingface/datasets/issues/2850 | Wound segmentation datasets | [] | ## Adding a Dataset
- **Name:** Wound segmentation datasets
- **Description:** annotated wound image dataset
- **Paper:** https://www.nature.com/articles/s41598-020-78799-w
- **Data:** https://github.com/uwm-bigdata/wound-segmentation
- **Motivation:** Interesting simple image dataset, useful for segmentation, with visibility due to http://www.miccai.org/special-interest-groups/challenges/ and https://fusc.grand-challenge.org/
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| 2,850 |
https://github.com/huggingface/datasets/issues/2849 | Add Open Catalyst Project Dataset | [] | ## Adding a Dataset
- **Name:** Open Catalyst 2020 (OC20) Dataset
- **Website:** https://opencatalystproject.org/
- **Data:** https://github.com/Open-Catalyst-Project/ocp/blob/master/DATASET.md
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| 2,849 |
https://github.com/huggingface/datasets/issues/2846 | Negative timezone | [
"Fixed by #2847."
] | ## Describe the bug
The load_dataset method do not accept a parquet file with a negative timezone, as it has the following regex:
```
"^(s|ms|us|ns),\s*tz=([a-zA-Z0-9/_+:]*)$"
```
So a valid timestap ```timestamp[us, tz=-03:00]``` returns an error when loading parquet files.
## Steps to reproduce the bug
```python
# Where the timestamp column has a tz of -03:00
datasets = load_dataset('parquet', data_files={'train': train_files, 'validation': validation_files,
'test': test_files}, cache_dir="./cache_teste/")
```
## Expected results
The -03:00 is a valid tz so the regex should accept this without raising an error.
## Actual results
As this regex disaproves a valid tz it raises the following error:
```python
raise ValueError(
f"{datasets_dtype} is not a validly formatted string representation of a pyarrow timestamp."
f"Examples include timestamp[us] or timestamp[us, tz=America/New_York]"
f"See: https://arrow.apache.org/docs/python/generated/pyarrow.timestamp.html#pyarrow.timestamp"
)
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.11.0
- Platform: Ubuntu 20.04
- Python version: 3.8
- PyArrow version: 5.0.0
| 2,846 |
https://github.com/huggingface/datasets/issues/2845 | [feature request] adding easy to remember `datasets.cache_dataset()` + `datasets.is_dataset_cached()` | [] | Often, there is a need to prepare a dataset but not use it immediately, e.g. think tests suite setup, so it'd be really useful to be able to do:
```
if not datasets.is_dataset_cached(ds): datasets.cache_dataset(ds)
```
This can already be done with:
```
builder = load_dataset_builder(ds)
if not os.path.idsir(builder.cache_dir):
builder.download_and_prepare()
```
but the current way is a way less intuitive and much harder to remember than the proposed API, IMHO.
One more way is to do:
```
_ = load_dataset(ds)
```
but it wastes resources loading the dataset when it's not needed.
this has been discussed at https://huggingface.slack.com/archives/C01229B19EX/p1630021912025800
Thank you!
@lhoestq
| 2,845 |
https://github.com/huggingface/datasets/issues/2842 | always requiring the username in the dataset name when there is one | [
"From what I can understand, you want the saved arrow file directory to have username as well instead of just dataset name if it was downloaded with the user prefix?",
"I don't think the user cares of how this is done, but the 2nd command should fail, IMHO, as its dataset name is invalid:\r\n```\r\n# first run\r\npython -c \"from datasets import load_dataset; load_dataset('stas/openwebtext-10k')\"\r\n# now run immediately\r\npython -c \"from datasets import load_dataset; load_dataset('openwebtext-10k')\"\r\n# the second command should fail, but it doesn't fail now.\r\n```\r\n\r\nMoreover, if someone were to create `openwebtext-10k` w/o the prefix, they will now get the wrong dataset, if they previously downloaded `stas/openwebtext-10k`.\r\n\r\nAnd if there are 2 users with the same dataset name `foo/ds` and `bar/ds` - currently this won't work to get the correct dataset.\r\n\r\nSo really there 3 unrelated issues hiding in the current behavior.",
"This has been fixed now, and we'll do a new release of the library today.\r\n\r\nNow the stas/openwebtext-10k dataset is cached at `.cache/huggingface/datasets/stas___openwebtext10k` and openwebtext-10k would be at `.cache/huggingface/datasets/openwebtext10k`. Since they are different, the cache won't fall back on loading the wrong one anymore.\r\n\r\nSame for the python script used to generate the dataset: stas/openwebtext-10k is cached at `.cache/huggingface/modules/datasets_modules/datasets/stas___openwebtext10k` and openwebtext-10k would be at `.cache/huggingface/modules/datasets_modules/datasets/openwebtext10k`",
"Amazing! Thank you for adding this improvement, @lhoestq!",
"(can be closed?)",
"Yes indeed :) thanks"
] | Me and now another person have been bitten by the `datasets`'s non-strictness on requiring a dataset creator's username when it's due.
So both of us started with `stas/openwebtext-10k`, somewhere along the lines lost `stas/` and continued using `openwebtext-10k` and it all was good until we published the software and things broke, since there is no `openwebtext-10k`
So this feature request is asking to tighten the checking and not allow dataset loading if it was downloaded with the user prefix, but then attempted to be used w/o it.
The same in code:
```
# first run
python -c "from datasets import load_dataset; load_dataset('stas/openwebtext-10k')"
# now run immediately
python -c "from datasets import load_dataset; load_dataset('openwebtext-10k')"
# the second command should fail, but it doesn't fail now.
```
Please let me know if I explained myself clearly.
Thank you! | 2,842 |
https://github.com/huggingface/datasets/issues/2841 | Adding GLUECoS Hinglish and Spanglish code-switching bemchmark | [
"Hi @yjernite I am interested in adding this dataset. \r\nIn the repo they have also added a code mixed MT task from English to Hinglish [here](https://github.com/microsoft/GLUECoS#code-mixed-machine-translation-task). I think this could be a good dataset addition in itself and then I can add the rest of the GLUECoS tasks as one dataset. What do you think?"
] | ## Adding a Dataset
- **Name:** GLUECoS
- **Description:** a Microsoft Benchmark to evaluate code-switching for only two language pairs but a variety of tasks
- **Paper:** https://aclanthology.org/2020.acl-main.329/
- **Data:** https://github.com/microsoft/GLUECoS
- **Motivation:** We currently only have [one other](https://huggingface.co/datasets/lince) dataset for code-switching
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| 2,841 |
https://github.com/huggingface/datasets/issues/2840 | How can I compute BLEU-4 score use `load_metric` ? | [] | I have found the sacrebleu metric. But, I do not know the difference between it and BLEU-4.
If I want to compute BLEU-4 score, what can i do? | 2,840 |
https://github.com/huggingface/datasets/issues/2839 | OpenWebText: NonMatchingSplitsSizesError | [
"Thanks for reporting, I'm updating the verifications metadata",
"I just regenerated the verifications metadata and noticed that nothing changed: the data file is fine (the checksum didn't change), and the number of examples is still 8013769. Not sure how you managed to get 7982430 examples.\r\n\r\nCan you try to delete your cache ( by default at `~/.cache/huggingface/datasets`) and try again please ?\r\nAlso, on which platform are you (linux/macos/windows) ?",
"I'll try without deleting the whole cache (we have large datasets already stored). I was under the impression that `download_mode=\"force_redownload\"` would bypass cache.\r\nSorry plateform should be linux (Redhat version 8.1)",
"Hi @thomasw21 , are you still having this issue after clearing your cache ?",
"Sorry I haven't had time to work on this. I'll close and re-open if I can't figure out why I'm having this issue. Thanks for taking a look !"
] | ## Describe the bug
When downloading `openwebtext`, I'm getting:
```
datasets.utils.info_utils.NonMatchingSplitsSizesError: [{'expected': SplitInfo(name='train', num_bytes=39769494896, num_examples=8013769, dataset_name='openwebtext'), 'recorded': SplitInfo(name='train', num_bytes=39611023912, num_examples=7982430, dataset_name='openwebtext')}]
```
I suspect that the file we download from has changed since the size doesn't look like to match with documentation
`Downloading: 0%| | 0.00/12.9G [00:00<?, ?B/s]` This suggest the total size is 12.9GB, whereas the one documented mentions `Size of downloaded dataset files: 12283.35 MB`.
## Steps to reproduce the bug
```python
from datasets import load_dataset
load_dataset("openwebtext", download_mode="force_redownload")
```
## Expected results
Loading is successful
## Actual results
Loading throws above error.
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.10.2
- Platform: linux (Redhat version 8.1)
- Python version: 3.8
- PyArrow version: 4.0.1
| 2,839 |
https://github.com/huggingface/datasets/issues/2837 | prepare_module issue when loading from read-only fs | [
"Hello, I opened #2887 to fix this."
] | ## Describe the bug
When we use prepare_module from a readonly file system, we create a FileLock using the `local_path`.
This path is not necessarily writable.
`lock_path = local_path + ".lock"`
## Steps to reproduce the bug
Run `load_dataset` on a readonly python loader file.
```python
ds = load_dataset(
python_loader, data_files={"train": train_path, "test": test_path}
)
```
where `python_loader` is a path to a file located in a readonly folder.
## Expected results
This should work I think?
## Actual results
```python
return load_dataset(
File "/usr/local/lib/python3.8/dist-packages/datasets/load.py", line 711, in load_dataset
module_path, hash, resolved_file_path = prepare_module(
File "/usr/local/lib/python3.8/dist-packages/datasets/load.py", line 465, in prepare_module
with FileLock(lock_path):
File "/usr/local/lib/python3.8/dist-packages/datasets/utils/filelock.py", line 314, in __enter__
self.acquire()
File "/usr/local/lib/python3.8/dist-packages/datasets/utils/filelock.py", line 263, in acquire
self._acquire()
File "/usr/local/lib/python3.8/dist-packages/datasets/utils/filelock.py", line 378, in _acquire
fd = os.open(self._lock_file, open_mode)
OSError: [Errno 30] Read-only file system: 'YOUR_FILE.py.lock'
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.7.0
- Platform: macOS-10.15.7-x86_64-i386-64bit
- Python version: 3.8.8
- PyArrow version: 3.0.0
| 2,837 |
https://github.com/huggingface/datasets/issues/2833 | IndexError when accessing first element of a Dataset if first RecordBatch is empty | [] | The computation of the offsets of the underlying Table of a Dataset has some issues if the first RecordBatch is empty.
```python
from datasets import Dataset
import pyarrow as pa
pa_table = pa.Table.from_pydict({"a": [1]})
pa_table2 = pa.Table.from_pydict({"a": []}, schema=pa_table.schema)
ds_table = pa.concat_tables([pa_table2, pa_table])
dataset = Dataset(ds_table)
print([len(b) for b in dataset.data._batches])
# [0, 1]
print(dataset.data._offsets)
# [0 0 1] (should be [0, 1])
dataset[0]
```
raises
```python
---------------------------------------------------------------------------
IndexError Traceback (most recent call last)
/usr/local/lib/python3.7/dist-packages/datasets/table.py in _interpolation_search(arr, x)
90 else:
91 i, j = i, k
---> 92 raise IndexError(f"Invalid query '{x}' for size {arr[-1] if len(arr) else 'none'}.")
93
94
IndexError: Invalid query '0' for size 1.
```
This can be fixed by ignoring empty batches when computing `table._batches` and `table._offsets`
cc @SaulLu | 2,833 |
https://github.com/huggingface/datasets/issues/2832 | Logging levels not taken into account | [
"I just take a look at all the outputs produced by `datasets` using the different log-levels.\r\nAs far as i can tell using `datasets==1.17.0` they overall issue seems to be fixed.\r\n\r\nHowever, I noticed that there is one tqdm based progress indicator appearing on STDERR that I can simply not suppress.\r\n```\r\nResolving data files: 100%|██████████| 652/652 [00:00<00:00, 1604.52it/s]\r\n```\r\n\r\nAccording to _get_origin_metadata_locally_or_by_urls it shold be supressable by using the `NOTSET` log-level\r\nhttps://github.com/huggingface/datasets/blob/1406a04c3e911cec2680d8bc513653e0cafcaaa4/src/datasets/data_files.py#L491-L501\r\nSadly when specifiing the log-level `NOTSET` it seems to has no effect.\r\n\r\nBut appart from it not having any effect I must admit that it seems unintuitive to me.\r\nI would suggest changing this such that it is only shown when the log-level is greater or equal to INFO.\r\n\r\nThis would conform better to INFO according to the [documentation](https://huggingface.co/docs/datasets/v1.0.0/package_reference/logging_methods.html#datasets.logging.set_verbosity_info).\r\n> This will display most of the logging information and tqdm bars.\r\n\r\nAny inputs on this?\r\nI will be happy to supply a PR if desired 👍 ",
"Hi! This should disable the tqdm output:\r\n```python\r\nimport datasets\r\ndatasets.set_progress_bar_enabled(False)\r\n```\r\n\r\nOn a side note: I believe the issue with logging (not tqdm) is still relevant on master."
] | ## Describe the bug
The `logging` module isn't working as intended relative to the levels to set.
## Steps to reproduce the bug
```python
from datasets import logging
logging.set_verbosity_debug()
logger = logging.get_logger()
logger.error("ERROR")
logger.warning("WARNING")
logger.info("INFO")
logger.debug("DEBUG"
```
## Expected results
I expect all logs to be output since I'm putting a `debug` level.
## Actual results
Only the two first logs are output.
## Environment info
- `datasets` version: 1.11.0
- Platform: Linux-5.13.9-arch1-1-x86_64-with-glibc2.33
- Python version: 3.9.6
- PyArrow version: 5.0.0
## To go further
This logging issue appears in `datasets` but not in `transformers`. It happens because there is no handler defined for the logger. When no handler is defined, the `logging` library will output a one-off error to stderr, using a `StderrHandler` with level `WARNING`.
`transformers` sets a default `StreamHandler` [here](https://github.com/huggingface/transformers/blob/5c6eca71a983bae2589eed01e5c04fcf88ba5690/src/transformers/utils/logging.py#L86) | 2,832 |
https://github.com/huggingface/datasets/issues/2831 | ArrowInvalid when mapping dataset with missing values | [
"Hi ! It fails because of the feature type inference.\r\n\r\nBecause the first 1000 examples all have null values in the \"match\" field, then it infers that the type for this field is `null` type before writing the data on disk. But as soon as it tries to map an example with a non-null \"match\" field, then it fails.\r\n\r\nTo fix that you can either:\r\n- increase the writer_batch_size to >2000 (default is 1000) so that some non-null values will be in the first batch written to disk\r\n```python\r\ndatasets = datasets.map(lambda e: {'labels': e['match']}, remove_columns=['id'], writer_batch_size=2000)\r\n```\r\n- OR force the feature type with:\r\n```python\r\nfrom datasets import Features, Value\r\n\r\nfeatures = Features({\r\n 'conflict': Value('int64'),\r\n 'date': Value('string'),\r\n 'headline': Value('string'),\r\n 'match': Value('float64'),\r\n 'label': Value('float64')\r\n})\r\ndatasets = datasets.map(lambda e: {'labels': e['match']}, remove_columns=['id'], features=features)\r\n```"
] | ## Describe the bug
I encountered an `ArrowInvalid` when mapping dataset with missing values.
Here are the files for a minimal example. The exception is only thrown when the first line in the csv has a missing value (if you move the last line to the top it isn't thrown).
[data_small.csv](https://github.com/huggingface/datasets/files/7037838/data_small.csv)
[data.csv](https://github.com/huggingface/datasets/files/7037842/data.csv)
## Steps to reproduce the bug
```python
from datasets import load_dataset
datasets = load_dataset("csv", data_files=['data_small.csv'])
datasets = datasets.map(lambda e: {'labels': e['match']},
remove_columns=['id'])
```
## Expected results
No error
## Actual results
```
File "pyarrow/error.pxi", line 84, in pyarrow.lib.check_status
pyarrow.lib.ArrowInvalid: Invalid null value
```
## Environment info
- `datasets` version: 1.5.0
- Platform: Linux-5.11.0-25-generic-x86_64-with-glibc2.29
- Python version: 3.8.10
- PyTorch version (GPU?): 1.7.1+cpu (False)
- Tensorflow version (GPU?): 2.4.1 (False)
- Using GPU in script?: no
- Using distributed or parallel set-up in script?: no
| 2,831 |
https://github.com/huggingface/datasets/issues/2829 | Optimize streaming from TAR archives | [
"Closed by: \r\n- #3066"
] | Hi ! As you know TAR has some constraints for data streaming. While it is optimized for buffering, the files in the TAR archive **need to be streamed in order**. It means that we can't choose which file to stream from, and this notation is to be avoided for TAR archives:
```
tar://books_large_p1.txt::https://storage.googleapis.com/huggingface-nlp/datasets/bookcorpus/bookcorpus.tar.bz2
```
Instead, I suggest we implement `iter_archive` for the `StreamingDownloadManager`.
The regular `DownloadManager` already has it.
Then we will have to update the json/txt/csv/etc. loaders to make them use `iter_archive` on TAR archives.
That's also what Tensorflow Datasets is doing in this case.
See this [dataset](https://github.com/tensorflow/datasets/blob/93895059c80a9e05805e8f32a2e310f66a23fc98/tensorflow_datasets/image_classification/flowers.py) for example.
Therefore instead of doing
```python
uncompressed = dl_manager.extract(tar_archive)
filename = "books_large_p1.txt"
with open(os.path.join(uncompressed, filename)) as f:
for line in f:
...
```
we'll do
```python
for filename, f in dl_manager.iter_archive(tar_archive):
for line in f:
...
``` | 2,829 |
https://github.com/huggingface/datasets/issues/2826 | Add a Text Classification dataset: KanHope | [
"Hi ! In your script it looks like you're trying to load the dataset `bn_hate_speech,`, not KanHope.\r\n\r\nMoreover the error `KeyError: ' '` means that you have a feature of type ClassLabel, but for a certain example of the dataset, it looks like the label is empty (it's just a string with a space). Can you make sure that the data don't have missing labels, and that your dataset script parses the labels correctly ?"
] | ## Adding a Dataset
- **Name:** *KanHope*
- **Description:** *A code-mixed English-Kannada dataset for Hope speech detection*
- **Paper:** *https://arxiv.org/abs/2108.04616* (I am the author of the paper}
- **Author:** *[AdeepH](https://github.com/adeepH)*
- **Data:** *https://github.com/adeepH/KanHope/tree/main/dataset*
- **Motivation:** *The dataset is amongst the very few resources available for code-mixed Dravidian languages*
- I tried following the steps as per the instructions. However, could not resolve an error. Any help would be appreciated.
- The dataset card and the scripts for the dataset *https://github.com/adeepH/datasets/tree/multilingual-hope-speech/datasets/mhs_eval*
```
Using custom data configuration default
Downloading and preparing dataset bn_hate_speech/default (download: Unknown size, generated: Unknown size, post-processed: Unknown size, total: Unknown size) to /root/.cache/huggingface/datasets/bn_hate_speech/default/0.0.0/5f417ddc89777278abd29988f909f39495f0ec802090f7d8fa63b5bffb121762...
---------------------------------------------------------------------------
KeyError Traceback (most recent call last)
<ipython-input-114-4a9cdb519e4c> in <module>()
1 from datasets import load_dataset
2
----> 3 data = load_dataset('/content/bn')
9 frames
/usr/local/lib/python3.7/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, script_version, use_auth_token, task, streaming, **config_kwargs)
850 ignore_verifications=ignore_verifications,
851 try_from_hf_gcs=try_from_hf_gcs,
--> 852 use_auth_token=use_auth_token,
853 )
854
/usr/local/lib/python3.7/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)
614 if not downloaded_from_gcs:
615 self._download_and_prepare(
--> 616 dl_manager=dl_manager, verify_infos=verify_infos, **download_and_prepare_kwargs
617 )
618 # Sync info
/usr/local/lib/python3.7/dist-packages/datasets/builder.py in _download_and_prepare(self, dl_manager, verify_infos, **prepare_split_kwargs)
691 try:
692 # Prepare split will record examples associated to the split
--> 693 self._prepare_split(split_generator, **prepare_split_kwargs)
694 except OSError as e:
695 raise OSError(
/usr/local/lib/python3.7/dist-packages/datasets/builder.py in _prepare_split(self, split_generator)
1107 disable=bool(logging.get_verbosity() == logging.NOTSET),
1108 ):
-> 1109 example = self.info.features.encode_example(record)
1110 writer.write(example, key)
1111 finally:
/usr/local/lib/python3.7/dist-packages/datasets/features.py in encode_example(self, example)
1015 """
1016 example = cast_to_python_objects(example)
-> 1017 return encode_nested_example(self, example)
1018
1019 def encode_batch(self, batch):
/usr/local/lib/python3.7/dist-packages/datasets/features.py in encode_nested_example(schema, obj)
863 if isinstance(schema, dict):
864 return {
--> 865 k: encode_nested_example(sub_schema, sub_obj) for k, (sub_schema, sub_obj) in utils.zip_dict(schema, obj)
866 }
867 elif isinstance(schema, (list, tuple)):
/usr/local/lib/python3.7/dist-packages/datasets/features.py in <dictcomp>(.0)
863 if isinstance(schema, dict):
864 return {
--> 865 k: encode_nested_example(sub_schema, sub_obj) for k, (sub_schema, sub_obj) in utils.zip_dict(schema, obj)
866 }
867 elif isinstance(schema, (list, tuple)):
/usr/local/lib/python3.7/dist-packages/datasets/features.py in encode_nested_example(schema, obj)
890 # ClassLabel will convert from string to int, TranslationVariableLanguages does some checks
891 elif isinstance(schema, (ClassLabel, TranslationVariableLanguages, Value, _ArrayXD)):
--> 892 return schema.encode_example(obj)
893 # Other object should be directly convertible to a native Arrow type (like Translation and Translation)
894 return obj
/usr/local/lib/python3.7/dist-packages/datasets/features.py in encode_example(self, example_data)
665 # If a string is given, convert to associated integer
666 if isinstance(example_data, str):
--> 667 example_data = self.str2int(example_data)
668
669 # Allowing -1 to mean no label.
/usr/local/lib/python3.7/dist-packages/datasets/features.py in str2int(self, values)
623 if value not in self._str2int:
624 value = str(value).strip()
--> 625 output.append(self._str2int[str(value)])
626 else:
627 # No names provided, try to integerize
KeyError: ' '
``` | 2,826 |
https://github.com/huggingface/datasets/issues/2825 | The datasets.map function does not load cached dataset after moving python script | [
"This also happened to me on COLAB.\r\nDetails:\r\nI ran the `run_mlm.py` in two different notebooks. \r\nIn the first notebook, I do tokenization since I can get 4 CPU cores without any GPUs, and save the cache into a folder which I copy to drive.\r\nIn the second notebook, I copy the cache folder from drive and re-run the run_mlm.py script (this time I uncomment the trainer code which happens after the tokenization)\r\n\r\nNote: I didn't change anything in the arguments, not even the preprocessing_num_workers\r\n ",
"Thanks for reporting ! This is indeed a bug, I'm looking into it",
"#2854 fixed the issue :)\r\n\r\nWe'll do a new release of `datasets` soon to make the fix available.\r\nIn the meantime, feel free to try it out by installing `datasets` from source\r\n\r\nIf you have other issues or any question, feel free to re-open the issue :)"
] | ## Describe the bug
The datasets.map function caches the processed data to a certain directory. When the map function is called another time with totally the same parameters, the cached data are supposed to be reloaded instead of re-processing. However, it doesn't reuse cached data sometimes. I use the common data processing in different tasks, the datasets are processed again, the only difference is that I run them in different files.
## Steps to reproduce the bug
Just run the following codes in different .py files.
```python
if __name__ == '__main__':
from datasets import load_dataset
from transformers import AutoTokenizer
raw_datasets = load_dataset("wikitext", "wikitext-2-raw-v1")
tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
def tokenize_function(examples):
return tokenizer(examples["text"], padding="max_length", truncation=True)
tokenized_datasets = raw_datasets.map(tokenize_function, batched=True)
```
## Expected results
The map function should reload data in the second or any later runs.
## Actual results
The processing happens in each run.
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.8.0
- Platform: linux
- Python version: 3.7.6
- PyArrow version: 3.0.0
This is the first time I report a bug. If there is any problem or confusing description, please let me know 😄.
| 2,825 |
https://github.com/huggingface/datasets/issues/2823 | HF_DATASETS_CACHE variable in Windows | [
"Agh - I'm a muppet. No quote marks are needed.\r\nset HF_DATASETS_CACHE = C:\\Datasets\r\nworks as intended."
] | I can't seem to use a custom Cache directory in Windows. I have tried:
set HF_DATASETS_CACHE = "C:\Datasets"
set HF_DATASETS_CACHE = "C:/Datasets"
set HF_DATASETS_CACHE = "C:\\Datasets"
set HF_DATASETS_CACHE = "r'C:\Datasets'"
set HF_DATASETS_CACHE = "\Datasets"
set HF_DATASETS_CACHE = "/Datasets"
In each instance I get the "[WinError 123] The filename, directory name, or volume label syntax is incorrect" error when attempting to load a dataset | 2,823 |
https://github.com/huggingface/datasets/issues/2821 | Cannot load linnaeus dataset | [
"Thanks for reporting ! #2852 fixed this error\r\n\r\nWe'll do a new release of `datasets` soon :)"
] | ## Describe the bug
The [linnaeus](https://huggingface.co/datasets/linnaeus) dataset cannot be loaded. To reproduce:
```
from datasets import load_dataset
datasets = load_dataset("linnaeus")
```
This results in:
```
Downloading and preparing dataset linnaeus/linnaeus (download: 17.36 MiB, generated: 8.74 MiB, post-processed: Unknown size, total: 26.10 MiB) to /root/.cache/huggingface/datasets/linnaeus/linnaeus/1.0.0/2ff05dbc256108233262f596e09e322dbc3db067202de14286913607cd9cb704...
---------------------------------------------------------------------------
ConnectionError Traceback (most recent call last)
<ipython-input-4-7ef3a88f6276> in <module>()
1 from datasets import load_dataset
2
----> 3 datasets = load_dataset("linnaeus")
11 frames
/usr/local/lib/python3.7/dist-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)
603 raise FileNotFoundError("Couldn't find file at {}".format(url))
604 _raise_if_offline_mode_is_enabled(f"Tried to reach {url}")
--> 605 raise ConnectionError("Couldn't reach {}".format(url))
606
607 # Try a second time
ConnectionError: Couldn't reach https://drive.google.com/u/0/uc?id=1OletxmPYNkz2ltOr9pyT0b0iBtUWxslh&export=download/
``` | 2,821 |
https://github.com/huggingface/datasets/issues/2820 | Downloading “reddit” dataset keeps timing out. | [
"```\r\nUsing custom data configuration default\r\nDownloading and preparing dataset reddit/default (download: 2.93 GiB, generated: 17.64 GiB, post-processed: Unknown size, total: 20.57 GiB) to /Volumes/My Passport for Mac/og-chat-data/reddit/default/1.0.0/98ba5abea674d3178f7588aa6518a5510dc0c6fa8176d9653a3546d5afcb3969...\r\nDownloading: 13%\r\n403M/3.14G [44:39<2:27:09, 310kB/s]\r\n---------------------------------------------------------------------------\r\ntimeout Traceback (most recent call last)\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/urllib3/response.py in _error_catcher(self)\r\n 437 try:\r\n--> 438 yield\r\n 439 \r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/urllib3/response.py in read(self, amt, decode_content, cache_content)\r\n 518 cache_content = False\r\n--> 519 data = self._fp.read(amt) if not fp_closed else b\"\"\r\n 520 if (\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/http/client.py in read(self, amt)\r\n 458 b = bytearray(amt)\r\n--> 459 n = self.readinto(b)\r\n 460 return memoryview(b)[:n].tobytes()\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/http/client.py in readinto(self, b)\r\n 502 # (for example, reading in 1k chunks)\r\n--> 503 n = self.fp.readinto(b)\r\n 504 if not n and b:\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/socket.py in readinto(self, b)\r\n 703 try:\r\n--> 704 return self._sock.recv_into(b)\r\n 705 except timeout:\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/ssl.py in recv_into(self, buffer, nbytes, flags)\r\n 1240 self.__class__)\r\n-> 1241 return self.read(nbytes, buffer)\r\n 1242 else:\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/ssl.py in read(self, len, buffer)\r\n 1098 if buffer is not None:\r\n-> 1099 return self._sslobj.read(len, buffer)\r\n 1100 else:\r\n\r\ntimeout: The read operation timed out\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nReadTimeoutError Traceback (most recent call last)\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/requests/models.py in generate()\r\n 757 try:\r\n--> 758 for chunk in self.raw.stream(chunk_size, decode_content=True):\r\n 759 yield chunk\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/urllib3/response.py in stream(self, amt, decode_content)\r\n 575 while not is_fp_closed(self._fp):\r\n--> 576 data = self.read(amt=amt, decode_content=decode_content)\r\n 577 \r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/urllib3/response.py in read(self, amt, decode_content, cache_content)\r\n 540 # Content-Length are caught.\r\n--> 541 raise IncompleteRead(self._fp_bytes_read, self.length_remaining)\r\n 542 \r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/contextlib.py in __exit__(self, type, value, traceback)\r\n 134 try:\r\n--> 135 self.gen.throw(type, value, traceback)\r\n 136 except StopIteration as exc:\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/urllib3/response.py in _error_catcher(self)\r\n 442 # there is yet no clean way to get at it from this context.\r\n--> 443 raise ReadTimeoutError(self._pool, None, \"Read timed out.\")\r\n 444 \r\n\r\nReadTimeoutError: HTTPSConnectionPool(host='zenodo.org', port=443): Read timed out.\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nConnectionError Traceback (most recent call last)\r\n/var/folders/3f/md0t9sgj6rz8xy01fskttqdc0000gn/T/ipykernel_89016/1133441872.py in <module>\r\n 1 from datasets import load_dataset\r\n 2 \r\n----> 3 dataset = load_dataset(\"reddit\", ignore_verifications=True, cache_dir=\"/Volumes/My Passport for Mac/og-chat-data\")\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/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, script_version, use_auth_token, task, streaming, **config_kwargs)\r\n 845 \r\n 846 # Download and prepare data\r\n--> 847 builder_instance.download_and_prepare(\r\n 848 download_config=download_config,\r\n 849 download_mode=download_mode,\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/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)\r\n 613 logger.warning(\"HF google storage unreachable. Downloading and preparing it from source\")\r\n 614 if not downloaded_from_gcs:\r\n--> 615 self._download_and_prepare(\r\n 616 dl_manager=dl_manager, verify_infos=verify_infos, **download_and_prepare_kwargs\r\n 617 )\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/datasets/builder.py in _download_and_prepare(self, dl_manager, verify_infos, **prepare_split_kwargs)\r\n 669 split_dict = SplitDict(dataset_name=self.name)\r\n 670 split_generators_kwargs = self._make_split_generators_kwargs(prepare_split_kwargs)\r\n--> 671 split_generators = self._split_generators(dl_manager, **split_generators_kwargs)\r\n 672 \r\n 673 # Checksums verification\r\n\r\n~/.cache/huggingface/modules/datasets_modules/datasets/reddit/98ba5abea674d3178f7588aa6518a5510dc0c6fa8176d9653a3546d5afcb3969/reddit.py in _split_generators(self, dl_manager)\r\n 73 def _split_generators(self, dl_manager):\r\n 74 \"\"\"Returns SplitGenerators.\"\"\"\r\n---> 75 dl_path = dl_manager.download_and_extract(_URL)\r\n 76 return [\r\n 77 datasets.SplitGenerator(\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/datasets/utils/download_manager.py in download_and_extract(self, url_or_urls)\r\n 287 extracted_path(s): `str`, extracted paths of given URL(s).\r\n 288 \"\"\"\r\n--> 289 return self.extract(self.download(url_or_urls))\r\n 290 \r\n 291 def get_recorded_sizes_checksums(self):\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/datasets/utils/download_manager.py in download(self, url_or_urls)\r\n 195 \r\n 196 start_time = datetime.now()\r\n--> 197 downloaded_path_or_paths = map_nested(\r\n 198 download_func,\r\n 199 url_or_urls,\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/datasets/utils/py_utils.py in map_nested(function, data_struct, dict_only, map_list, map_tuple, map_numpy, num_proc, types)\r\n 194 # Singleton\r\n 195 if not isinstance(data_struct, dict) and not isinstance(data_struct, types):\r\n--> 196 return function(data_struct)\r\n 197 \r\n 198 disable_tqdm = bool(logger.getEffectiveLevel() > logging.INFO) or not utils.is_progress_bar_enabled()\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/datasets/utils/download_manager.py in _download(self, url_or_filename, download_config)\r\n 218 # append the relative path to the base_path\r\n 219 url_or_filename = url_or_path_join(self._base_path, url_or_filename)\r\n--> 220 return cached_path(url_or_filename, download_config=download_config)\r\n 221 \r\n 222 def iter_archive(self, path):\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/datasets/utils/file_utils.py in cached_path(url_or_filename, download_config, **download_kwargs)\r\n 286 if is_remote_url(url_or_filename):\r\n 287 # URL, so get it from the cache (downloading if necessary)\r\n--> 288 output_path = get_from_cache(\r\n 289 url_or_filename,\r\n 290 cache_dir=cache_dir,\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/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)\r\n 643 ftp_get(url, temp_file)\r\n 644 else:\r\n--> 645 http_get(\r\n 646 url,\r\n 647 temp_file,\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/datasets/utils/file_utils.py in http_get(url, temp_file, proxies, resume_size, headers, cookies, timeout, max_retries)\r\n 451 disable=bool(logging.get_verbosity() == logging.NOTSET),\r\n 452 )\r\n--> 453 for chunk in response.iter_content(chunk_size=1024):\r\n 454 if chunk: # filter out keep-alive new chunks\r\n 455 progress.update(len(chunk))\r\n\r\n/usr/local/anaconda3/envs/og-data-env/lib/python3.9/site-packages/requests/models.py in generate()\r\n 763 raise ContentDecodingError(e)\r\n 764 except ReadTimeoutError as e:\r\n--> 765 raise ConnectionError(e)\r\n 766 else:\r\n 767 # Standard file-like object.\r\n\r\nConnectionError: HTTPSConnectionPool(host='zenodo.org', port=443): Read timed out.\r\n```",
"Hey @lhoestq should I try to fix this issue ?",
"It also doesn't seem to be \"smart caching\" and I received an error about a file not being found...",
"To be clear, the error I get when I try to \"re-instantiate\" the download after failure is: \r\n```\r\nOSError: Cannot find data file. \r\nOriginal error:\r\n[Errno 20] Not a directory: <HOME>/.cache/huggingface/datasets/downloads/1ec12301abba4daa60eb3a90e53529b5b173296b22dc3bef3186e205c75e594c/corpus-webis-tldr-17.json'\r\n```",
"Here is a new error:\r\n```\r\nConnectionError: Couldn't reach https://zenodo.org/record/1043504/files/corpus-webis-tldr-17.zip?download=1\r\n```",
"Hi ! Since https://github.com/huggingface/datasets/pull/2803 we've changed the time out from 10sec to 100sec.\r\nThis should prevent the `ReadTimeoutError`. Feel free to try it out by installing `datasets` from source\r\n```\r\npip install git+https://github.com/huggingface/datasets.git\r\n```\r\n\r\nWhen re-running your code you said you get a `OSError`, could you try deleting the file at the path returned by the error ? (the one after `[Errno 20] Not a directory:`). Ideally when a download fails you should be able to re-run it without error; there might be an issue here.\r\n\r\nFinally not sure what we can do about `ConnectionError`, this must be an issue from zenodo. If it happens you simply need to try again\r\n",
"@lhoestq thanks for the update. The directory specified by the OSError ie. \r\n```\r\n1ec12301abba4daa60eb3a90e53529b5b173296b22dc3bef3186e205c75e594c/corpus-webis-tldr-17.json \r\n```\r\n was not actually in that directory so I can't delete it. ",
"Oh, then could you try deleting the parent directory `1ec12301abba4daa60eb3a90e53529b5b173296b22dc3bef3186e205c75e594c` instead ?\r\nThis way the download manager will know that it has to uncompress the data again",
"It seems to have worked. It only took like 20min! I think the extra timeout length did the trick! One thing is that it downloaded a total of 41gb instead of 20gb but at least it finished. ",
"Great ! The timeout change will be available in the next release of `datasets` :)"
] | ## Describe the bug
A clear and concise description of what the bug is.
Everytime I try and download the reddit dataset it times out before finishing and I have to try again.
There is some timeout error that I will post once it happens again.
## Steps to reproduce the bug
```python
from datasets import load_dataset
dataset = load_dataset("reddit", ignore_verifications=True, cache_dir="/Volumes/My Passport for Mac/og-chat-data")
```
## Expected results
A clear and concise description of the expected results.
I would expect the download to finish, or at least provide a parameter to extend the read timeout window.
## Actual results
Specify the actual results or traceback.
Shown below in error message.
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.11.0
- Platform: macOS
- Python version: 3.9.6 (conda env)
- PyArrow version: N/A
| 2,820 |
https://github.com/huggingface/datasets/issues/2818 | cannot load data from my loacal path | [
"Hi ! The `data_files` parameter must be a string, a list/tuple or a python dict.\r\n\r\nCan you check the type of your `config.train_path` please ? Or use `data_files=str(config.train_path)` ?"
] | ## Describe the bug
I just want to directly load data from my local path,but find a bug.And I compare it with pandas to provide my local path is real.
here is my code
```python3
# print my local path
print(config.train_path)
# read data and print data length
tarin=pd.read_csv(config.train_path)
print(len(tarin))
# loading data by load_dataset
data = load_dataset('csv',data_files=config.train_path)
print(len(data))
```
## Steps to reproduce the bug
```python
C:\Users\wie\Documents\项目\文本分类\data\train.csv
7613
Traceback (most recent call last):
File "c:/Users/wie/Documents/项目/文本分类/lib/DataPrecess.py", line 17, in <module>
data = load_dataset('csv',data_files=config.train_path)
File "C:\Users\wie\Miniconda3\lib\site-packages\datasets\load.py", line 830, in load_dataset
**config_kwargs,
File "C:\Users\wie\Miniconda3\lib\site-packages\datasets\load.py", line 710, in load_dataset_builder
**config_kwargs,
File "C:\Users\wie\Miniconda3\lib\site-packages\datasets\builder.py", line 271, in __init__
**config_kwargs,
File "C:\Users\wie\Miniconda3\lib\site-packages\datasets\builder.py", line 386, in _create_builder_config
config_kwargs, custom_features=custom_features, use_auth_token=self.use_auth_token
File "C:\Users\wie\Miniconda3\lib\site-packages\datasets\builder.py", line 156, in create_config_id
raise ValueError("Please provide a valid `data_files` in `DatasetBuilder`")
ValueError: Please provide a valid `data_files` in `DatasetBuilder`
```
## 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: 1.11.0
- Platform: win10
- Python version: 3.7.9
- PyArrow version: 5.0.0
| 2,818 |
https://github.com/huggingface/datasets/issues/2816 | Add Mostly Basic Python Problems Dataset | [
"I started working on that."
] | ## Adding a Dataset
- **Name:** Mostly Basic Python Problems Dataset
- **Description:** The benchmark consists of around 1,000 crowd-sourced Python programming problems, designed to be solvable by entry level programmers, covering programming fundamentals, standard library functionality, and so on. Each problem consists of a task description, code solution and 3 automated test cases.
- **Paper:** *link to the dataset paper if available*
- **Data:** https://github.com/google-research/google-research/tree/master/mbpp
- **Motivation:** Simple, small dataset related to coding problems.
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| 2,816 |
https://github.com/huggingface/datasets/issues/2813 | Remove compression from xopen | [
"After discussing with @lhoestq, a reasonable alternative:\r\n- `download_manager.extract(urlpath)` adds prefixes to `urlpath` in the same way as `fsspec` does for protocols, but we implement custom prefixes for all compression formats: \r\n `bz2::http://domain.org/filename.bz2`\r\n- `xopen` parses the `urlpath` and extracts the `compression` parameter and passes it to `fsspec.open`:\r\n `fsspec.open(\"http://domain.org/filename.bz2\", compression=\"bz2\")`\r\n\r\nPros:\r\n- clean solution that continues giving support to all compression formats\r\n- no breaking change when opening non-decompressed files: if no compression-protocol-like is passed, fsspec.open does not uncompress (passes compression=None)\r\n\r\nCons:\r\n- we create a \"private\" convention for the format of `urlpath`: although similar to `fsspec` protocols, we add custom prefixes for the `compression` argument"
] | We implemented support for streaming with 2 requirements:
- transparent use for the end user: just needs to pass the parameter `streaming=True`
- no additional work for the contributors: previous loading scripts should also work in streaming mode with no (or minor) changes; and new loading scripts should not involve additional code to support streaming
In order to fulfill these requirements, streaming implementation patched some Python functions:
- the `open(urlpath)` function was patched with `fsspec.open(urlpath)`
- the `os.path.join(urlpath, *others)` function was patched in order to add to `urlpath` hops (`::`) and extractor protocols (`zip://`), which are required by `fsspec.open`
Recently, we implemented support for streaming all archive+compression formats: zip, tar, gz, bz2, lz4, xz, zst; tar.gz, tar.bz2,...
Under the hood, the implementation:
- passes an additional parameter `compression` to `fsspec.open`, so that it performs the decompression on the fly: `fsspec.open(urlpath, compression=...)`
Some concerns have been raised about passing the parameter `compression` to `fsspec.open`:
- https://github.com/huggingface/datasets/pull/2786#discussion_r689550254
- #2811
The main argument is that if `open` decompresses the file and afterwards we call `gzip.open` on it, that will raise an error in `oscar` dataset:
```python
gzip.open(open(urlpath
```
While this is true:
- it is not natural/usual to call `open` inside `gzip.open` (never seen this before)
- indeed, this was recently (2 months ago) coded that way in `datasets` in order to allow streaming support (with previous implementation of streaming)
In this particular case, there is a natural fix solution: #2811:
- Revert the `open` inside the `gzip.open` (change done 2 months ago): `gzip.open(open(urlpath` => `gzip.open(urlpath`
- Patch `gzip.open(urlpath` with `fsspec.open(urlpath, compression="gzip"`
Are there other issues apart from this?
Note that there is an issue just because the open inside of the gzip.open. There is no issue in the other cases where datasets loading scripts use just
- `gzip.open`
- `open` (after having called dl_manager.download_and_extract)
TODO:
- [ ] Is this really an issue? Please enumerate the `datasets` loading scripts where this is problematic.
- For the moment, there are only 3 datasets where we have an `open` inside a `gzip.open`:
- oscar (since 23 June), mc4 (since 2 July) and c4 (since 2 July)
- In the 3 datasets, the only reason to put an open inside a gzip.open was indeed to force supporting streaming
- [ ] If this is indeed an issue, which are the possible alternatives? Pros/cons? | 2,813 |
https://github.com/huggingface/datasets/issues/2812 | arXiv Dataset verification problem | [] | ## Describe the bug
`dataset_infos.json` for `arxiv_dataset` contains a fixed number of training examples, however the data (downloaded from an external source) is updated every week with additional examples.
Therefore, loading the dataset without `ignore_verifications=True` results in a verification error. | 2,812 |
https://github.com/huggingface/datasets/issues/2808 | Enable streaming for Wikipedia corpora | [
"Closing as this has been addressed in https://github.com/huggingface/datasets/pull/5689."
] | **Is your feature request related to a problem? Please describe.**
Several of the [Wikipedia corpora](https://huggingface.co/datasets?search=wiki) on the Hub involve quite large files that would be a good candidate for streaming. Currently it is not possible to stream these corpora:
```python
from datasets import load_dataset
# Throws ValueError: Builder wikipedia is not streamable.
wiki_dataset_streamed = load_dataset("wikipedia", "20200501.en", split="train", streaming=True)
```
Given that these corpora are derived from Wikipedia dumps in XML format which are then processed with Apache Beam, I am not sure whether streaming is possible in principle. The goal of this issue is to discuss whether this feature even makes sense :)
**Describe the solution you'd like**
It would be nice to be able to stream Wikipedia corpora from the Hub with something like
```python
from datasets import load_dataset
wiki_dataset_streamed = load_dataset("wikipedia", "20200501.en", split="train", streaming=True)
``` | 2,808 |
https://github.com/huggingface/datasets/issues/2799 | Loading JSON throws ArrowNotImplementedError | [
"Hi @lewtun, thanks for reporting.\r\n\r\nApparently, `pyarrow.json` tries to cast timestamp-like fields in your JSON file to pyarrow timestamp type, and it fails with `ArrowNotImplementedError`.\r\n\r\nI will investigate if there is a way to tell pyarrow not to try that timestamp casting.",
"I think the issue is more complex than that...\r\n\r\nI just took one of your JSON lines and pyarrow.json read it without problem.",
"> I just took one of your JSON lines an pyarrow.json read it without problem.\r\n\r\nyes, and for some peculiar reason the error is non-deterministic (i was eventually able to load the whole dataset by just re-running the `load_dataset` cell multiple times 🤔)\r\n\r\nthanks for looking into this 🙏 !",
"I think the error is generated by the `pyarrow.json.read()` option: `read_options=paj.ReadOptions(block_size=block_size)`...\r\ncc: @lhoestq ",
"The code works fine on my side.\r\nNot sure what's going on here :/\r\n\r\nI remember we did a few changes in the JSON loader in #2638 , did you do an update `datasets` when debugging this ?\r\n",
"OK after upgrading `datasets` to v1.12.1 the issue seems to have gone away. Closing this now :)",
"Oops, I spoke too soon 😓 \r\n\r\nAfter deleting the cache and trying the above code snippet again I am hitting the same error. You can also reproduce it in the Colab notebook I linked to in the issue description. ",
"@albertvillanova @lhoestq I noticed the same issue using datasets v1.12.1. Is there an update on when this could be fixed?",
"Apparently it's possible to make it work by increasing the `block_size`, let me open a PR",
"I just opened a PR with a fix, feel free to install `datasets` from source from source and let me know if it helps",
"@zijwang did PR #3000 solve the problem for you? It did for me, so it all is good on your end we can close this issue. Thanks again to @lhoestq for the pyarrow magic 🤯 "
] | ## Describe the bug
I have created a [dataset](https://huggingface.co/datasets/lewtun/github-issues-test) of GitHub issues in line-separated JSON format and am finding that I cannot load it with the `json` loading script (see stack trace below).
Curiously, there is no problem loading the dataset with `pandas` which suggests some incorrect type inference is being made on the `datasets` side. For example, the stack trace indicates that some URL fields are being parsed as timestamps.
You can find a Colab notebook which reproduces the error [here](https://colab.research.google.com/drive/1YUCM0j1vx5ZrouQbYSzal6RwB4-Aoh4o?usp=sharing).
**Edit:** If one repeatedly tries to load the dataset, it _eventually_ works but I think it would still be good to understand why it fails in the first place :)
## Steps to reproduce the bug
```python
from datasets import load_dataset
from huggingface_hub import hf_hub_url
import pandas as pd
# returns https://huggingface.co/datasets/lewtun/github-issues-test/resolve/main/issues-datasets.jsonl
data_files = hf_hub_url(repo_id="lewtun/github-issues-test", filename="issues-datasets.jsonl", repo_type="dataset")
# throws ArrowNotImplementedError
dset = load_dataset("json", data_files=data_files, split="test")
# no problem with pandas ...
df = pd.read_json(data_files, orient="records", lines=True)
df.head()
```
## Expected results
I can load any line-separated JSON file, similar to `pandas`.
## Actual results
```
---------------------------------------------------------------------------
ArrowNotImplementedError Traceback (most recent call last)
<ipython-input-7-5b8e82b6c3a2> in <module>()
----> 1 dset = load_dataset("json", data_files=data_files, split="test")
9 frames
/usr/local/lib/python3.7/dist-packages/pyarrow/error.pxi in pyarrow.lib.check_status()
ArrowNotImplementedError: JSON conversion to struct<url: timestamp[s], html_url: timestamp[s], labels_url: timestamp[s], id: int64, node_id: timestamp[s], number: int64, title: timestamp[s], description: timestamp[s], creator: struct<login: timestamp[s], id: int64, node_id: timestamp[s], avatar_url: timestamp[s], gravatar_id: timestamp[s], url: timestamp[s], html_url: timestamp[s], followers_url: timestamp[s], following_url: timestamp[s], gists_url: timestamp[s], starred_url: timestamp[s], subscriptions_url: timestamp[s], organizations_url: timestamp[s], repos_url: timestamp[s], events_url: timestamp[s], received_events_url: timestamp[s], type: timestamp[s], site_admin: bool>, open_issues: int64, closed_issues: int64, state: timestamp[s], created_at: timestamp[s], updated_at: timestamp[s], due_on: timestamp[s], closed_at: timestamp[s]> is not supported
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.11.0
- Platform: Linux-5.4.104+-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.11
- PyArrow version: 3.0.0
| 2,799 |
https://github.com/huggingface/datasets/issues/2797 | Make creating/editing dataset cards easier, by editing on site and dumping info from test command. | [] | **Is your feature request related to a problem? Please describe.**
Creating and editing dataset cards should be but not that easy
- If other else know Some information I don't know (bias of dataset, dataset curation, supported dataset, ...), he/she should know the description on hf.co comes from README.md under github huggingface/datasets/datasets/the dataset, and willing to make a pr to add or fix information.
- Many information is also saved in `dataset_info.json` (citaion, description), but still need to write it down to README.md again.
- Contributor need to pip install and start a local server just for tagging the dataset's size. And contributor may be creating the dataset on lab's server, which can't open a browser.
- if any one proposes a new tag, it doesn't show in the list that another creator see. (a stackoverflow way may be ideal)
- dataset card generator web app doesn't generate the necessary subsecion `Contributions` for us.
**Describe the solution you'd like**
- Everyone (or at least the author/contributor) can edit the description, information, tags of the dataset, on hf.co website. Just like wikipedia+stackoverflow
- We can infer the actual data size, citation, data instance, ... from `dataset_info.json` and `dataset.arrow` via `dataset-cli test`
| 2,797 |
https://github.com/huggingface/datasets/issues/2794 | Warnings and documentation about pickling incorrect | [] | ## Describe the bug
I have a docs bug and a closely related docs enhancement suggestion!
### Bug
The warning and documentation say "either `dill` or `pickle`" for fingerprinting. But it seems that `dill`, which is installed by `datasets` by default, _must_ work, or else the fingerprinting fails.
Warning:
https://github.com/huggingface/datasets/blob/450b9174765374111e5c6daab0ed294bc3d9b639/src/datasets/fingerprint.py#L262
Docs:
> For a transform to be hashable, it needs to be pickleable using dill or pickle.
> – [docs](https://huggingface.co/docs/datasets/processing.html#fingerprinting)
For my code, `pickle` works, but `dill` fails. The `dill` failure has already been reported in https://github.com/huggingface/datasets/issues/2643. However, the `dill` failure causes a hashing failure in the datasets library, without any backing off to `pickle`. This implies that it's not the case that either `dill` **or** `pickle` can work, but that `dill` must work if it is installed. I think this is more accurate wording, since it is installed and used by default:
https://github.com/huggingface/datasets/blob/c93525dc291346e54212567fa72d7d607befe937/setup.py#L83
... and the hashing will fail if it fails.
### Enhancement
I think it'd be very helpful to add to the documentation how to debug hashing failures. It took me a while to figure out how to diagnose this. There is a very nice two-liner by @lhoestq in https://github.com/huggingface/datasets/issues/2516#issuecomment-865173139:
```python
from datasets.fingerprint import Hasher
Hasher.hash(my_object)
```
I think add this to the docs will help future users quickly debug any hashing troubles of their own :-)
## Steps to reproduce the bug
`dill` but not `pickle` hashing failure in https://github.com/huggingface/datasets/issues/2643
## Expected results
If either `dill` or `pickle` can successfully hash, the hashing will succeed.
## Actual results
If `dill` or `pickle` cannot hash, the hashing fails.
## Environment info
- `datasets` version: 1.9.0
- Platform: Linux-5.8.0-1038-gcp-x86_64-with-glibc2.31
- Python version: 3.9.6
- PyArrow version: 4.0.1
| 2,794 |
https://github.com/huggingface/datasets/issues/2788 | How to sample every file in a list of files making up a split in a dataset when loading? | [
"Hi ! This is not possible just with `load_dataset`.\r\n\r\nYou can do something like this instead:\r\n```python\r\nseed=42\r\ndata_files_dict = {\r\n \"train\": [train_file1, train_file2],\r\n \"test\": [test_file1, test_file2],\r\n \"val\": [val_file1, val_file2]\r\n}\r\ndataset = datasets.load_dataset(\r\n \"csv\",\r\n data_files=data_files_dict,\r\n).shuffle(seed=seed)\r\n\r\nsample_dataset = {splitname: split.select(range(8)) for splitname, split in dataset.items()}\r\n```\r\n\r\nAnother alternative is loading each file separately with `split=\"train[:8]\"` and then use `concatenate_datasets` to merge the sample of each file."
] | I am loading a dataset with multiple train, test, and validation files like this:
```
data_files_dict = {
"train": [train_file1, train_file2],
"test": [test_file1, test_file2],
"val": [val_file1, val_file2]
}
dataset = datasets.load_dataset(
"csv",
data_files=data_files_dict,
split=['train[:8]', 'test[:8]', 'val[:8]']
)
```
However, this only selects the first 8 rows from train_file1, test_file1, val_file1, since they are the first files in the lists.
I'm trying to formulate a split argument that can sample from each file specified in my list of files that make up each split.
Is this type of splitting supported? If so, how can I do it? | 2,788 |
https://github.com/huggingface/datasets/issues/2787 | ConnectionError: Couldn't reach https://raw.githubusercontent.com | [
"the bug code locate in ：\r\n if data_args.task_name is not None:\r\n # Downloading and loading a dataset from the hub.\r\n datasets = load_dataset(\"glue\", data_args.task_name, cache_dir=model_args.cache_dir)",
"Hi @jinec,\r\n\r\nFrom time to time we get this kind of `ConnectionError` coming from the github.com website: https://raw.githubusercontent.com\r\n\r\nNormally, it should work if you wait a little and then retry.\r\n\r\nCould you please confirm if the problem persists?",
"cannot connect，even by Web browser，please check that there is some problems。",
"I can access https://raw.githubusercontent.com/huggingface/datasets/1.7.0/datasets/glue/glue.py without problem...",
"> I can access https://raw.githubusercontent.com/huggingface/datasets/1.7.0/datasets/glue/glue.py without problem...\r\n\r\nI can not access https://raw.githubusercontent.com/huggingface/datasets either, I am in China",
"Finally i can access it, by the superfast software. Thanks",
"> Finally i can access it, by the superfast software. Thanks\r\n\r\nExcuse me, I have the same problem as you, could you please tell me how to solve it？",
"It is not related to the area, the ConnectionError with http://raw.githubuserconent.com has persisted with load_data function, datasets module. However, it can be set to either wget or ssl snippet to download dataset from github as following. \r\n\r\n`$ wget https://raw.githubusercontent.com/... --no-check-certificate`\r\n\r\n\r\nor \r\n\r\nfor the tfds, nltk or pandas.read_csv downloading as follows. \r\n\r\n```\r\nimport ssl\r\n\r\ntry:\r\n _create_unverified_https_context = ssl._create_unverified_context\r\nexcept AttributeError:\r\n pass\r\nelse:\r\n ssl._create_default_https_context = _create_unverified_https_context\r\n```\r\n\r\nSo it is most probably the problem of github rather than users \r\n",
"> > I can access https://raw.githubusercontent.com/huggingface/datasets/1.7.0/datasets/glue/glue.py without problem...\r\n> \r\n> I can not access https://raw.githubusercontent.com/huggingface/datasets either, I am in China\r\n\r\n所以老哥怎么解决这个问题呢"
] | Hello,
I am trying to run run_glue.py and it gives me this error -
Traceback (most recent call last):
File "E:/BERT/pytorch_hugging/transformers/examples/pytorch/text-classification/run_glue.py", line 546, in <module>
main()
File "E:/BERT/pytorch_hugging/transformers/examples/pytorch/text-classification/run_glue.py", line 250, in main
datasets = load_dataset("glue", data_args.task_name, cache_dir=model_args.cache_dir)
File "C:\install\Anaconda3\envs\huggingface\lib\site-packages\datasets\load.py", line 718, in load_dataset
use_auth_token=use_auth_token,
File "C:\install\Anaconda3\envs\huggingface\lib\site-packages\datasets\load.py", line 320, in prepare_module
local_path = cached_path(file_path, download_config=download_config)
File "C:\install\Anaconda3\envs\huggingface\lib\site-packages\datasets\utils\file_utils.py", line 291, in cached_path
use_auth_token=download_config.use_auth_token,
File "C:\install\Anaconda3\envs\huggingface\lib\site-packages\datasets\utils\file_utils.py", line 623, in get_from_cache
raise ConnectionError("Couldn't reach {}".format(url))
ConnectionError: Couldn't reach https://raw.githubusercontent.com/huggingface/datasets/1.7.0/datasets/glue/glue.py
Trying to do python run_glue.py --model_name_or_path
bert-base-cased
--task_name
mrpc
--do_train
--do_eval
--max_seq_length
128
--per_device_train_batch_size
32
--learning_rate
2e-5
--num_train_epochs
3
--output_dir
./tmp/mrpc/
Is this something on my end? From what I can tell, this was re-fixeded by @fullyz a few months ago.
Thank you!
| 2,787 |
https://github.com/huggingface/datasets/issues/2781 | Latest v2.0.0 release of sacrebleu has broken some metrics | [] | ## Describe the bug
After `sacrebleu` v2.0.0 release (see changes here: https://github.com/mjpost/sacrebleu/pull/152/files#diff-2553a315bb1f7e68c9c1b00d56eaeb74f5205aeb3a189bc3e527b122c6078795L17-R15), some of `datasets` metrics are broken:
- Default tokenizer `sacrebleu.DEFAULT_TOKENIZER` no longer exists:
- #2739
- #2778
- Bleu tokenizers are no longer accessible with `sacrebleu.TOKENIZERS`:
- #2779
- `corpus_bleu` args have been renamed from `(sys_stream, ref_streams)` to `(hipotheses, references)`:
- #2782 | 2,781 |
https://github.com/huggingface/datasets/issues/2776 | document `config.HF_DATASETS_OFFLINE` and precedence | [] | https://github.com/huggingface/datasets/pull/1976 implemented `HF_DATASETS_OFFLINE`, but:
1. `config.HF_DATASETS_OFFLINE` is not documented
2. the precedence is not documented (env, config)
I'm thinking it probably should be similar to what it says https://huggingface.co/docs/datasets/loading_datasets.html#from-the-huggingface-hub about `datasets.config.IN_MEMORY_MAX_SIZE`:
Quote:
> The default in 🤗 Datasets is to memory-map the dataset on disk unless you set datasets.config.IN_MEMORY_MAX_SIZE different from 0 bytes (default). In that case, the dataset will be copied in-memory if its size is smaller than datasets.config.IN_MEMORY_MAX_SIZE bytes, and memory-mapped otherwise. This behavior can be enabled by setting either the configuration option datasets.config.IN_MEMORY_MAX_SIZE (higher precedence) or the environment variable HF_DATASETS_IN_MEMORY_MAX_SIZE (lower precedence) to nonzero.
Context: trying to use `config.HF_DATASETS_OFFLINE` here:
https://github.com/bigscience-workshop/Megatron-DeepSpeed/pull/48
but are uncertain if it's safe, since it's not documented as a public API.
Thank you!
@lhoestq, @albertvillanova | 2,776 |
https://github.com/huggingface/datasets/issues/2775 | `generate_random_fingerprint()` deterministic with 🤗Transformers' `set_seed()` | [
"I dug into what I believe is the root of this issue and added a repro in my comment. If this is better addressed as a cross-team issue, let me know and I can open an issue in the Transformers repo",
"Hi !\r\n\r\nIMO we shouldn't try to modify `set_seed` from transformers but maybe make `datasets` have its own RNG just to generate random fingerprints.\r\n\r\nAny opinion on this @LysandreJik ?",
"Yes, this sounds good @lhoestq "
] | ## Describe the bug
**Update:** I dug into this to try to reproduce the underlying issue, and I believe it's that `set_seed()` from the `transformers` library makes the "random" fingerprint identical each time. I believe this is still a bug, because `datasets` is used exactly this way in `transformers` after `set_seed()` has been called, and I think that using `set_seed()` is a standard procedure to aid reproducibility. I've added more details to reproduce this below.
Hi there! I'm using my own local dataset and custom preprocessing function. My preprocessing function seems to be unpickle-able, perhaps because it is from a closure (will debug this separately). I get this warning, which is expected:
https://github.com/huggingface/datasets/blob/450b9174765374111e5c6daab0ed294bc3d9b639/src/datasets/fingerprint.py#L260-L265
However, what's not expected is that the `datasets` actually _does_ seem to cache and reuse this dataset between runs! After that line, the next thing that's logged looks like:
```text
Loading cached processed dataset at /home/xxx/.cache/huggingface/datasets/csv/default-xxx/0.0.0/xxx/cache-xxx.arrow
```
The path is exactly the same each run (e.g., last 26 runs).
This becomes a problem because I'll pass in the `--max_eval_samples` flag to the HuggingFace example script I'm running off of ([run_swag.py](https://github.com/huggingface/transformers/blob/master/examples/pytorch/multiple-choice/run_swag.py)). The fact that the cached dataset is reused means this flag gets ignored. I'll try to load 100 examples, and it will load the full cached 1,000,000.
I think that
https://github.com/huggingface/datasets/blob/450b9174765374111e5c6daab0ed294bc3d9b639/src/datasets/fingerprint.py#L248
... is actually consistent because randomness is being controlled in HuggingFace/Transformers for reproducibility. I've added a demo of this below.
## Steps to reproduce the bug
```python
# Contents of print_fingerprint.py
from transformers import set_seed
from datasets.fingerprint import generate_random_fingerprint
set_seed(42)
print(generate_random_fingerprint())
```
```bash
for i in {0..10}; do
python print_fingerprint.py
done
1c80317fa3b1799d
1c80317fa3b1799d
1c80317fa3b1799d
1c80317fa3b1799d
1c80317fa3b1799d
1c80317fa3b1799d
1c80317fa3b1799d
1c80317fa3b1799d
1c80317fa3b1799d
1c80317fa3b1799d
1c80317fa3b1799d
```
## Expected results
After the "random hash" warning is emitted, a random hash is generated, and no outdated cached datasets are reused.
## Actual results
After the "random hash" warning is emitted, an identical hash is generated each time, and an outdated cached dataset is reused each run.
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.9.0
- Platform: Linux-5.8.0-1038-gcp-x86_64-with-glibc2.31
- Python version: 3.9.6
- PyArrow version: 4.0.1 | 2,775 |
https://github.com/huggingface/datasets/issues/2773 | Remove dataset_infos.json | [
"This was closed by:\r\n- #4926"
] | **Is your feature request related to a problem? Please describe.**
As discussed, there are infos in the `dataset_infos.json` which are redundant and we could have them only in the README file.
Others could be migrated to the README, like: "dataset_size", "size_in_bytes", "download_size", "splits.split_name.[num_bytes, num_examples]",...
However, there are others that do not seem too meaningful in the README, like the checksums.
**Describe the solution you'd like**
Open a discussion to decide what to do with the `dataset_infos.json` files: which information to be migrated and/or which information to be kept.
cc: @julien-c @lhoestq | 2,773 |
https://github.com/huggingface/datasets/issues/2772 | Remove returned feature constrain | [] | In the current version, the returned value of the map function has to be list or ndarray. However, this makes it unsuitable for many tasks. In NLP, many features are sparse like verb words, noun chunks, if we want to assign different values to different words, which will result in a large sparse matrix if we only score useful words like verb words.
Mostly, when using it on large scale, saving it as a whole takes a lot of disk storage and making it hard to read, the normal method is saving it in sparse form. However, the NumPy does not support sparse, therefore I have to use PyTorch or scipy to transform a matrix into special sparse form, which is not a form that can be transformed into list or ndarry. This violates the feature constraints of the map function.
I do appreciate the convenience of Datasets package, but I do not think the compulsory datatype constrain is necessary, in some cases, we just cannot transform it into a list or ndarray due to some reasons. Any way to fix this? Or what I can do to disable the compulsory datatype constrain?
| 2,772 |
https://github.com/huggingface/datasets/issues/2768 | `ArrowInvalid: Added column's length must match table's length.` after using `select` | [
"Hi,\r\n\r\nthe `select` method creates an indices mapping and doesn't modify the underlying PyArrow table by default for better performance. To modify the underlying table after the `select` call, call `flatten_indices` on the dataset object as follows:\r\n```python\r\nfrom datasets import load_dataset\r\n\r\nds = load_dataset(\"tweets_hate_speech_detection\")['train'].select(range(128))\r\nds = ds.flatten_indices()\r\nds = ds.add_column('ones', [1]*128)\r\n```",
"Thanks for the question @lvwerra. And thanks for the answer @mariosasko. ^^"
] | ## Describe the bug
I would like to add a column to a downsampled dataset. However I get an error message saying the length don't match with the length of the unsampled dataset indicated. I suspect that the dataset size is not updated when calling `select`.
## Steps to reproduce the bug
```python
from datasets import load_dataset
ds = load_dataset("tweets_hate_speech_detection")['train'].select(range(128))
ds = ds.add_column('ones', [1]*128)
```
## Expected results
I would expect a new column named `ones` filled with `1`. When I check the length of `ds` it says `128`. Interestingly, it works when calling `ds = ds.map(lambda x: x)` before adding the column.
## Actual results
Specify the actual results or traceback.
```python
---------------------------------------------------------------------------
ArrowInvalid Traceback (most recent call last)
/var/folders/l4/2905jygx4tx5jv8_kn03vxsw0000gn/T/ipykernel_6301/868709636.py in <module>
1 from datasets import load_dataset
2 ds = load_dataset("tweets_hate_speech_detection")['train'].select(range(128))
----> 3 ds = ds.add_column('ones', [0]*128)
~/git/semantic-clustering/env/lib/python3.8/site-packages/datasets/arrow_dataset.py in wrapper(*args, **kwargs)
183 }
184 # apply actual function
--> 185 out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
186 datasets: List["Dataset"] = list(out.values()) if isinstance(out, dict) else [out]
187 # re-apply format to the output
~/git/semantic-clustering/env/lib/python3.8/site-packages/datasets/fingerprint.py in wrapper(*args, **kwargs)
395 # Call actual function
396
--> 397 out = func(self, *args, **kwargs)
398
399 # Update fingerprint of in-place transforms + update in-place history of transforms
~/git/semantic-clustering/env/lib/python3.8/site-packages/datasets/arrow_dataset.py in add_column(self, name, column, new_fingerprint)
2965 column_table = InMemoryTable.from_pydict({name: column})
2966 # Concatenate tables horizontally
-> 2967 table = ConcatenationTable.from_tables([self._data, column_table], axis=1)
2968 # Update features
2969 info = self.info.copy()
~/git/semantic-clustering/env/lib/python3.8/site-packages/datasets/table.py in from_tables(cls, tables, axis)
715 table_blocks = to_blocks(table)
716 blocks = _extend_blocks(blocks, table_blocks, axis=axis)
--> 717 return cls.from_blocks(blocks)
718
719 @property
~/git/semantic-clustering/env/lib/python3.8/site-packages/datasets/table.py in from_blocks(cls, blocks)
663 return cls(table, blocks)
664 else:
--> 665 table = cls._concat_blocks_horizontally_and_vertically(blocks)
666 return cls(table, blocks)
667
~/git/semantic-clustering/env/lib/python3.8/site-packages/datasets/table.py in _concat_blocks_horizontally_and_vertically(cls, blocks)
623 if not tables:
624 continue
--> 625 pa_table_horizontally_concatenated = cls._concat_blocks(tables, axis=1)
626 pa_tables_to_concat_vertically.append(pa_table_horizontally_concatenated)
627 return cls._concat_blocks(pa_tables_to_concat_vertically, axis=0)
~/git/semantic-clustering/env/lib/python3.8/site-packages/datasets/table.py in _concat_blocks(blocks, axis)
612 else:
613 for name, col in zip(table.column_names, table.columns):
--> 614 pa_table = pa_table.append_column(name, col)
615 return pa_table
616 else:
~/git/semantic-clustering/env/lib/python3.8/site-packages/pyarrow/table.pxi in pyarrow.lib.Table.append_column()
~/git/semantic-clustering/env/lib/python3.8/site-packages/pyarrow/table.pxi in pyarrow.lib.Table.add_column()
~/git/semantic-clustering/env/lib/python3.8/site-packages/pyarrow/error.pxi in pyarrow.lib.pyarrow_internal_check_status()
~/git/semantic-clustering/env/lib/python3.8/site-packages/pyarrow/error.pxi in pyarrow.lib.check_status()
ArrowInvalid: Added column's length must match table's length. Expected length 31962 but got length 128
```
## Environment info
- `datasets` version: 1.11.0
- Platform: macOS-10.16-x86_64-i386-64bit
- Python version: 3.8.5
- PyArrow version: 5.0.0
| 2,768 |
https://github.com/huggingface/datasets/issues/2767 | equal operation to perform unbatch for huggingface datasets | [
"Hi @lhoestq \r\nMaybe this is clearer to explain like this, currently map function, map one example to \"one\" modified one, lets assume we want to map one example to \"multiple\" examples, in which we do not know in advance how many examples they would be per each entry. I greatly appreciate telling me how I can handle this operation, thanks a lot",
"Hi,\r\nthis is also my question on how to perform similar operation as \"unbatch\" in tensorflow in great huggingface dataset library. \r\nthanks.",
"Hi,\r\n\r\n`Dataset.map` in the batched mode allows you to map a single row to multiple rows. So to perform \"unbatch\", you can do the following:\r\n```python\r\nimport collections\r\n\r\ndef unbatch(batch):\r\n new_batch = collections.defaultdict(list)\r\n keys = batch.keys()\r\n for values in zip(*batch.values()):\r\n ex = {k: v for k, v in zip(keys, values)}\r\n inputs = f\"record query: {ex['query']} entities: {', '.join(ex['entities'])} passage: {ex['passage']}\"\r\n new_batch[\"inputs\"].extend([inputs] * len(ex[\"answers\"]))\r\n new_batch[\"targets\"].extend(ex[\"answers\"])\r\n return new_batch\r\n\r\ndset = dset.map(unbatch, batched=True, remove_columns=dset.column_names)\r\n```",
"Dear @mariosasko \r\nFirst, thank you very much for coming back to me on this, I appreciate it a lot. I tried this solution, I am getting errors, do you mind\r\ngiving me one test example to be able to run your code, to understand better the format of the inputs to your function?\r\nin this function https://github.com/google-research/text-to-text-transfer-transformer/blob/3c58859b8fe72c2dbca6a43bc775aa510ba7e706/t5/data/preprocessors.py#L952 they copy each example to the number of \"answers\", do you mean one should not do the copying part and use directly your function? \r\n\r\n\r\nthank you very much for your help and time.",
"Hi @mariosasko \r\nI think finally I got this, I think you mean to do things in one step, here is the full example for completeness:\r\n\r\n```\r\ndef unbatch(batch):\r\n new_batch = collections.defaultdict(list)\r\n keys = batch.keys()\r\n for values in zip(*batch.values()):\r\n ex = {k: v for k, v in zip(keys, values)}\r\n # updates the passage.\r\n passage = ex['passage']\r\n passage = re.sub(r'(\\.|\\?|\\!|\\\"|\\')\\n@highlight\\n', r'\\1 ', passage)\r\n passage = re.sub(r'\\n@highlight\\n', '. ', passage)\r\n inputs = f\"record query: {ex['query']} entities: {', '.join(ex['entities'])} passage: {passage}\"\r\n # duplicates the samples based on number of answers.\r\n num_answers = len(ex[\"answers\"])\r\n num_duplicates = np.maximum(1, num_answers)\r\n new_batch[\"inputs\"].extend([inputs] * num_duplicates) #len(ex[\"answers\"]))\r\n new_batch[\"targets\"].extend(ex[\"answers\"] if num_answers > 0 else [\"<unk>\"])\r\n return new_batch\r\n\r\ndata = datasets.load_dataset('super_glue', 'record', split=\"train\", script_version=\"master\")\r\ndata = data.map(unbatch, batched=True, remove_columns=data.column_names)\r\n```\r\n\r\nThanks a lot again, this was a super great way to do it."
] | Hi
I need to use "unbatch" operation in tensorflow on a huggingface dataset, I could not find this operation, could you kindly direct me how I can do it, here is the problem I am trying to solve:
I am considering "record" dataset in SuperGlue and I need to replicate each entery of the dataset for each answer, to make it similar to what T5 originally did:
https://github.com/google-research/text-to-text-transfer-transformer/blob/3c58859b8fe72c2dbca6a43bc775aa510ba7e706/t5/data/preprocessors.py#L925
Here please find an example:
For example, a typical example from ReCoRD might look like
{
'passsage': 'This is the passage.',
'query': 'A @placeholder is a bird.',
'entities': ['penguin', 'potato', 'pigeon'],
'answers': ['penguin', 'pigeon'],
}
and I need a prosessor which would turn this example into the following two examples:
{
'inputs': 'record query: A @placeholder is a bird. entities: penguin, '
'potato, pigeon passage: This is the passage.',
'targets': 'penguin',
}
and
{
'inputs': 'record query: A @placeholder is a bird. entities: penguin, '
'potato, pigeon passage: This is the passage.',
'targets': 'pigeon',
}
For doing this, one need unbatch, as each entry can map to multiple samples depending on the number of answers, I am not sure how to perform this operation with huggingface datasets library and greatly appreciate your help
@lhoestq
Thank you very much.
| 2,767 |
https://github.com/huggingface/datasets/issues/2765 | BERTScore Error | [
"Hi,\r\n\r\nThe `use_fast_tokenizer` argument has been recently added to the bert-score lib. I've opened a PR with the fix. In the meantime, you can try to downgrade the version of bert-score with the following command to make the code work:\r\n```\r\npip uninstall bert-score\r\npip install \"bert-score<0.3.10\"\r\n```"
] | ## Describe the bug
A clear and concise description of what the bug is.
## Steps to reproduce the bug
```python
predictions = ["hello there", "general kenobi"]
references = ["hello there", "general kenobi"]
bert = load_metric('bertscore')
bert.compute(predictions=predictions, references=references,lang='en')
```
# Bug
`TypeError: get_hash() missing 1 required positional argument: 'use_fast_tokenizer'`
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version:
- Platform: Colab
- Python version:
- PyArrow version:
| 2,765 |
https://github.com/huggingface/datasets/issues/2763 | English wikipedia datasets is not clean | [
"Hi ! Certain users might need these data (for training or simply to explore/index the dataset).\r\n\r\nFeel free to implement a map function that gets rid of these paragraphs and process the wikipedia dataset with it before training"
] | ## Describe the bug
Wikipedia english dumps contain many wikipedia paragraphs like "References", "Category:" and "See Also" that should not be used for training.
## Steps to reproduce the bug
```python
# Sample code to reproduce the bug
from datasets import load_dataset
w = load_dataset('wikipedia', '20200501.en')
print(w['train'][0]['text'])
```
> 'Yangliuqing () is a market town in Xiqing District, in the western suburbs of Tianjin, People\'s Republic of China. Despite its relatively small size, it has been named since 2006 in the "famous historical and cultural market towns in China".\n\nIt is best known in China for creating nianhua or Yangliuqing nianhua. For more than 400 years, Yangliuqing has in effect specialised in the creation of these woodcuts for the New Year. wood block prints using vivid colourschemes to portray traditional scenes of children\'s games often interwoven with auspiciouse objects.\n\n, it had 27 residential communities () and 25 villages under its administration.\n\nShi Family Grand Courtyard\n\nShi Family Grand Courtyard (Tiānjīn Shí Jiā Dà Yuàn, 天津石家大院) is situated in Yangliuqing Town of Xiqing District, which is the former residence of wealthy merchant Shi Yuanshi - the 4th son of Shi Wancheng, one of the eight great masters in Tianjin. First built in 1875, it covers over 6,000 square meters, including large and small yards and over 200 folk houses, a theater and over 275 rooms that served as apartments and places of business and worship for this powerful family. Shifu Garden, which finished its expansion in October 2003, covers 1,200 square meters, incorporates the elegance of imperial garden and delicacy of south garden. Now the courtyard of Shi family covers about 10,000 square meters, which is called the first mansion in North China. Now it serves as the folk custom museum in Yangliuqing, which has a large collection of folk custom museum in Yanliuqing, which has a large collection of folk art pieces like Yanliuqing New Year pictures, brick sculpture.\n\nShi\'s ancestor came from Dong\'e County in Shandong Province, engaged in water transport of grain. As the wealth gradually accumulated, the Shi Family moved to Yangliuqing and bought large tracts of land and set up their residence. Shi Yuanshi came from the fourth generation of the family, who was a successful businessman and a good household manager, and the residence was thus enlarged for several times until it acquired the present scale. It is believed to be the first mansion in the west of Tianjin.\n\nThe residence is symmetric based on the axis formed by a passageway in the middle, on which there are four archways. On the east side of the courtyard, there are traditional single-story houses with rows of rooms around the four sides, which was once the living area for the Shi Family. The rooms on north side were the accountants\' office. On the west are the major constructions including the family hall for worshipping Buddha, theater and the south reception room. On both sides of the residence are side yard rooms for maids and servants.\n\nToday, the Shi mansion, located in the township of Yangliuqing to the west of central Tianjin, stands as a surprisingly well-preserved monument to China\'s pre-revolution mercantile spirit. It also serves as an on-location shoot for many of China\'s popular historical dramas. Many of the rooms feature period furniture, paintings and calligraphy, and the extensive Shifu Garden.\n\nPart of the complex has been turned into the Yangliuqing Museum, which includes displays focused on symbolic aspects of the courtyards\' construction, local folk art and customs, and traditional period furnishings and crafts.\n\n**See also \n\nList of township-level divisions of Tianjin\n\nReferences \n\n http://arts.cultural-china.com/en/65Arts4795.html\n\nCategory:Towns in Tianjin'**
## Expected results
I expect no junk in the data.
## Actual results
Specify the actual results or traceback.
## Environment info
- `datasets` version: 1.10.2
- Platform: macOS-10.15.7-x86_64-i386-64bit
- Python version: 3.8.5
- PyArrow version: 3.0.0
| 2,763 |
https://github.com/huggingface/datasets/issues/2762 | Add RVL-CDIP dataset | [
"cc @nateraw ",
"#self-assign",
"[labels_only.tar.gz](https://docs.google.com/uc?authuser=0&id=0B0NKIRwUL9KYcXo3bV9LU0t3SGs&export=download) on the RVL-CDIP website does not work for me.\r\n\r\n> 404. That’s an error. The requested URL was not found on this server.\r\n\r\nI contacted the author ( Adam Harley) regarding this, and he told me that the link works for him. Not sure what the issue is. But Adam shared the file (labels_only.tar.gz) with me as an attachment.\r\n\r\nAre we allowed to host this file(labels_only.tar.gz) elsewhere and use that link instead ?\r\n\r\nThank you.\r\n"
] | ## Adding a Dataset
- **Name:** RVL-CDIP
- **Description:** The RVL-CDIP (Ryerson Vision Lab Complex Document Information Processing) dataset consists of 400,000 grayscale images in 16 classes, with 25,000 images per class. There are 320,000 training images, 40,000 validation images, and 40,000 test images. The images are sized so their largest dimension does not exceed 1000 pixels.
- **Paper:** https://www.cs.cmu.edu/~aharley/icdar15/
- **Data:** https://www.cs.cmu.edu/~aharley/rvl-cdip/
- **Motivation:** I'm currently adding LayoutLMv2 and LayoutXLM to HuggingFace Transformers. LayoutLM (v1) already exists in the library. This dataset has a large value for document image classification (i.e. classifying scanned documents). LayoutLM models obtain SOTA on this dataset, so would be great to directly use it in notebooks.
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| 2,762 |
https://github.com/huggingface/datasets/issues/2761 | Error loading C4 realnewslike dataset | [
"Hi @danshirron, \r\n`c4` was updated few days back by @lhoestq. The new configs are `['en', 'en.noclean', 'en.realnewslike', 'en.webtextlike'].` You'll need to remove any older version of this dataset you previously downloaded and then run `load_dataset` again with new configuration.",
"@bhavitvyamalik @lhoestq , just tried the above and got:\r\n>>> a=datasets.load_dataset('c4','en.realnewslike')\r\nDownloading: 3.29kB [00:00, 1.66MB/s] \r\nDownloading: 2.40MB [00:00, 12.6MB/s] \r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/home/dshirron/.local/lib/python3.8/site-packages/datasets/load.py\", line 819, in load_dataset\r\n builder_instance = load_dataset_builder(\r\n File \"/home/dshirron/.local/lib/python3.8/site-packages/datasets/load.py\", line 701, in load_dataset_builder\r\n builder_instance: DatasetBuilder = builder_cls(\r\n File \"/home/dshirron/.local/lib/python3.8/site-packages/datasets/builder.py\", line 1049, in __init__\r\n super(GeneratorBasedBuilder, self).__init__(*args, **kwargs)\r\n File \"/home/dshirron/.local/lib/python3.8/site-packages/datasets/builder.py\", line 268, in __init__\r\n self.config, self.config_id = self._create_builder_config(\r\n File \"/home/dshirron/.local/lib/python3.8/site-packages/datasets/builder.py\", line 360, in _create_builder_config\r\n raise ValueError(\r\nValueError: BuilderConfig en.realnewslike not found. Available: ['en', 'realnewslike', 'en.noblocklist', 'en.noclean']\r\n>>> \r\n\r\ndatasets version is 1.11.0\r\n",
"I think I had an older version of datasets installed and that's why I commented the old configurations in my last comment, my bad! I re-checked and updated it to latest version (`datasets==1.11.0`) and it's showing `available configs: ['en', 'realnewslike', 'en.noblocklist', 'en.noclean']`. \r\n\r\nI tried `raw_datasets = load_dataset('c4', 'realnewslike')` and the download started. Make sure you don't have any old copy of this dataset and you download it fresh using the latest version of datasets. Sorry for the mix up!",
"It works. I probably had some issue with the cache. after cleaning it im able to download the dataset. Thanks"
] | ## Describe the bug
Error loading C4 realnewslike dataset. Validation part mismatch
## Steps to reproduce the bug
```python
raw_datasets = load_dataset('c4', 'realnewslike', cache_dir=model_args.cache_dir)
## Expected results
success on data loading
## Actual results
Downloading: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 15.3M/15.3M [00:00<00:00, 28.1MB/s]Traceback (most recent call last):
File "run_mlm_tf.py", line 794, in <module>
main()
File "run_mlm_tf.py", line 425, in main
raw_datasets = load_dataset(data_args.dataset_name, data_args.dataset_config_name, cache_dir=model_args.cache_dir) File "/home/dshirron/.local/lib/python3.8/site-packages/datasets/load.py", line 843, in load_dataset
builder_instance.download_and_prepare(
File "/home/dshirron/.local/lib/python3.8/site-packages/datasets/builder.py", line 608, in download_and_prepare
self._download_and_prepare(
File "/home/dshirron/.local/lib/python3.8/site-packages/datasets/builder.py", line 698, in _download_and_prepare verify_splits(self.info.splits, split_dict) File "/home/dshirron/.local/lib/python3.8/site-packages/datasets/utils/info_utils.py", line 74, in verify_splits
raise NonMatchingSplitsSizesError(str(bad_splits))
datasets.utils.info_utils.NonMatchingSplitsSizesError: [{'expected': SplitInfo(name='validation', num_bytes=38165657946, num_examples=13799838, dataset_name='c4'), 'recorded': SplitInfo(name='validation', num_bytes=37875873, num_examples=13863, dataset_name='c4')}]
## Environment info
- `datasets` version: 1.10.2
- Platform: Linux-5.4.0-58-generic-x86_64-with-glibc2.29
- Python version: 3.8.10
- PyArrow version: 4.0.1 | 2,761 |
https://github.com/huggingface/datasets/issues/2760 | Add Nuswide dataset | [] | ## Adding a Dataset
- **Name:** *NUSWIDE*
- **Description:** *[A Real-World Web Image Dataset from National University of Singapore](https://lms.comp.nus.edu.sg/wp-content/uploads/2019/research/nuswide/NUS-WIDE.html)*
- **Paper:** *[here](https://lms.comp.nus.edu.sg/wp-content/uploads/2019/research/nuswide/nuswide-civr2009.pdf)*
- **Data:** *[here](https://github.com/wenting-zhao/nuswide)*
- **Motivation:** *This dataset is a benchmark in the Text Retrieval task.*
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| 2,760 |
https://github.com/huggingface/datasets/issues/2757 | Unexpected type after `concatenate_datasets` | [
"Hi @JulesBelveze, thanks for your question.\r\n\r\nNote that 🤗 `datasets` internally store their data in Apache Arrow format.\r\n\r\nHowever, when accessing dataset columns, by default they are returned as native Python objects (lists in this case).\r\n\r\nIf you would like their columns to be returned in a more suitable format for your use case (torch arrays), you can use the method `set_format()`:\r\n```python\r\nconcat_dataset.set_format(type=\"torch\")\r\n```\r\n\r\nYou have detailed information in our docs:\r\n- [Using a Dataset with PyTorch/Tensorflow](https://huggingface.co/docs/datasets/torch_tensorflow.html)\r\n- [Dataset.set_format()](https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.set_format)",
"Thanks @albertvillanova it indeed did the job 😃 \r\nThanks for your answer!"
] | ## Describe the bug
I am trying to concatenate two `Dataset` using `concatenate_datasets` but it turns out that after concatenation the features are casted from `torch.Tensor` to `list`.
It then leads to a weird tensors when trying to convert it to a `DataLoader`. However, if I use each `Dataset` separately everything behave as expected.
## Steps to reproduce the bug
```python
>>> featurized_teacher
Dataset({
features: ['t_labels', 't_input_ids', 't_token_type_ids', 't_attention_mask'],
num_rows: 502
})
>>> for f in featurized_teacher.features:
print(featurized_teacher[f].shape)
torch.Size([502])
torch.Size([502, 300])
torch.Size([502, 300])
torch.Size([502, 300])
>>> featurized_student
Dataset({
features: ['s_features', 's_labels'],
num_rows: 502
})
>>> for f in featurized_student.features:
print(featurized_student[f].shape)
torch.Size([502, 64])
torch.Size([502])
```
The shapes seem alright to me. Then the results after concatenation are as follow:
```python
>>> concat_dataset = datasets.concatenate_datasets([featurized_student, featurized_teacher], axis=1)
>>> type(concat_dataset["t_labels"])
<class 'list'>
```
One would expect to obtain the same type as the one before concatenation.
Am I doing something wrong here? Any idea on how to fix this unexpected behavior?
## Environment info
- `datasets` version: 1.9.0
- Platform: macOS-10.14.6-x86_64-i386-64bit
- Python version: 3.9.5
- PyArrow version: 3.0.0
| 2,757 |
https://github.com/huggingface/datasets/issues/2750 | Second concatenation of datasets produces errors | [
"@albertvillanova ",
"Hi @Aktsvigun, thanks for reporting.\r\n\r\nI'm investigating this.",
"Hi @albertvillanova ,\r\nany update on this? Can I probably help in some way?",
"Hi @Aktsvigun! We are planning to address this issue before our next release, in a couple of weeks at most. 😅 \r\n\r\nIn the meantime, if you would like to contribute, feel free to open a Pull Request. You are welcome. Here you can find more information: [How to contribute to Datasets?](CONTRIBUTING.md)",
"I can't reproduce the bug on master. I believe this issue was fixed by https://github.com/huggingface/datasets/pull/3551."
] | Hi,
I am need to concatenate my dataset with others several times, and after I concatenate it for the second time, the features of features (e.g. tags names) are collapsed. This hinders, for instance, the usage of tokenize function with `data.map`.
```
from datasets import load_dataset, concatenate_datasets
data = load_dataset('trec')['train']
concatenated = concatenate_datasets([data, data])
concatenated_2 = concatenate_datasets([concatenated, concatenated])
print('True features of features:', concatenated.features)
print('\nProduced features of features:', concatenated_2.features)
```
outputs
```
True features of features: {'label-coarse': ClassLabel(num_classes=6, names=['DESC', 'ENTY', 'ABBR', 'HUM', 'NUM', 'LOC'], names_file=None, id=None), 'label-fine': ClassLabel(num_classes=47, names=['manner', 'cremat', 'animal', 'exp', 'ind', 'gr', 'title', 'def', 'date', 'reason', 'event', 'state', 'desc', 'count', 'other', 'letter', 'religion', 'food', 'country', 'color', 'termeq', 'city', 'body', 'dismed', 'mount', 'money', 'product', 'period', 'substance', 'sport', 'plant', 'techmeth', 'volsize', 'instru', 'abb', 'speed', 'word', 'lang', 'perc', 'code', 'dist', 'temp', 'symbol', 'ord', 'veh', 'weight', 'currency'], names_file=None, id=None), 'text': Value(dtype='string', id=None)}
Produced features of features: {'label-coarse': Value(dtype='int64', id=None), 'label-fine': Value(dtype='int64', id=None), 'text': Value(dtype='string', id=None)}
```
I am using `datasets` v.1.11.0 | 2,750 |
https://github.com/huggingface/datasets/issues/2749 | Raise a proper exception when trying to stream a dataset that requires to manually download files | [
"Hi @severo, thanks for reporting.\r\n\r\nAs discussed, datasets requiring manual download should be:\r\n- programmatically identifiable\r\n- properly handled with more clear error message when trying to load them with streaming\r\n\r\nIn relation with programmatically identifiability, note that for datasets requiring manual download, their builder have a property `manual_download_instructions` which is not None:\r\n```python\r\n# Dataset requiring manual download:\r\nbuilder.manual_download_instructions is not None\r\n```",
"Thanks @albertvillanova "
] | ## Describe the bug
At least for 'reclor', 'telugu_books', 'turkish_movie_sentiment', 'ubuntu_dialogs_corpus', 'wikihow', trying to `load_dataset` in streaming mode raises a `TypeError` without any detail about why it fails.
## Steps to reproduce the bug
```python
from datasets import load_dataset
dataset = load_dataset("reclor", streaming=True)
```
## Expected results
Ideally: raise a specific exception, something like `ManualDownloadError`.
Or at least give the reason in the message, as when we load in normal mode:
```python
from datasets import load_dataset
dataset = load_dataset("reclor")
```
```
AssertionError: The dataset reclor with config default requires manual data.
Please follow the manual download instructions: to use ReClor you need to download it manually. Please go to its homepage (http://whyu.me/reclor/) fill the google
form and you will receive a download link and a password to extract it.Please extract all files in one folder and use the path folder in datasets.load_dataset('reclor', data_dir='path/to/folder/folder_name')
.
Manual data can be loaded with `datasets.load_dataset(reclor, data_dir='<path/to/manual/data>')
```
## Actual results
```
TypeError: expected str, bytes or os.PathLike object, not NoneType
```
## Environment info
- `datasets` version: 1.11.0
- Platform: macOS-11.5-x86_64-i386-64bit
- Python version: 3.8.11
- PyArrow version: 4.0.1
| 2,749 |
https://github.com/huggingface/datasets/issues/2746 | Cannot load `few-nerd` dataset | [
"Hi @Mehrad0711,\r\n\r\nI'm afraid there is no \"canonical\" Hugging Face dataset named \"few-nerd\".\r\n\r\nThere are 2 kinds of datasets hosted at the Hugging Face Hub:\r\n- canonical datasets (their identifier contains no slash \"/\"): we, the Hugging Face team, supervise their implementation and we make sure they work correctly by means of our test suite\r\n- community datasets (their identifier contains a slash \"/\", where before the slash it is the username or the organization name): those datasets are uploaded to the Hub by the community, and we, the Hugging Face team, do not supervise them; it is the responsibility of the user/organization implementing them properly if they want them to be used by other users.\r\n\r\nIn this specific case, there is no \"canonical\" dataset named \"few-nerd\". On the other hand, there are two \"community\" datasets named \"few-nerd\":\r\n- [\"nbroad/few-nerd\"](https://huggingface.co/datasets/nbroad/few-nerd)\r\n- [\"dfki-nlp/few-nerd\"](https://huggingface.co/datasets/dfki-nlp/few-nerd)\r\n\r\nIf they were properly implemented, you should be able to load them this way:\r\n```python\r\n# \"nbroad/few-nerd\" community dataset\r\nds = load_dataset(\"nbroad/few-nerd\", \"supervised\")\r\n\r\n# \"dfki-nlp/few-nerd\" community dataset\r\nds = load_dataset(\"dfki-nlp/few-nerd\", \"supervised\")\r\n```\r\n\r\nHowever, they are not correctly implemented and both of them give errors:\r\n- \"nbroad/few-nerd\":\r\n ```\r\n TypeError: expected str, bytes or os.PathLike object, not dict\r\n ```\r\n- \"dfki-nlp/few-nerd\":\r\n ```\r\n ConnectionError: Couldn't reach https://cloud.tsinghua.edu.cn/f/09265750ae6340429827/?dl=1\r\n ```\r\n\r\nYou could try to contact their users/organizations to inform them about their bugs and ask them if they are planning to fix them. Alternatively you could try to implement your own script for this dataset.",
"Thanks @albertvillanova for your detailed explanation! I will resort to my own scripts for now. ",
"Hello, @Mehrad0711; Hi, @albertvillanova !\r\nI am the maintainer of the `dfki/few-nerd\" dataset script, sorry for the very late reply and hope this message finds you well!\r\nWe should use\r\n```\r\ndataset = load_dataset(\"dfki-nlp/few-nerd\", name=\"supervised\")\r\n```\r\ninstead of not specifying the \"name\" argument, where name is from `[\"supervised\", \"inter\", \"intra\"]`. Otherwise the method just treats \"supervised\" as `split`, which we reserve after specifying the name, since for each name, there are three splits: train, dev and test.\r\n\r\nAlso we use Tsinghua server source to download data files since it is the official source referred in the paper where the dataset is released (even though it is cc-by-sa-4.0 licensed, means we can copy the data anywhere after mentioning the license\r\n). Sometimes the server just runs down due to high pressure, kinda weird (we encountered the same server problem serveral times a month when we conducted experiments on Few-NERD XD). I tried the script just now and it works perfectly!\r\n```\r\n>> dataset\r\nDatasetDict({\r\n train: Dataset({\r\n features: ['id', 'tokens', 'ner_tags', 'fine_ner_tags'],\r\n num_rows: 131767\r\n })\r\n validation: Dataset({\r\n features: ['id', 'tokens', 'ner_tags', 'fine_ner_tags'],\r\n num_rows: 18824\r\n })\r\n test: Dataset({\r\n features: ['id', 'tokens', 'ner_tags', 'fine_ner_tags'],\r\n num_rows: 37648\r\n })\r\n})\r\n>>> dataset[\"train\"]\r\nDataset({\r\n features: ['id', 'tokens', 'ner_tags', 'fine_ner_tags'],\r\n num_rows: 131767\r\n})\r\n>>> dataset[\"train\"][0]\r\n{'id': '0', 'tokens': ['Paul', 'International', 'airport', '.'], 'ner_tags': [0, 0, 0, 0], 'fine_ner_tags': [0, 0, 0, 0]}\r\n```\r\nAnyways if you cannot stand the pain with the server and its slow download speed, you can also download the `dfki/few-nerd.py` script from HF and change the `_URLs` to your personal drive (after you once successfully download the data and upload to your cloud drive), and then load the .py script locally.\r\n\r\nHope this reply can still be any help. If you still have problems with it, feel free to ask here and I am glad to help!\r\nBest wishes.",
"Hi @chen-yuxuan, thanks for your answer.\r\n\r\nJust a few comments:\r\n\r\n- Please, note that as we use `datasets.load_dataset` implementation, we can pass the configuration name as the second positional argument (no need to pass explicitly `name=`) and it downloads the 3 splits:\r\n```python\r\n In [4]: ds = load_dataset(\"dfki-nlp/few-nerd\", \"supervised\")\r\nDownloading: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 11.5k/11.5k [00:00<00:00, 2.85MB/s]\r\nDownloading and preparing dataset few_nerd/supervised to .cache\\huggingface\\datasets\\dfki-nlp___few_nerd\\supervised\\0.0.0\\e40882b71f037a4a1f232025899170fbe8113cd2f4a26dddd2add7222a077255...\r\nDownloading: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 14.6M/14.6M [01:16<00:00, 190kB/s]\r\nDownloading: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 11.9M/11.9M [01:14<00:00, 160kB/s]\r\nDownloading: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 12.0M/12.0M [01:04<00:00, 186kB/s]\r\n100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [03:58<00:00, 79.45s/it]\r\n100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00, 3.11it/s]\r\n```\r\n\r\n- On the other hand, please note that your script does not work on Windows machines, because you call `open()` without passing the encoding parameter:\r\n```\r\n~\\.cache\\huggingface\\modules\\datasets_modules\\datasets\\dfki-nlp___few-nerd\\e40882b71f037a4a1f232025899170fbe8113cd2f4a26dddd2add7222a077255\\few-nerd.py in <genexpr>(.0)\r\n 276 assert filepath[-4:] == \".txt\"\r\n 277\r\n--> 278 num_lines = sum(1 for _ in open(filepath))\r\n 279 id = 0\r\n 280\r\n\r\n.venv\\lib\\encodings\\cp1252.py in decode(self, input, final)\r\n 21 class IncrementalDecoder(codecs.IncrementalDecoder):\r\n 22 def decode(self, input, final=False):\r\n---> 23 return codecs.charmap_decode(input,self.errors,decoding_table)[0]\r\n 24\r\n 25 class StreamWriter(Codec,codecs.StreamWriter):\r\n\r\nUnicodeDecodeError: 'charmap' codec can't decode byte 0x8d in position 5238: character maps to <undefined>\r\n```\r\n\r\nIf you would like your script to be usable on Windows machines, you should pass `encoding=\"utf-8\"` to every `open()` function:\r\n- line 278: `num_lines = sum(1 for _ in open(filepath, encoding=\"utf-8\"))`\r\n- line 281: `with open(filepath, \"r\", encoding=\"utf-8\")`",
"Thank you @albertvillanova for your detailed feedback!\r\n\r\n> no need to pass explicitly `name=`\r\n\r\nGood catch! I thought `split` stands before `name` in the argument list... but now it is all clear to me, sounds cool! Thanks for the explanation.\r\n\r\nAnyways in our old code it still looks bit confusing if we only want one split but the function downloads all, so to allow efficient downloading, I optimized the code a bit so that only the specified split data is downloaded. now we get\r\n```\r\n>>> x = load_dataset(\"dfki-nlp/few-nerd\", \"supervised\")\r\nDownloading and preparing dataset few_nerd/supervised to /home/user/.cache/huggingface/datasets/few_nerd/supervised/0.0.0/8e7ab598946cd5b395dcec6ea239123c8dff5b58b8e1c03b0c595b540248a885...\r\nDownloading: 100%|███████████████████████████████████████████████████████████████████| 14.6M/14.6M [01:01<00:00, 238kB/s]\r\n100%|██████████████████████████████████████████████████████████████████████| 3359329/3359329 [00:12<00:00, 275462.84it/s]\r\n100%|████████████████████████████████████████████████████████████████████████| 482037/482037 [00:01<00:00, 278633.64it/s]\r\n100%|████████████████████████████████████████████████████████████████████████| 958765/958765 [00:03<00:00, 267472.83it/s]\r\nDataset few_nerd downloaded and prepared to /home/user/.cache/huggingface/datasets/few_nerd/supervised/0.0.0/8e7ab598946cd5b395dcec6ea239123c8dff5b58b8e1c03b0c595b540248a885. Subsequent calls will reuse this data.\r\n```\r\nwhere only one progress bar indicates downloading, and the three others just indicate pre-processing for the train, dev, test set.\r\n\r\nFor the encoding issue, I have made corresponding changes for the two lines you pointed out. However, I have no windows machine at hand, I would really appreciate it if you could help test on your end.\r\n\r\nAll the updates are uploaded to HF under `dfki-nlp` account where I am working for. \r\nThank you again for your kind help!\r\n",
"Hi @chen-yuxuan,\r\n\r\nI have tested on Windows and now it works perfectly, after the fixing of the encoding issue:\r\n```python\r\nIn [1]: from datasets import load_dataset\r\n\r\nIn [2]: ds = load_dataset(\"dfki-nlp/few-nerd\", \"supervised\")\r\nDownloading: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 11.5k/11.5k [00:00<?, ?B/s]\r\nDownloading and preparing dataset few_nerd/supervised to C:\\Users\\username\\.cache\\huggingface\\datasets\\dfki-nlp___few_nerd\\supervised\\0.0.0\\e1ceeaee82073fea12206e4461c7cfcd67e68c8f3ebeca179bddcacee00c4511...\r\n100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3359329/3359329 [00:25<00:00, 129427.23it/s]\r\n100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 482037/482037 [00:03<00:00, 134513.66it/s]\r\n100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 958765/958765 [00:06<00:00, 143152.35it/s]\r\nDataset few_nerd downloaded and prepared to C:\\Users\\username\\.cache\\huggingface\\datasets\\dfki-nlp___few_nerd\\supervised\\0.0.0\\e1ceeaee82073fea12206e4461c7cfcd67e68c8f3ebeca179bddcacee00c4511. Subsequent calls will reuse this data.765 [00:06<00:00, 139045.03it/s]\r\n100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00, 174.71it/s]\r\n\r\nIn [3]: ds\r\nOut[3]:\r\nDatasetDict({\r\n train: Dataset({\r\n features: ['id', 'tokens', 'ner_tags', 'fine_ner_tags'],\r\n num_rows: 131767\r\n })\r\n validation: Dataset({\r\n features: ['id', 'tokens', 'ner_tags', 'fine_ner_tags'],\r\n num_rows: 18824\r\n })\r\n test: Dataset({\r\n features: ['id', 'tokens', 'ner_tags', 'fine_ner_tags'],\r\n num_rows: 37648\r\n })\r\n})\r\n```"
] | ## Describe the bug
Cannot load `few-nerd` dataset.
## Steps to reproduce the bug
```python
from datasets import load_dataset
load_dataset('few-nerd', 'supervised')
```
## Actual results
Executing above code will give the following error:
```
Using the latest cached version of the module from /Users/Mehrad/.cache/huggingface/modules/datasets_modules/datasets/few-nerd/62464ace912a40a0f33a11a8310f9041c9dc3590ff2b3c77c14d83ca53cfec53 (last modified on Wed Jun 2 11:34:25 2021) since it couldn't be found locally at /Users/Mehrad/Documents/GitHub/genienlp/few-nerd/few-nerd.py, or remotely (FileNotFoundError).
Downloading and preparing dataset few_nerd/supervised (download: Unknown size, generated: Unknown size, post-processed: Unknown size, total: Unknown size) to /Users/Mehrad/.cache/huggingface/datasets/few_nerd/supervised/0.0.0/62464ace912a40a0f33a11a8310f9041c9dc3590ff2b3c77c14d83ca53cfec53...
Traceback (most recent call last):
File "/Users/Mehrad/opt/anaconda3/lib/python3.7/site-packages/datasets/builder.py", line 693, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/Users/Mehrad/opt/anaconda3/lib/python3.7/site-packages/datasets/builder.py", line 1107, in _prepare_split
disable=bool(logging.get_verbosity() == logging.NOTSET),
File "/Users/Mehrad/opt/anaconda3/lib/python3.7/site-packages/tqdm/std.py", line 1133, in __iter__
for obj in iterable:
File "/Users/Mehrad/.cache/huggingface/modules/datasets_modules/datasets/few-nerd/62464ace912a40a0f33a11a8310f9041c9dc3590ff2b3c77c14d83ca53cfec53/few-nerd.py", line 196, in _generate_examples
with open(filepath, encoding="utf-8") as f:
FileNotFoundError: [Errno 2] No such file or directory: '/Users/Mehrad/.cache/huggingface/datasets/downloads/supervised/train.json'
```
The bug is probably in identifying and downloading the dataset. If I download the json splits directly from [link](https://github.com/nbroad1881/few-nerd/tree/main/uncompressed) and put them under the downloads directory, they will be processed into arrow format correctly.
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.11.0
- Python version: 3.8
- PyArrow version: 1.0.1
| 2,746 |
https://github.com/huggingface/datasets/issues/2743 | Dataset JSON is incorrect | [
"As discussed, the metadata JSON files must be regenerated because the keys were nor properly generated and they will not be read by the builder:\r\n> Indeed there is some problem/bug while reading the datasets_info.json file: there is a mismatch with the config.name keys in the file...\r\nIn the meanwhile, in order to be able to use the datasets_info.json file content, you can create the builder without passing the name :\r\n```\r\nIn [25]: builder = datasets.load_dataset_builder(\"journalists_questions\")\r\nIn [26]: builder.info.splits\r\nOut[26]: {'train': SplitInfo(name='train', num_bytes=342296, num_examples=10077, dataset_name='journalists_questions')}\r\n```\r\n\r\nAfter regenerating the metadata JSON file for this dataset, I get the right key:\r\n```\r\n{\"plain_text\": {\"description\": \"The journalists_questions corpus (\r\n```",
"Thanks!"
] | ## Describe the bug
The JSON file generated for https://github.com/huggingface/datasets/blob/573f3d35081cee239d1b962878206e9abe6cde91/datasets/journalists_questions/journalists_questions.py is https://github.com/huggingface/datasets/blob/573f3d35081cee239d1b962878206e9abe6cde91/datasets/journalists_questions/dataset_infos.json.
The only config should be `plain_text`, but the first key in the JSON is `journalists_questions` (the dataset id) instead.
```json
{
"journalists_questions": {
"description": "The journalists_questions corpus (version 1.0) is a collection of 10K human-written Arabic\ntweets manually labeled for question identification over Arabic tweets posted by journalists.\n",
...
```
## Steps to reproduce the bug
Look at the files.
## Expected results
The first key should be `plain_text`:
```json
{
"plain_text": {
"description": "The journalists_questions corpus (version 1.0) is a collection of 10K human-written Arabic\ntweets manually labeled for question identification over Arabic tweets posted by journalists.\n",
...
```
## Actual results
```json
{
"journalists_questions": {
"description": "The journalists_questions corpus (version 1.0) is a collection of 10K human-written Arabic\ntweets manually labeled for question identification over Arabic tweets posted by journalists.\n",
...
```
| 2,743 |
https://github.com/huggingface/datasets/issues/2742 | Improve detection of streamable file types | [
"maybe we should rather attempt to download a `Range` from the server and see if it works?"
] | **Is your feature request related to a problem? Please describe.**
```python
from datasets import load_dataset_builder
from datasets.utils.streaming_download_manager import StreamingDownloadManager
builder = load_dataset_builder("journalists_questions", name="plain_text")
builder._split_generators(StreamingDownloadManager(base_path=builder.base_path))
```
raises
```
NotImplementedError: Extraction protocol for file at https://drive.google.com/uc?export=download&id=1CBrh-9OrSpKmPQBxTK_ji6mq6WTN_U9U is not implemented yet
```
But the file at https://drive.google.com/uc?export=download&id=1CBrh-9OrSpKmPQBxTK_ji6mq6WTN_U9U is a text file and it can be streamed:
```bash
curl --header "Range: bytes=0-100" -L https://drive.google.com/uc\?export\=download\&id\=1CBrh-9OrSpKmPQBxTK_ji6mq6WTN_U9U
506938088174940160 yes 1
302221719412830209 yes 1
289761704907268096 yes 1
513820885032378369 yes %
```
Yet, it's wrongly categorized as a file type that cannot be streamed because the test is currently based on 1. the presence of a file extension at the end of the URL (here: no extension), and 2. the inclusion of this extension in a list of supported formats.
**Describe the solution you'd like**
In the case of an URL (instead of a local path), ask for the MIME type, and decide on that value? Note that it would not work in that case, because the value of `content_type` is `text/html; charset=UTF-8`.
**Describe alternatives you've considered**
Add a variable in the dataset script to set the data format by hand.
| 2,742 |
https://github.com/huggingface/datasets/issues/2741 | Add Hypersim dataset | [] | ## Adding a Dataset
- **Name:** Hypersim
- **Description:** photorealistic synthetic dataset for holistic indoor scene understanding
- **Paper:** *link to the dataset paper if available*
- **Data:** https://github.com/apple/ml-hypersim
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| 2,741 |
https://github.com/huggingface/datasets/issues/2737 | SacreBLEU update | [
"Hi @devrimcavusoglu, \r\nI tried your code with latest version of `datasets`and `sacrebleu==1.5.1` and it's running fine after changing one small thing:\r\n```\r\nsacrebleu = datasets.load_metric('sacrebleu')\r\npredictions = [\"It is a guide to action which ensures that the military always obeys the commands of the party\"]\r\nreferences = [[\"It is a guide to action that ensures that the military will forever heed Party commands\"]] # double brackets here should do the work\r\nresults = sacrebleu.compute(predictions=predictions, references=references)\r\nprint(results)\r\noutput: {'score': 41.180376356915765, 'counts': [11, 8, 6, 4], 'totals': [18, 17, 16, 15], 'precisions': [61.111111111111114, 47.05882352941177, 37.5, 26.666666666666668], 'bp': 1.0, 'sys_len': 18, 'ref_len': 16}\r\n```",
"@bhavitvyamalik hmm. I forgot double brackets, but still didn't work when used it with double brackets. It may be an isseu with platform (using win-10 currently), or versions. What is your platform and your version info for datasets, python, and sacrebleu ?",
"You can check that here, I've reproduced your code in [Google colab](https://colab.research.google.com/drive/1X90fHRgMLKczOVgVk7NDEw_ciZFDjaCM?usp=sharing). Looks like there was some issue in `sacrebleu` which was fixed later from what I've found [here](https://github.com/pytorch/fairseq/issues/2049#issuecomment-622367967). Upgrading `sacrebleu` to latest version should work.",
"It seems that next release of `sacrebleu` (v2.0.0) will break our `datasets` implementation to compute it. See my Google Colab: https://colab.research.google.com/drive/1SKmvvjQi6k_3OHsX5NPkZdiaJIfXyv9X?usp=sharing\r\n\r\nI'm reopening this Issue and making a Pull Request to fix it.",
"> It seems that next release of `sacrebleu` (v2.0.0) will break our `datasets` implementation to compute it. See my Google Colab: https://colab.research.google.com/drive/1SKmvvjQi6k_3OHsX5NPkZdiaJIfXyv9X?usp=sharing\r\n> \r\n> I'm reopening this Issue and making a Pull Request to fix it.\r\n\r\nHow did you solve him"
] | With the latest release of [sacrebleu](https://github.com/mjpost/sacrebleu), `datasets.metrics.sacrebleu` is broken, and getting error.
AttributeError: module 'sacrebleu' has no attribute 'DEFAULT_TOKENIZER'
this happens since in new version of sacrebleu there is no `DEFAULT_TOKENIZER`, but sacrebleu.py tries to import it anyways. This can be fixed currently with fixing `sacrebleu==1.5.0`
## Steps to reproduce the bug
```python
sacrebleu= datasets.load_metric('sacrebleu')
predictions = ["It is a guide to action which ensures that the military always obeys the commands of the party"]
references = ["It is a guide to action that ensures that the military will forever heed Party commands"]
results = sacrebleu.compute(predictions=predictions, references=references)
print(results)
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.11.0
- Platform: Windows-10-10.0.19041-SP0
- Python version: Python 3.8.0
- PyArrow version: 5.0.0
| 2,737 |
https://github.com/huggingface/datasets/issues/2736 | Add Microsoft Building Footprints dataset | [
"Motivation: this can be a useful dataset for researchers working on climate change adaptation, urban studies, geography, etc. I'll see if I can figure out how to add it!"
] | ## Adding a Dataset
- **Name:** Microsoft Building Footprints
- **Description:** With the goal to increase the coverage of building footprint data available as open data for OpenStreetMap and humanitarian efforts, we have released millions of building footprints as open data available to download free of charge.
- **Paper:** *link to the dataset paper if available*
- **Data:** https://www.microsoft.com/en-us/maps/building-footprints
- **Motivation:** this can be a useful dataset for researchers working on climate change adaptation, urban studies, geography, etc.
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
Reported by: @sashavor | 2,736 |
https://github.com/huggingface/datasets/issues/2735 | Add Open Buildings dataset | [] | ## Adding a Dataset
- **Name:** Open Buildings
- **Description:** A dataset of building footprints to support social good applications.
Building footprints are useful for a range of important applications, from population estimation, urban planning and humanitarian response, to environmental and climate science. This large-scale open dataset contains the outlines of buildings derived from high-resolution satellite imagery in order to support these types of uses. The project being based in Ghana, the current focus is on the continent of Africa.
See: "Mapping Africa's Buildings with Satellite Imagery" https://ai.googleblog.com/2021/07/mapping-africas-buildings-with.html
- **Paper:** https://arxiv.org/abs/2107.12283
- **Data:** https://sites.research.google/open-buildings/
- **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).
Reported by: @osanseviero | 2,735 |
https://github.com/huggingface/datasets/issues/2730 | Update CommonVoice with new release | [
"cc @patrickvonplaten?",
"Does anybody know if there is a bundled link, which would allow direct data download instead of manual? \r\nSomething similar to: `https://voice-prod-bundler-ee1969a6ce8178826482b88e843c335139bd3fb4.s3.amazonaws.com/cv-corpus-6.1-2020-12-11/ab.tar.gz` ? cc @patil-suraj \r\n",
"Also see: https://github.com/common-voice/common-voice-bundler/issues/15"
] | ## Adding a Dataset
- **Name:** CommonVoice mid-2021 release
- **Description:** more data in CommonVoice: Languages that have increased the most by percentage are Thai (almost 20x growth, from 12 hours to 250 hours), Luganda (almost 9x growth, from 8 to 80), Esperanto (7x growth, from 100 to 840), and Tamil (almost 8x, from 24 to 220).
- **Paper:** https://discourse.mozilla.org/t/common-voice-2021-mid-year-dataset-release/83812
- **Data:** https://commonvoice.mozilla.org/en/datasets
- **Motivation:** More data and more varied. I think we just need to add configs in the existing dataset script.
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| 2,730 |
https://github.com/huggingface/datasets/issues/2728 | Concurrent use of same dataset (already downloaded) | [
"Launching simultaneous job relying on the same datasets try some writing issue. I guess it is unexpected since I only need to load some already downloaded file.",
"If i have two jobs that use the same dataset. I got :\r\n\r\n\r\n File \"compute_measures.py\", line 181, in <module>\r\n train_loader, val_loader, test_loader = get_dataloader(args)\r\n File \"/gpfsdswork/projects/rech/toto/intRAOcular/dataset_utils.py\", line 69, in get_dataloader\r\n dataset_train = load_dataset('paws', \"labeled_final\", split='train', download_mode=\"reuse_cache_if_exists\")\r\n File \"/gpfslocalsup/pub/anaconda-py3/2020.02/envs/pytorch-gpu-1.7.0/lib/python3.7/site-packages/datasets/load.py\", line 748, in load_dataset\r\n use_auth_token=use_auth_token,\r\n File \"/gpfslocalsup/pub/anaconda-py3/2020.02/envs/pytorch-gpu-1.7.0/lib/python3.7/site-packages/datasets/builder.py\", line 582, in download_and_prepare\r\n self._save_info()\r\n File \"/gpfslocalsup/pub/anaconda-py3/2020.02/envs/pytorch-gpu-1.7.0/lib/python3.7/site-packages/datasets/builder.py\", line 690, in _save_info\r\n self.info.write_to_directory(self._cache_dir)\r\n File \"/gpfslocalsup/pub/anaconda-py3/2020.02/envs/pytorch-gpu-1.7.0/lib/python3.7/site-packages/datasets/info.py\", line 195, in write_to_directory\r\n with open(os.path.join(dataset_info_dir, config.LICENSE_FILENAME), \"wb\") as f:\r\nFileNotFoundError: [Errno 2] No such file or directory: '/gpfswork/rech/toto/datasets/paws/labeled_final/1.1.0/09d8fae989bb569009a8f5b879ccf2924d3e5cd55bfe2e89e6dab1c0b50ecd34.incomplete/LICENSE'",
"You can probably have a solution much faster than me (first time I use the library). But I suspect some write function are used when loading the dataset from cache.",
"I have the same issue:\r\n```\r\nTraceback (most recent call last):\r\n File \"/dccstor/tslm/envs/anaconda3/envs/trf-a100/lib/python3.9/site-packages/datasets/builder.py\", line 652, in _download_and_prepare\r\n self._prepare_split(split_generator, **prepare_split_kwargs)\r\n File \"/dccstor/tslm/envs/anaconda3/envs/trf-a100/lib/python3.9/site-packages/datasets/builder.py\", line 1040, in _prepare_split\r\n with ArrowWriter(features=self.info.features, path=fpath) as writer:\r\n File \"/dccstor/tslm/envs/anaconda3/envs/trf-a100/lib/python3.9/site-packages/datasets/arrow_writer.py\", line 192, in __init__\r\n self.stream = pa.OSFile(self._path, \"wb\")\r\n File \"pyarrow/io.pxi\", line 829, in pyarrow.lib.OSFile.__cinit__\r\n File \"pyarrow/io.pxi\", line 844, in pyarrow.lib.OSFile._open_writable\r\n File \"pyarrow/error.pxi\", line 122, in pyarrow.lib.pyarrow_internal_check_status\r\n File \"pyarrow/error.pxi\", line 97, in pyarrow.lib.check_status\r\nFileNotFoundError: [Errno 2] Failed to open local file '/dccstor/tslm-gen/.cache/csv/default-387f1f95c084d4df/0.0.0/2dc6629a9ff6b5697d82c25b73731dd440507a69cbce8b425db50b751e8fcfd0.incomplete/csv-validation.arrow'. Detail: [errno 2] No such file or directory\r\nDuring handling of the above exception, another exception occurred:\r\nTraceback (most recent call last):\r\n File \"/dccstor/tslm/elron/tslm-gen/train.py\", line 510, in <module>\r\n main()\r\n File \"/dccstor/tslm/elron/tslm-gen/train.py\", line 246, in main\r\n datasets = prepare_dataset(dataset_args, logger)\r\n File \"/dccstor/tslm/elron/tslm-gen/data.py\", line 157, in prepare_dataset\r\n datasets = load_dataset(extension, data_files=data_files, split=dataset_split, cache_dir=dataset_args.dataset_cache_dir, na_filter=False, download_mode=dataset_args.dataset_generate_mode)\r\n File \"/dccstor/tslm/envs/anaconda3/envs/trf-a100/lib/python3.9/site-packages/datasets/load.py\", line 742, in load_dataset\r\n builder_instance.download_and_prepare(\r\n File \"/dccstor/tslm/envs/anaconda3/envs/trf-a100/lib/python3.9/site-packages/datasets/builder.py\", line 574, in download_and_prepare\r\n self._download_and_prepare(\r\n File \"/dccstor/tslm/envs/anaconda3/envs/trf-a100/lib/python3.9/site-packages/datasets/builder.py\", line 654, in _download_and_prepare\r\n raise OSError(\r\nOSError: Cannot find data file. \r\nOriginal error:\r\n[Errno 2] Failed to open local file '/dccstor/tslm-gen/.cache/csv/default-387f1f95c084d4df/0.0.0/2dc6629a9ff6b5697d82c25b73731dd440507a69cbce8b425db50b751e8fcfd0.incomplete/csv-validation.arrow'. Detail: [errno 2] No such file or directory\r\n```"
] | ## Describe the bug
When launching several jobs at the same time loading the same dataset trigger some errors see (last comments).
## Steps to reproduce the bug
export HF_DATASETS_CACHE=/gpfswork/rech/toto/datasets
for MODEL in "bert-base-uncased" "roberta-base" "distilbert-base-cased"; do # "bert-base-uncased" "bert-large-cased" "roberta-large" "albert-base-v1" "albert-large-v1"; do
for TASK_NAME in "mrpc" "rte" 'imdb' "paws" "mnli"; do
export OUTPUT_DIR=${MODEL}_${TASK_NAME}
sbatch --job-name=${OUTPUT_DIR} \
--gres=gpu:1 \
--no-requeue \
--cpus-per-task=10 \
--hint=nomultithread \
--time=1:00:00 \
--output=jobinfo/${OUTPUT_DIR}_%j.out \
--error=jobinfo/${OUTPUT_DIR}_%j.err \
--qos=qos_gpu-t4 \
--wrap="module purge; module load pytorch-gpu/py3/1.7.0 ; export HF_DATASETS_OFFLINE=1; export HF_DATASETS_CACHE=/gpfswork/rech/toto/datasets; python compute_measures.py --seed=$SEED --saving_path=results --batch_size=$BATCH_SIZE --task_name=$TASK_NAME --model_name=/gpfswork/rech/toto/transformers_models/$MODEL"
done
done
```python
# Sample code to reproduce the bug
dataset_train = load_dataset('imdb', split='train', download_mode="reuse_cache_if_exists")
dataset_train = dataset_train.map(lambda e: tokenizer(e['text'], truncation=True, padding='max_length'),
batched=True).select(list(range(args.filter)))
dataset_val = load_dataset('imdb', split='train', download_mode="reuse_cache_if_exists")
dataset_val = dataset_val.map(lambda e: tokenizer(e['text'], truncation=True, padding='max_length'),
batched=True).select(list(range(args.filter, args.filter + 5000)))
dataset_test = load_dataset('imdb', split='test', download_mode="reuse_cache_if_exists")
dataset_test = dataset_test.map(lambda e: tokenizer(e['text'], truncation=True, padding='max_length'),
batched=True)
```
## Expected results
I believe I am doing something wrong with the objects.
## Actual results
Traceback (most recent call last):
File "/gpfslocalsup/pub/anaconda-py3/2020.02/envs/pytorch-gpu-1.7.0/lib/python3.7/site-packages/datasets/builder.py", line 652, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/gpfslocalsup/pub/anaconda-py3/2020.02/envs/pytorch-gpu-1.7.0/lib/python3.7/site-packages/datasets/builder.py", line 983, in _prepare_split
check_duplicates=True,
File "/gpfslocalsup/pub/anaconda-py3/2020.02/envs/pytorch-gpu-1.7.0/lib/python3.7/site-packages/datasets/arrow_writer.py", line 192, in __init__
self.stream = pa.OSFile(self._path, "wb")
File "pyarrow/io.pxi", line 829, in pyarrow.lib.OSFile.__cinit__
File "pyarrow/io.pxi", line 844, in pyarrow.lib.OSFile._open_writable
File "pyarrow/error.pxi", line 122, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 97, in pyarrow.lib.check_status
FileNotFoundError: [Errno 2] Failed to open local file '/gpfswork/rech/tts/unm25jp/datasets/paws/labeled_final/1.1.0/09d8fae989bb569009a8f5b879ccf2924d3e5cd55bfe2e89e6dab1c0b50ecd34.incomplete/paws-test.arrow'. Detail: [errno 2] No such file or directory
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "compute_measures.py", line 181, in <module>
train_loader, val_loader, test_loader = get_dataloader(args)
File "/gpfsdswork/projects/rech/toto/intRAOcular/dataset_utils.py", line 69, in get_dataloader
dataset_train = load_dataset('paws', "labeled_final", split='train', download_mode="reuse_cache_if_exists")
File "/gpfslocalsup/pub/anaconda-py3/2020.02/envs/pytorch-gpu-1.7.0/lib/python3.7/site-packages/datasets/load.py", line 748, in load_dataset
use_auth_token=use_auth_token,
File "/gpfslocalsup/pub/anaconda-py3/2020.02/envs/pytorch-gpu-1.7.0/lib/python3.7/site-packages/datasets/builder.py", line 575, in download_and_prepare
dl_manager=dl_manager, verify_infos=verify_infos, **download_and_prepare_kwargs
File "/gpfslocalsup/pub/anaconda-py3/2020.02/envs/pytorch-gpu-1.7.0/lib/python3.7/site-packages/datasets/builder.py", line 658, in _download_and_prepare
+ str(e)
OSError: Cannot find data file.
Original error:
[Errno 2] Failed to open local file '/gpfswork/rech/toto/datasets/paws/labeled_final/1.1.0/09d8fae989bb569009a8f5b879ccf2924d3e5cd55bfe2e89e6dab1c0b50ecd34.incomplete/paws-test.arrow'. Detail: [errno 2] No such file or directory
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: datasets==1.8.0
- Platform: linux (jeanzay)
- Python version: pyarrow==2.0.0
- PyArrow version: 3.7.8
| 2,728 |
https://github.com/huggingface/datasets/issues/2727 | Error in loading the Arabic Billion Words Corpus | [
"I modified the dataset loading script to catch the `IndexError` and inspect the records at which the error is happening, and I found this:\r\nFor the `Techreen` config, the error happens in 36 records when trying to find the `Text` or `Dateline` tags. All these 36 records look something like:\r\n```\r\n<Techreen>\r\n <ID>TRN_ARB_0248167</ID>\r\n <URL>http://tishreen.news.sy/tishreen/public/read/248240</URL>\r\n <Headline>Removed, because the original articles was in English</Headline>\r\n</Techreen>\r\n```\r\n\r\nand all the 288 faulty records in the `Almustaqbal` config look like:\r\n```\r\n<Almustaqbal>\r\n <ID>MTL_ARB_0028398</ID>\r\n \r\n <URL>http://www.almustaqbal.com/v4/article.aspx?type=NP&ArticleID=179015</URL>\r\n <Headline> Removed because it is not available in the original site</Headline>\r\n</Almustaqbal>\r\n```\r\n\r\nso the error is happening because the articles were removed and so the associated records lack the `Text` tag.\r\n\r\nIn this case, I think we just need to catch the `IndexError` and ignore (pass) it.\r\n",
"Thanks @M-Salti for reporting this issue and for your investigation.\r\n\r\nIndeed, those `IndexError` should be catched and the corresponding record should be ignored.\r\n\r\nI'm opening a Pull Request to fix it."
] | ## Describe the bug
I get `IndexError: list index out of range` when trying to load the `Techreen` and `Almustaqbal` configs of the dataset.
## Steps to reproduce the bug
```python
load_dataset("arabic_billion_words", "Techreen")
load_dataset("arabic_billion_words", "Almustaqbal")
```
## Expected results
The datasets load succefully.
## Actual results
```python
_extract_tags(self, sample, tag)
139 if len(out) > 0:
140 break
--> 141 return out[0]
142
143 def _clean_text(self, text):
IndexError: list index out of range
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.10.2
- Platform: Ubuntu 18.04.5 LTS
- Python version: 3.7.11
- PyArrow version: 3.0.0
| 2,727 |
https://github.com/huggingface/datasets/issues/2724 | 404 Error when loading remote data files from private repo | [
"I guess the issue is when computing the ETags of the remote files. Indeed `use_auth_token` must be passed to `request_etags` here:\r\n\r\nhttps://github.com/huggingface/datasets/blob/35b5e4bc0cb2ed896e40f3eb2a4aa3de1cb1a6c5/src/datasets/builder.py#L160-L160",
"Yes, I remember having properly implemented that: \r\n- https://github.com/huggingface/datasets/commit/7a9c62f7cef9ecc293f629f859d4375a6bd26dc8#diff-f933ce41f71c6c0d1ce658e27de62cbe0b45d777e9e68056dd012ac3eb9324f7R160\r\n- https://github.com/huggingface/datasets/pull/2628/commits/6350a03b4b830339a745f7b1da46ece784ca734c\r\n\r\nBut a subsequent refactoring accidentally removed it...",
"I have opened a PR to fix it @lewtun."
] | ## Describe the bug
When loading remote data files from a private repo, a 404 error is raised.
## Steps to reproduce the bug
```python
url = hf_hub_url("lewtun/asr-preds-test", "preds.jsonl", repo_type="dataset")
dset = load_dataset("json", data_files=url, use_auth_token=True)
# HTTPError: 404 Client Error: Not Found for url: https://huggingface.co/datasets/lewtun/asr-preds-test/resolve/main/preds.jsonl
```
## Expected results
Load dataset.
## Actual results
404 Error.
| 2,724 |
https://github.com/huggingface/datasets/issues/2722 | Missing cache file | [
"This could be solved by going to the glue/ directory and delete sst2 directory, then load the dataset again will help you redownload the dataset.",
"Hi ! Not sure why this file was missing, but yes the way to fix this is to delete the sst2 directory and to reload the dataset"
] | Strangely missing cache file after I restart my program again.
`glue_dataset = datasets.load_dataset('glue', 'sst2')`
`FileNotFoundError: [Errno 2] No such file or directory: /Users/chris/.cache/huggingface/datasets/glue/sst2/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96d6053ad/dataset_info.json'`
| 2,722 |
https://github.com/huggingface/datasets/issues/2719 | Use ETag in streaming mode to detect resource updates | [] | **Is your feature request related to a problem? Please describe.**
I want to cache data I generate from processing a dataset I've loaded in streaming mode, but I've currently no way to know if the remote data has been updated or not, thus I don't know when to invalidate my cache.
**Describe the solution you'd like**
Take the ETag of the data files into account and provide it (directly or through a hash) to give a signal that I can invalidate my cache.
**Describe alternatives you've considered**
None
| 2,719 |
https://github.com/huggingface/datasets/issues/2716 | Calling shuffle on IterableDataset will disable batching in case any functions were mapped | [
"Hi :) Good catch ! Feel free to open a PR if you want to contribute, this would be very welcome ;)",
"Have raised the PR [here](https://github.com/huggingface/datasets/pull/2717)",
"Fixed by #2717."
] | When using dataset in streaming mode, if one applies `shuffle` method on the dataset and `map` method for which `batched=True` than the batching operation will not happen, instead `batched` will be set to `False`
I did RCA on the dataset codebase, the problem is emerging from [this line of code](https://github.com/huggingface/datasets/blob/d25a0bf94d9f9a9aa6cabdf5b450b9c327d19729/src/datasets/iterable_dataset.py#L197) here as it is
`self.ex_iterable.shuffle_data_sources(seed), function=self.function, batch_size=self.batch_size`, as one can see it is missing batched argument, which means that the iterator fallsback to default constructor value, which in this case is `False`.
To remedy the problem we can change this line to
`self.ex_iterable.shuffle_data_sources(seed), function=self.function, batched=self.batched, batch_size=self.batch_size`
| 2,716 |
https://github.com/huggingface/datasets/issues/2714 | add more precise information for size | [
"We already have this information in the dataset_infos.json files of each dataset.\r\nMaybe we can parse these files in the backend to return their content with the endpoint at huggingface.co/api/datasets\r\n\r\nFor now if you want to access this info you have to load the json for each dataset. For example:\r\n- for a dataset on github like `squad` \r\n- https://raw.githubusercontent.com/huggingface/datasets/master/datasets/squad/dataset_infos.json\r\n- for a community dataset on the hub like `lhoestq/squad`:\r\n https://huggingface.co/datasets/lhoestq/squad/resolve/main/dataset_infos.json"
] | For the import into ELG, we would like a more precise description of the size of the dataset, instead of the current size categories. The size can be expressed in bytes, or any other preferred size unit. As suggested in the slack channel, perhaps this could be computed with a regex for existing datasets. | 2,714 |
https://github.com/huggingface/datasets/issues/2709 | Missing documentation for wnut_17 (ner_tags) | [
"Hi @maxpel, thanks for reporting this issue.\r\n\r\nIndeed, the documentation in the dataset card is not complete. I’m opening a Pull Request to fix it.\r\n\r\nAs the paper explains, there are 6 entity types and we have ordered them alphabetically: `corporation`, `creative-work`, `group`, `location`, `person` and `product`. \r\n\r\nEach of these entity types has 2 possible IOB2 format tags: \r\n- `B-`: to indicate that the token is the beginning of an entity name, and the \r\n- `I-`: to indicate that the token is inside an entity name. \r\n\r\nAdditionally, there is the standalone IOB2 tag \r\n- `O`: that indicates that the token belongs to no named entity. \r\n\r\nIn total there are 13 possible tags, which correspond to the following integer numbers:\r\n\r\n0. `O`\r\n1. `B-corporation`\r\n2. `I-corporation`\r\n3. `B-creative-work`\r\n4. `I-creative-work`\r\n5. `B-group`\r\n6. `I-group`\r\n7. `B-location`\r\n8. `I-location`\r\n9. `B-person`\r\n10. `I-person`\r\n11. `B-product`\r\n12. `I-product`"
] | On the info page of the wnut_17 data set (https://huggingface.co/datasets/wnut_17), the model output of ner-tags is only documented for these 5 cases:
`ner_tags: a list of classification labels, with possible values including O (0), B-corporation (1), I-corporation (2), B-creative-work (3), I-creative-work (4).`
I trained a model with the data and it gives me 13 classes:
```
"id2label": {
"0": 0,
"1": 1,
"2": 2,
"3": 3,
"4": 4,
"5": 5,
"6": 6,
"7": 7,
"8": 8,
"9": 9,
"10": 10,
"11": 11,
"12": 12
}
"label2id": {
"0": 0,
"1": 1,
"10": 10,
"11": 11,
"12": 12,
"2": 2,
"3": 3,
"4": 4,
"5": 5,
"6": 6,
"7": 7,
"8": 8,
"9": 9
}
```
The paper (https://www.aclweb.org/anthology/W17-4418.pdf) explains those 6 categories, but the ordering does not match:
```
1. person
2. location (including GPE, facility)
3. corporation
4. product (tangible goods, or well-defined
services)
5. creative-work (song, movie, book and
so on)
6. group (subsuming music band, sports team,
and non-corporate organisations)
```
I would be very helpful for me, if somebody could clarify the model ouputs and explain the "B-" and "I-" prefixes to me.
Really great work with that and the other packages, I couldn't believe that training the model with that data was basically a one-liner! | 2,709 |
https://github.com/huggingface/datasets/issues/2708 | QASC: incomplete training set | [
"Hi @danyaljj, thanks for reporting.\r\n\r\nUnfortunately, I have not been able to reproduce your problem. My train split has 8134 examples:\r\n```ipython\r\nIn [10]: ds[\"train\"]\r\nOut[10]:\r\nDataset({\r\n features: ['id', 'question', 'choices', 'answerKey', 'fact1', 'fact2', 'combinedfact', 'formatted_question'],\r\n num_rows: 8134\r\n})\r\n\r\nIn [11]: ds[\"train\"].shape\r\nOut[11]: (8134, 8)\r\n```\r\nand the content of the last 5 examples is:\r\n```ipython\r\nIn [12]: for i in range(8129, 8134):\r\n ...: print(json.dumps(ds[\"train\"][i]))\r\n ...:\r\n{\"id\": \"3KAKFY4PGU1LGXM77JAK2700NGCI3X\", \"question\": \"Chitin can be used for protection by whom?\", \"choices\": {\"text\": [\"Fungi\", \"People\", \"Man\", \"Fish\", \"trees\", \"Dogs\", \"animal\", \"Birds\"], \"label\": [\"A\", \"B\",\r\n \"C\", \"D\", \"E\", \"F\", \"G\", \"H\"]}, \"answerKey\": \"D\", \"fact1\": \"scales are used for protection by scaled animals\", \"fact2\": \"Fish scales are also composed of chitin.\", \"combinedfact\": \"Chitin can be used for prote\r\nction by fish.\", \"formatted_question\": \"Chitin can be used for protection by whom? (A) Fungi (B) People (C) Man (D) Fish (E) trees (F) Dogs (G) animal (H) Birds\"}\r\n{\"id\": \"336YQZE83VDAQVZ26HW59X51JZ9M5M\", \"question\": \"Which type of animal uses plates for protection?\", \"choices\": {\"text\": [\"squids\", \"reptiles\", \"sea urchins\", \"fish\", \"amphibians\", \"Frogs\", \"mammals\", \"salm\r\non\"], \"label\": [\"A\", \"B\", \"C\", \"D\", \"E\", \"F\", \"G\", \"H\"]}, \"answerKey\": \"B\", \"fact1\": \"scales are used for protection by scaled animals\", \"fact2\": \"Reptiles have scales or plates.\", \"combinedfact\": \"Reptiles use\r\n their plates for protection.\", \"formatted_question\": \"Which type of animal uses plates for protection? (A) squids (B) reptiles (C) sea urchins (D) fish (E) amphibians (F) Frogs (G) mammals (H) salmon\"}\r\n{\"id\": \"3WZ36BJEV3FGS66VGOOUYX0LN8GTBU\", \"question\": \"What are used for protection by fish?\", \"choices\": {\"text\": [\"scales\", \"fins\", \"streams.\", \"coral\", \"gills\", \"Collagen\", \"mussels\", \"whiskers\"], \"label\": [\"\r\nA\", \"B\", \"C\", \"D\", \"E\", \"F\", \"G\", \"H\"]}, \"answerKey\": \"A\", \"fact1\": \"scales are used for protection by scaled animals\", \"fact2\": \"Fish are backboned aquatic animals.\", \"combinedfact\": \"scales are used for prote\r\nction by fish \", \"formatted_question\": \"What are used for protection by fish? (A) scales (B) fins (C) streams. (D) coral (E) gills (F) Collagen (G) mussels (H) whiskers\"}\r\n{\"id\": \"3Z2R0DQ0JHDKFAO2706OYIXGNA4E28\", \"question\": \"What are pangolins covered in?\", \"choices\": {\"text\": [\"tunicates\", \"Echinoids\", \"shells\", \"exoskeleton\", \"blastoids\", \"barrel-shaped\", \"protection\", \"white\"\r\n], \"label\": [\"A\", \"B\", \"C\", \"D\", \"E\", \"F\", \"G\", \"H\"]}, \"answerKey\": \"G\", \"fact1\": \"scales are used for protection by scaled animals\", \"fact2\": \"Pangolins have an elongate and tapering body covered above with ov\r\nerlapping scales.\", \"combinedfact\": \"Pangolins are covered in overlapping protection.\", \"formatted_question\": \"What are pangolins covered in? (A) tunicates (B) Echinoids (C) shells (D) exoskeleton (E) blastoids\r\n (F) barrel-shaped (G) protection (H) white\"}\r\n{\"id\": \"3PMBY0YE272GIWPNWIF8IH5RBHVC9S\", \"question\": \"What are covered with protection?\", \"choices\": {\"text\": [\"apples\", \"trees\", \"coral\", \"clams\", \"roses\", \"wings\", \"hats\", \"fish\"], \"label\": [\"A\", \"B\", \"C\", \"D\r\n\", \"E\", \"F\", \"G\", \"H\"]}, \"answerKey\": \"H\", \"fact1\": \"scales are used for protection by scaled animals\", \"fact2\": \"Fish are covered with scales.\", \"combinedfact\": \"Fish are covered with protection\", \"formatted_q\r\nuestion\": \"What are covered with protection? (A) apples (B) trees (C) coral (D) clams (E) roses (F) wings (G) hats (H) fish\"}\r\n```\r\n\r\nCould you please load again your dataset and print its shape, like this:\r\n```python\r\nds = load_dataset(\"qasc\", split=\"train)\r\nprint(ds.shape)\r\n```\r\nand confirm which is your output?",
"Hmm .... it must have been a mistake on my side. Sorry for the hassle! "
] | ## Describe the bug
The training instances are not loaded properly.
## Steps to reproduce the bug
```python
from datasets import load_dataset
dataset = load_dataset("qasc", script_version='1.10.2')
def load_instances(split):
instances = dataset[split]
print(f"split: {split} - size: {len(instances)}")
for x in instances:
print(json.dumps(x))
load_instances('test')
load_instances('validation')
load_instances('train')
```
## results
For test and validation, we can see the examples in the output (which is good!):
```
split: test - size: 920
{"answerKey": "", "choices": {"label": ["A", "B", "C", "D", "E", "F", "G", "H"], "text": ["Anthax", "under water", "uterus", "wombs", "two", "moles", "live", "embryo"]}, "combinedfact": "", "fact1": "", "fact2": "", "formatted_question": "What type of birth do therian mammals have? (A) Anthax (B) under water (C) uterus (D) wombs (E) two (F) moles (G) live (H) embryo", "id": "3C44YUNSI1OBFBB8D36GODNOZN9DPA", "question": "What type of birth do therian mammals have?"}
{"answerKey": "", "choices": {"label": ["A", "B", "C", "D", "E", "F", "G", "H"], "text": ["Corvidae", "arthropods", "birds", "backbones", "keratin", "Jurassic", "front paws", "Parakeets."]}, "combinedfact": "", "fact1": "", "fact2": "", "formatted_question": "By what time had mouse-sized viviparous mammals evolved? (A) Corvidae (B) arthropods (C) birds (D) backbones (E) keratin (F) Jurassic (G) front paws (H) Parakeets.", "id": "3B1NLC6UGZVERVLZFT7OUYQLD1SGPZ", "question": "By what time had mouse-sized viviparous mammals evolved?"}
{"answerKey": "", "choices": {"label": ["A", "B", "C", "D", "E", "F", "G", "H"], "text": ["Reduced friction", "causes infection", "vital to a good life", "prevents water loss", "camouflage from consumers", "Protection against predators", "spur the growth of the plant", "a smooth surface"]}, "combinedfact": "", "fact1": "", "fact2": "", "formatted_question": "What does a plant's skin do? (A) Reduced friction (B) causes infection (C) vital to a good life (D) prevents water loss (E) camouflage from consumers (F) Protection against predators (G) spur the growth of the plant (H) a smooth surface", "id": "3QRYMNZ7FYGITFVSJET3PS0F4S0NT9", "question": "What does a plant's skin do?"}
...
```
However, only a few instances are loaded for the training split, which is not correct.
## Environment info
- `datasets` version: '1.10.2'
- Platform: MaxOS
- Python version:3.7
- PyArrow version: 3.0.0
| 2,708 |
https://github.com/huggingface/datasets/issues/2707 | 404 Not Found Error when loading LAMA dataset | [
"Hi @dwil2444! I was able to reproduce your error when I downgraded to v1.1.2. Updating to the latest version of Datasets fixed the error for me :)",
"Hi @dwil2444, thanks for reporting.\r\n\r\nCould you please confirm which `datasets` version you were using and if the problem persists after you update it to the latest version: `pip install -U datasets`?\r\n\r\nThanks @stevhliu for the hint to fix this! ;)",
"@stevhliu @albertvillanova updating to the latest version of datasets did in fact fix this issue. Thanks a lot for your help!"
] | The [LAMA](https://huggingface.co/datasets/viewer/?dataset=lama) probing dataset is not available for download:
Steps to Reproduce:
1. `from datasets import load_dataset`
2. `dataset = load_dataset('lama', 'trex')`.
Results:
`FileNotFoundError: Couldn't find file locally at lama/lama.py, or remotely at https://raw.githubusercontent.com/huggingface/datasets/1.1.2/datasets/lama/lama.py or https://s3.amazonaws.com/datasets.huggingface.co/datasets/datasets/lama/lama.py` | 2,707 |
https://github.com/huggingface/datasets/issues/2705 | 404 not found error on loading WIKIANN dataset | [
"Hi @ronbutan, thanks for reporting.\r\n\r\nYou are right: we have recently found that the link to the original PAN-X dataset (also called WikiANN), hosted at Dropbox, is no longer working.\r\n\r\nWe have opened an issue in the GitHub repository of the original dataset (afshinrahimi/mmner#4) and we have also contacted the author by email to ask if they are planning to fix this issue. See the details here: https://github.com/huggingface/datasets/issues/2691#issuecomment-885463027\r\n\r\nI close this issue because it is the same as in #2691. Feel free to subscribe to that other issue to be informed about any updates."
] | ## Describe the bug
Unable to retreive wikiann English dataset
## Steps to reproduce the bug
```python
from datasets import list_datasets, load_dataset, list_metrics, load_metric
WIKIANN = load_dataset("wikiann","en")
```
## Expected results
Colab notebook should display successful download status
## Actual results
FileNotFoundError: Couldn't find file at https://www.dropbox.com/s/12h3qqog6q4bjve/panx_dataset.tar?dl=1
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.10.1
- Platform: Linux-5.4.104+-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.11
- PyArrow version: 3.0.0
| 2,705 |
https://github.com/huggingface/datasets/issues/2703 | Bad message when config name is missing | [] | When loading a dataset that have several configurations, we expect to see an error message if the user doesn't specify a config name.
However in `datasets` 1.10.0 and 1.10.1 it doesn't show the right message:
```python
import datasets
datasets.load_dataset("glue")
```
raises
```python
AttributeError: 'BuilderConfig' object has no attribute 'text_features'
```
instead of
```python
ValueError: Config name is missing.
Please pick one among the available configs: ['cola', 'sst2', 'mrpc', 'qqp', 'stsb', 'mnli', 'mnli_mismatched', 'mnli_matched', 'qnli', 'rte', 'wnli', 'ax']
Example of usage:
`load_dataset('glue', 'cola')`
``` | 2,703 |
https://github.com/huggingface/datasets/issues/2700 | from datasets import Dataset is failing | [
"Hi @kswamy15, thanks for reporting.\r\n\r\nWe are fixing this critical issue and making an urgent patch release of the `datasets` library today.\r\n\r\nIn the meantime, you can circumvent this issue by updating the `tqdm` library: `!pip install -U tqdm`"
] | ## Describe the bug
A clear and concise description of what the bug is.
## Steps to reproduce the bug
```python
# Sample code to reproduce the bug
from datasets import Dataset
```
## Expected results
A clear and concise description of the expected results.
## Actual results
Specify the actual results or traceback.
/usr/local/lib/python3.7/dist-packages/datasets/utils/file_utils.py in <module>()
25 import posixpath
26 import requests
---> 27 from tqdm.contrib.concurrent import thread_map
28
29 from .. import __version__, config, utils
ModuleNotFoundError: No module named 'tqdm.contrib.concurrent'
---------------------------------------------------------------------------
NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.
To view examples of installing some common dependencies, click the
"Open Examples" button below.
---------------------------------------------------------------------------
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: latest version as of 07/21/2021
- Platform: Google Colab
- Python version: 3.7
- PyArrow version:
| 2,700 |
https://github.com/huggingface/datasets/issues/2699 | cannot combine splits merging and streaming? | [
"Hi ! That's missing indeed. We'll try to implement this for the next version :)\r\n\r\nI guess we just need to implement #2564 first, and then we should be able to add support for splits combinations",
"is there an update on this? ran into the same issue on 2.17.1.\r\n\r\nOn a similar note, the keyword `split=\"all\"` also does not work as intended when `streaming=True`. ",
"No update so far, especially since we haven't implemented an efficient way to query `split=train[50%:]` for example. The addition of two splits should be easy though, since we have `concatenate_datasets()`",
"Can you concatenate_datasets that are being streamed now? I was led to believe concatenation works on non streaming datasets only.",
"Yes `concatenate_datasets` works for datasets loaded in streaming mode as well"
] | this does not work:
`dataset = datasets.load_dataset('mc4','iw',split='train+validation',streaming=True)`
with error:
`ValueError: Bad split: train+validation. Available splits: ['train', 'validation']`
these work:
`dataset = datasets.load_dataset('mc4','iw',split='train+validation')`
`dataset = datasets.load_dataset('mc4','iw',split='train',streaming=True)`
`dataset = datasets.load_dataset('mc4','iw',split='validation',streaming=True)`
i could not find a reference to this in the documentation and the error message is confusing. also would be nice to allow streaming for the merged splits | 2,699 |
https://github.com/huggingface/datasets/issues/2695 | Cannot import load_dataset on Colab | [
"I'm facing the same issue on Colab today too.\r\n\r\n```\r\nModuleNotFoundError Traceback (most recent call last)\r\n<ipython-input-4-5833ac0f5437> in <module>()\r\n 3 \r\n 4 from ray import tune\r\n----> 5 from datasets import DatasetDict, Dataset\r\n 6 from datasets import load_dataset, load_metric\r\n 7 from dataclasses import dataclass\r\n\r\n7 frames\r\n/usr/local/lib/python3.7/dist-packages/datasets/utils/file_utils.py in <module>()\r\n 25 import posixpath\r\n 26 import requests\r\n---> 27 from tqdm.contrib.concurrent import thread_map\r\n 28 \r\n 29 from .. import __version__, config, utils\r\n\r\nModuleNotFoundError: No module named 'tqdm.contrib.concurrent'\r\n\r\n---------------------------------------------------------------------------\r\nNOTE: If your import is failing due to a missing package, you can\r\nmanually install dependencies using either !pip or !apt.\r\n\r\nTo view examples of installing some common dependencies, click the\r\n\"Open Examples\" button below.\r\n---------------------------------------------------------------------------\r\n```",
"@phosseini \r\nI think it is related to [1.10.0](https://github.com/huggingface/datasets/actions/runs/1052653701) release done 3 hours ago. (cc: @lhoestq )\r\nFor now I just downgraded to 1.9.0 and it is working fine.",
"> @phosseini\r\n> I think it is related to [1.10.0](https://github.com/huggingface/datasets/actions/runs/1052653701) release done 3 hours ago. (cc: @lhoestq )\r\n> For now I just downgraded to 1.9.0 and it is working fine.\r\n\r\nSame here, downgraded to 1.9.0 for now and works fine.",
"Hi, \r\n\r\nupdating tqdm to the newest version resolves the issue for me. You can do this as follows in Colab:\r\n```\r\n!pip install tqdm --upgrade\r\n```",
"Hi @bayartsogt-ya and @phosseini, thanks for reporting.\r\n\r\nWe are fixing this critical issue and making an urgent patch release of the `datasets` library today.\r\n\r\nIn the meantime, as pointed out by @mariosasko, you can circumvent this issue by updating the `tqdm` library: \r\n```\r\n!pip install -U tqdm\r\n```"
] | ## Describe the bug
Got tqdm concurrent module not found error during importing load_dataset from datasets.
## Steps to reproduce the bug
Here [colab notebook](https://colab.research.google.com/drive/1pErWWnVP4P4mVHjSFUtkePd8Na_Qirg4?usp=sharing) to reproduce the error
On colab:
```python
!pip install datasets
from datasets import load_dataset
```
## Expected results
Works without error
## Actual results
Specify the actual results or traceback.
```
ModuleNotFoundError Traceback (most recent call last)
<ipython-input-2-8cc7de4c69eb> in <module>()
----> 1 from datasets import load_dataset, load_metric, Metric, MetricInfo, Features, Value
2 from sklearn.metrics import mean_squared_error
/usr/local/lib/python3.7/dist-packages/datasets/__init__.py in <module>()
31 )
32
---> 33 from .arrow_dataset import Dataset, concatenate_datasets
34 from .arrow_reader import ArrowReader, ReadInstruction
35 from .arrow_writer import ArrowWriter
/usr/local/lib/python3.7/dist-packages/datasets/arrow_dataset.py in <module>()
40 from tqdm.auto import tqdm
41
---> 42 from datasets.tasks.text_classification import TextClassification
43
44 from . import config, utils
/usr/local/lib/python3.7/dist-packages/datasets/tasks/__init__.py in <module>()
1 from typing import Optional
2
----> 3 from ..utils.logging import get_logger
4 from .automatic_speech_recognition import AutomaticSpeechRecognition
5 from .base import TaskTemplate
/usr/local/lib/python3.7/dist-packages/datasets/utils/__init__.py in <module>()
19
20 from . import logging
---> 21 from .download_manager import DownloadManager, GenerateMode
22 from .file_utils import DownloadConfig, cached_path, hf_bucket_url, is_remote_url, temp_seed
23 from .mock_download_manager import MockDownloadManager
/usr/local/lib/python3.7/dist-packages/datasets/utils/download_manager.py in <module>()
24
25 from .. import config
---> 26 from .file_utils import (
27 DownloadConfig,
28 cached_path,
/usr/local/lib/python3.7/dist-packages/datasets/utils/file_utils.py in <module>()
25 import posixpath
26 import requests
---> 27 from tqdm.contrib.concurrent import thread_map
28
29 from .. import __version__, config, utils
ModuleNotFoundError: No module named 'tqdm.contrib.concurrent'
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.10.0
- Platform: Colab
- Python version: 3.7.11
- PyArrow version: 3.0.0
| 2,695 |
https://github.com/huggingface/datasets/issues/2691 | xtreme / pan-x cannot be downloaded | [
"Hi @severo, thanks for reporting.\r\n\r\nHowever I have not been able to reproduce this issue. Could you please confirm if the problem persists for you?\r\n\r\nMaybe Dropbox (where the data source is hosted) was temporarily unavailable when you tried.",
"Hmmm, the file (https://www.dropbox.com/s/dl/12h3qqog6q4bjve/panx_dataset.tar) really seems to be unavailable... I tried from various connexions and machines and got the same 404 error. Maybe the dataset has been loaded from the cache in your case?",
"Yes @severo, weird... I could access the file when I answered to you, but now I cannot longer access it either... Maybe it was from the cache as you point out.\r\n\r\nAnyway, I have opened an issue in the GitHub repository responsible for the original dataset: https://github.com/afshinrahimi/mmner/issues/4\r\nI have also contacted the maintainer by email.\r\n\r\nI'll keep you informed with their answer.",
"Reply from the author/maintainer: \r\n> Will fix the issue and let you know during the weekend.",
"The author told that apparently Dropbox has changed their policy and no longer allow downloading the file without having signed in first. The author asked Hugging Face to host their dataset."
] | ## Describe the bug
Dataset xtreme / pan-x cannot be loaded
Seems related to https://github.com/huggingface/datasets/pull/2326
## Steps to reproduce the bug
```python
dataset = load_dataset("xtreme", "PAN-X.fr")
```
## Expected results
Load the dataset
## Actual results
```
FileNotFoundError: Couldn't find file at https://www.dropbox.com/s/12h3qqog6q4bjve/panx_dataset.tar?dl=1
```
## Environment info
- `datasets` version: 1.9.0
- Platform: macOS-11.4-x86_64-i386-64bit
- Python version: 3.8.11
- PyArrow version: 4.0.1
| 2,691 |
https://github.com/huggingface/datasets/issues/2689 | cannot save the dataset to disk after rename_column | [
"Hi ! That's because you are trying to overwrite a file that is already open and being used.\r\nIndeed `foo/dataset.arrow` is open and used by your `dataset` object.\r\n\r\nWhen you do `rename_column`, the resulting dataset reads the data from the same arrow file.\r\nIn other cases like when using `map` on the other hand, the resulting dataset reads the data from another arrow file that is the result of the map transform.\r\n\r\nTherefore overwriting a dataset after `rename_column` is not possible, but it is possible after `map`, since `rename_column` doesn't switch to using another arrow file (the actual data stay the same).",
"Ok, thanks for clearing it up :)",
"so what would be the right way to read a dataset, then change something and then overwrite it with the new version?",
"You have to write to a new directory, then delete the former directory, and finally rename the new directory."
] | ## Describe the bug
If you use `rename_column` and do no other modification, you will be unable to save the dataset using `save_to_disk`
## Steps to reproduce the bug
```python
# Sample code to reproduce the bug
In [1]: from datasets import Dataset, load_from_disk
In [5]: dataset=Dataset.from_dict({'foo': [0]})
In [7]: dataset.save_to_disk('foo')
In [8]: dataset=load_from_disk('foo')
In [10]: dataset=dataset.rename_column('foo', 'bar')
In [11]: dataset.save_to_disk('foo')
---------------------------------------------------------------------------
PermissionError Traceback (most recent call last)
<ipython-input-11-a3bc0d4fc339> in <module>
----> 1 dataset.save_to_disk('foo')
/mnt/beegfs/projects/meerqat/anaconda3/envs/meerqat/lib/python3.7/site-packages/datasets/arrow_dataset.py in save_to_disk(self, dataset_path
, fs)
597 if Path(dataset_path, config.DATASET_ARROW_FILENAME) in cache_files_paths:
598 raise PermissionError(
--> 599 f"Tried to overwrite {Path(dataset_path, config.DATASET_ARROW_FILENAME)} but a dataset can't overwrite itself."
600 )
601 if Path(dataset_path, config.DATASET_INDICES_FILENAME) in cache_files_paths:
PermissionError: Tried to overwrite foo/dataset.arrow but a dataset can't overwrite itself.
```
N. B. I created the dataset from dict to enable easy reproduction but the same happens if you load an existing dataset (e.g. starting from `In [8]`)
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.8.0
- Platform: Linux-3.10.0-1160.11.1.el7.x86_64-x86_64-with-centos-7.9.2009-Core
- Python version: 3.7.10
- PyArrow version: 3.0.0
| 2,689 |
https://github.com/huggingface/datasets/issues/2688 | hebrew language codes he and iw should be treated as aliases | [
"Hi @eyaler, thanks for reporting.\r\n\r\nWhile you are true with respect the Hebrew language tag (\"iw\" is deprecated and \"he\" is the preferred value), in the \"mc4\" dataset (which is a derived dataset) we have kept the language tags present in the original dataset: [Google C4](https://www.tensorflow.org/datasets/catalog/c4).",
"For discoverability on the website I updated the YAML tags at the top of the mC4 dataset card https://github.com/huggingface/datasets/commit/38288087b1b02f97586e0346e8f28f4960f1fd37\r\n\r\nOnce the website is updated, mC4 will be listed in https://huggingface.co/datasets?filter=languages:he\r\n\r\n"
] | https://huggingface.co/datasets/mc4 not listed when searching for hebrew datasets (he) as it uses the older language code iw, preventing discoverability. | 2,688 |
https://github.com/huggingface/datasets/issues/2683 | Cache directories changed due to recent changes in how config kwargs are handled | [] | Since #2659 I can see weird cache directory names with hashes in the config id, even though no additional config kwargs are passed. For example:
```python
from datasets import load_dataset_builder
c4_builder = load_dataset_builder("c4", "en")
print(c4_builder.cache_dir)
# /Users/quentinlhoest/.cache/huggingface/datasets/c4/en-174d3b7155eb68db/0.0.0/...
# instead of
# /Users/quentinlhoest/.cache/huggingface/datasets/c4/en/0.0.0/...
```
This issue could be annoying since it would simply ignore old cache directories for users, and regenerate datasets
cc @stas00 this is what you experienced a few days ago
| 2,683 |
https://github.com/huggingface/datasets/issues/2681 | 5 duplicate datasets | [
"Yes this was documented in the PR that added this hf->paperswithcode mapping (https://github.com/huggingface/datasets/pull/2404) and AFAICT those are slightly distinct datasets so I think it's a wontfix\r\n\r\nFor context on the paperswithcode mapping you can also refer to https://github.com/huggingface/huggingface_hub/pull/43 which contains a lot of background discussion ",
"Thanks for the antecedents. I close."
] | ## Describe the bug
In 5 cases, I could find a dataset on Paperswithcode which references two Hugging Face datasets as dataset loaders. They are:
- https://paperswithcode.com/dataset/multinli -> https://huggingface.co/datasets/multi_nli and https://huggingface.co/datasets/multi_nli_mismatch
<img width="838" alt="Capture d’écran 2021-07-20 à 16 33 58" src="https://user-images.githubusercontent.com/1676121/126342757-4625522a-f788-41a3-bd1f-2a8b9817bbf5.png">
- https://paperswithcode.com/dataset/squad -> https://huggingface.co/datasets/squad and https://huggingface.co/datasets/squad_v2
- https://paperswithcode.com/dataset/narrativeqa -> https://huggingface.co/datasets/narrativeqa and https://huggingface.co/datasets/narrativeqa_manual
- https://paperswithcode.com/dataset/hate-speech-and-offensive-language -> https://huggingface.co/datasets/hate_offensive and https://huggingface.co/datasets/hate_speech_offensive
- https://paperswithcode.com/dataset/newsph-nli -> https://huggingface.co/datasets/newsph and https://huggingface.co/datasets/newsph_nli
Possible solutions:
- don't fix (it works)
- for each pair of duplicate datasets, remove one, and create an alias to the other.
## Steps to reproduce the bug
Visit the Paperswithcode links, and look at the "Dataset Loaders" section
## Expected results
There should only be one reference to a Hugging Face dataset loader
## Actual results
Two Hugging Face dataset loaders
| 2,681 |
https://github.com/huggingface/datasets/issues/2679 | Cannot load the blog_authorship_corpus due to codec errors | [
"Hi @izaskr, thanks for reporting.\r\n\r\nHowever the traceback you joined does not correspond to the codec error message: it is about other error `NonMatchingSplitsSizesError`. Maybe you missed some important part of your traceback...\r\n\r\nI'm going to have a look at the dataset anyway...",
"Hi @izaskr, thanks again for having reported this issue.\r\n\r\nAfter investigation, I have created a Pull Request (#2685) to fix several issues with this dataset:\r\n- the `NonMatchingSplitsSizesError`\r\n- the `UnicodeDecodeError`\r\n\r\nOnce the Pull Request merged into master, you will be able to load this dataset if you install `datasets` from our GitHub repository master branch. Otherwise, you will be able to use it after our next release, by updating `datasets`: `pip install -U datasets`.",
"@albertvillanova \r\nCan you shed light on how this fix works?\r\n\r\nWe're experiencing a similar issue. \r\n\r\nIf we run several runs (eg in a Wandb sweep) the first run \"works\" but then we get `NonMatchingSplitsSizesError`\r\n\r\n| run num | actual train examples # | expected example # | recorded example # |\r\n| ------- | -------------- | ----------------- | -------- |\r\n| 1 | 100 | 100 | 100 |\r\n| 2 | 102 | 100 | 102 |\r\n| 3 | 100 | 100 | 202 | \r\n| 4 | 40 | 100 | 40 |\r\n| 5 | 40 | 100 | 40 |\r\n| 6 | 40 | 100 | 40 | \r\n\r\n\r\nThe second through the nth all crash with \r\n\r\n```\r\ndatasets.utils.info_utils.NonMatchingSplitsSizesError: [{'expected': SplitInfo(name='train', num_bytes=19980970, num_examples=100, dataset_name='cies'), 'recorded': SplitInfo(name='train', num_bytes=40163811, num_examples=202, dataset_name='cies')}]\r\n\r\n```"
] | ## Describe the bug
A codec error is raised while loading the blog_authorship_corpus.
## Steps to reproduce the bug
```
from datasets import load_dataset
raw_datasets = load_dataset("blog_authorship_corpus")
```
## Expected results
Loading the dataset without errors.
## Actual results
An error similar to the one below was raised for (what seems like) every XML file.
/home/izaskr/.cache/huggingface/datasets/downloads/extracted/7cf52524f6517e168604b41c6719292e8f97abbe8f731e638b13423f4212359a/blogs/788358.male.24.Arts.Libra.xml cannot be loaded. Error message: 'utf-8' codec can't decode byte 0xe7 in position 7551: invalid continuation byte
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/izaskr/anaconda3/envs/local_vae_older/lib/python3.8/site-packages/datasets/load.py", line 856, in load_dataset
builder_instance.download_and_prepare(
File "/home/izaskr/anaconda3/envs/local_vae_older/lib/python3.8/site-packages/datasets/builder.py", line 583, in download_and_prepare
self._download_and_prepare(
File "/home/izaskr/anaconda3/envs/local_vae_older/lib/python3.8/site-packages/datasets/builder.py", line 671, in _download_and_prepare
verify_splits(self.info.splits, split_dict)
File "/home/izaskr/anaconda3/envs/local_vae_older/lib/python3.8/site-packages/datasets/utils/info_utils.py", line 74, in verify_splits
raise NonMatchingSplitsSizesError(str(bad_splits))
datasets.utils.info_utils.NonMatchingSplitsSizesError: [{'expected': SplitInfo(name='train', num_bytes=610252351, num_examples=532812, dataset_name='blog_authorship_corpus'), 'recorded': SplitInfo(name='train', num_bytes=614706451, num_examples=535568, dataset_name='blog_authorship_corpus')}, {'expected': SplitInfo(name='validation', num_bytes=37500394, num_examples=31277, dataset_name='blog_authorship_corpus'), 'recorded': SplitInfo(name='validation', num_bytes=32553710, num_examples=28521, dataset_name='blog_authorship_corpus')}]
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.9.0
- Platform: Linux-4.15.0-132-generic-x86_64-with-glibc2.10
- Python version: 3.8.8
- PyArrow version: 4.0.1
| 2,679 |
https://github.com/huggingface/datasets/issues/2678 | Import Error in Kaggle notebook | [
"This looks like an issue with PyArrow. Did you try reinstalling it ?",
"@lhoestq I did, and then let pip handle the installation in `pip import datasets`. I also tried using conda but it gives the same error.\r\n\r\nEdit: pyarrow version on kaggle is 4.0.0, it gets replaced with 4.0.1. So, I don't think uninstalling will change anything.\r\n```\r\nInstall Trace of datasets:\r\n\r\nCollecting datasets\r\n Downloading datasets-1.9.0-py3-none-any.whl (262 kB)\r\n |████████████████████████████████| 262 kB 834 kB/s eta 0:00:01\r\nRequirement already satisfied: dill in /opt/conda/lib/python3.7/site-packages (from datasets) (0.3.4)\r\nCollecting pyarrow!=4.0.0,>=1.0.0\r\n Downloading pyarrow-4.0.1-cp37-cp37m-manylinux2014_x86_64.whl (21.8 MB)\r\n |████████████████████████████████| 21.8 MB 6.2 MB/s eta 0:00:01\r\nRequirement already satisfied: importlib-metadata in /opt/conda/lib/python3.7/site-packages (from datasets) (3.4.0)\r\nRequirement already satisfied: huggingface-hub<0.1.0 in /opt/conda/lib/python3.7/site-packages (from datasets) (0.0.8)\r\nRequirement already satisfied: pandas in /opt/conda/lib/python3.7/site-packages (from datasets) (1.2.4)\r\nRequirement already satisfied: requests>=2.19.0 in /opt/conda/lib/python3.7/site-packages (from datasets) (2.25.1)\r\nRequirement already satisfied: fsspec>=2021.05.0 in /opt/conda/lib/python3.7/site-packages (from datasets) (2021.6.1)\r\nRequirement already satisfied: multiprocess in /opt/conda/lib/python3.7/site-packages (from datasets) (0.70.12.2)\r\nRequirement already satisfied: packaging in /opt/conda/lib/python3.7/site-packages (from datasets) (20.9)\r\nCollecting xxhash\r\n Downloading xxhash-2.0.2-cp37-cp37m-manylinux2010_x86_64.whl (243 kB)\r\n |████████████████████████████████| 243 kB 23.7 MB/s eta 0:00:01\r\nRequirement already satisfied: numpy>=1.17 in /opt/conda/lib/python3.7/site-packages (from datasets) (1.19.5)\r\nRequirement already satisfied: tqdm>=4.27 in /opt/conda/lib/python3.7/site-packages (from datasets) (4.61.1)\r\nRequirement already satisfied: filelock in /opt/conda/lib/python3.7/site-packages (from huggingface-hub<0.1.0->datasets) (3.0.12)\r\nRequirement already satisfied: urllib3<1.27,>=1.21.1 in /opt/conda/lib/python3.7/site-packages (from requests>=2.19.0->datasets) (1.26.5)\r\nRequirement already satisfied: idna<3,>=2.5 in /opt/conda/lib/python3.7/site-packages (from requests>=2.19.0->datasets) (2.10)\r\nRequirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.7/site-packages (from requests>=2.19.0->datasets) (2021.5.30)\r\nRequirement already satisfied: chardet<5,>=3.0.2 in /opt/conda/lib/python3.7/site-packages (from requests>=2.19.0->datasets) (4.0.0)\r\nRequirement already satisfied: typing-extensions>=3.6.4 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata->datasets) (3.7.4.3)\r\nRequirement already satisfied: zipp>=0.5 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata->datasets) (3.4.1)\r\nRequirement already satisfied: pyparsing>=2.0.2 in /opt/conda/lib/python3.7/site-packages (from packaging->datasets) (2.4.7)\r\nRequirement already satisfied: python-dateutil>=2.7.3 in /opt/conda/lib/python3.7/site-packages (from pandas->datasets) (2.8.1)\r\nRequirement already satisfied: pytz>=2017.3 in /opt/conda/lib/python3.7/site-packages (from pandas->datasets) (2021.1)\r\nRequirement already satisfied: six>=1.5 in /opt/conda/lib/python3.7/site-packages (from python-dateutil>=2.7.3->pandas->datasets) (1.15.0)\r\nInstalling collected packages: xxhash, pyarrow, datasets\r\n Attempting uninstall: pyarrow\r\n Found existing installation: pyarrow 4.0.0\r\n Uninstalling pyarrow-4.0.0:\r\n Successfully uninstalled pyarrow-4.0.0\r\nSuccessfully installed datasets-1.9.0 pyarrow-4.0.1 xxhash-2.0.2\r\nWARNING: Running pip as root will break packages and permissions. You should install packages reliably by using venv: https://pip.pypa.io/warnings/venv\r\n```",
"You may need to restart your kaggle notebook after installing a newer version of `pyarrow`.\r\n\r\nIf it doesn't work we'll probably have to create an issue on [arrow's JIRA](https://issues.apache.org/jira/projects/ARROW/issues/), and maybe ask kaggle why it could fail",
"> You may need to restart your kaggle notebook before after installing a newer version of `pyarrow`.\r\n> \r\n> If it doesn't work we'll probably have to create an issue on [arrow's JIRA](https://issues.apache.org/jira/projects/ARROW/issues/), and maybe ask kaggle why it could fail\r\n\r\nIt works after restarting.\r\nMy bad, I forgot to restart the notebook. Sorry for the trouble!"
] | ## Describe the bug
Not able to import datasets library in kaggle notebooks
## Steps to reproduce the bug
```python
!pip install datasets
import datasets
```
## Expected results
No such error
## Actual results
```
ImportError Traceback (most recent call last)
<ipython-input-9-652e886d387f> in <module>
----> 1 import datasets
/opt/conda/lib/python3.7/site-packages/datasets/__init__.py in <module>
31 )
32
---> 33 from .arrow_dataset import Dataset, concatenate_datasets
34 from .arrow_reader import ArrowReader, ReadInstruction
35 from .arrow_writer import ArrowWriter
/opt/conda/lib/python3.7/site-packages/datasets/arrow_dataset.py in <module>
36 import pandas as pd
37 import pyarrow as pa
---> 38 import pyarrow.compute as pc
39 from multiprocess import Pool, RLock
40 from tqdm.auto import tqdm
/opt/conda/lib/python3.7/site-packages/pyarrow/compute.py in <module>
16 # under the License.
17
---> 18 from pyarrow._compute import ( # noqa
19 Function,
20 FunctionOptions,
ImportError: /opt/conda/lib/python3.7/site-packages/pyarrow/_compute.cpython-37m-x86_64-linux-gnu.so: undefined symbol: _ZNK5arrow7compute15KernelSignature8ToStringEv
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.9.0
- Platform: Kaggle
- Python version: 3.7.10
- PyArrow version: 4.0.1
| 2,678 |
https://github.com/huggingface/datasets/issues/2677 | Error when downloading C4 | [
"Hi Thanks for reporting !\r\nIt looks like these files are not correctly reported in the list of expected files to download, let me fix that ;)",
"Alright this is fixed now. We'll do a new release soon to make the fix available.\r\n\r\nIn the meantime feel free to simply pass `ignore_verifications=True` to `load_dataset` to skip this error",
"@lhoestq thank you for such a quick feedback!"
] | Hi,
I am trying to download `en` corpus from C4 dataset. However, I get an error caused by validation files download (see image). My code is very primitive:
`datasets.load_dataset('c4', 'en')`
Is this a bug or do I have some configurations missing on my server?
Thanks!
<img width="1014" alt="Снимок экрана 2021-07-20 в 11 37 17" src="https://user-images.githubusercontent.com/36672861/126289448-6e0db402-5f3f-485a-bf74-eb6e0271fc25.png"> | 2,677 |
https://github.com/huggingface/datasets/issues/2670 | Using sharding to parallelize indexing | [] | **Is your feature request related to a problem? Please describe.**
Creating an elasticsearch index on large dataset could be quite long and cannot be parallelized on shard (the index creation is colliding)
**Describe the solution you'd like**
When working on dataset shards, if an index already exists, its mapping should be checked and if compatible, the indexing process should continue with the shard data.
Additionally, at the end of the process, the `_indexes` dict should be send back to the original dataset object (from which the shards have been created) to allow to use the index for later filtering on the whole dataset.
**Describe alternatives you've considered**
Each dataset shard could created independent partial indices. then on the whole dataset level, indices should be all referred in `_indexes` dict and be used in querying through `get_nearest_examples()`. The drawback is that the scores will be computed independently on the partial indices leading to inconsistent values for most scoring based on corpus level statistics (tf/idf, BM25).
**Additional context**
The objectives is to parallelize the index creation to speed-up the process (ie surcharging the ES server which is fine to handle large load) while later enabling search on the whole dataset. | 2,670 |
https://github.com/huggingface/datasets/issues/2669 | Metric kwargs are not passed to underlying external metric f1_score | [
"Hi @BramVanroy, thanks for reporting.\r\n\r\nFirst, note that `\"min\"` is not an allowed value for `average`. According to scikit-learn [documentation](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.f1_score.html), `average` can only take the values: `{\"micro\", \"macro\", \"samples\", \"weighted\", \"binary\"} or None, default=\"binary\"`.\r\n\r\nSecond, you should take into account that all additional metric-specific argument should be passed in the method `compute` (and not in the method `load_metric`). You can find more information in our documentation: https://huggingface.co/docs/datasets/using_metrics.html#computing-the-metric-scores\r\n\r\nSo for example, if you would like to calculate the macro-averaged F1 score, you should use:\r\n```python\r\nimport datasets\r\n\r\nf1 = datasets.load_metric(\"f1\", keep_in_memory=True)\r\nf1.add_batch(predictions=[0,2,3], references=[1, 2, 3])\r\nf1.compute(average=\"macro\")\r\n```",
"Thanks, that was it. A bit strange though, since `load_metric` had an argument `metric_init_kwargs`. I assume that that's for specific initialisation arguments whereas `average` is for the function itself."
] | ## Describe the bug
When I want to use F1 score with average="min", this keyword argument does not seem to be passed through to the underlying sklearn metric. This is evident because [sklearn](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.f1_score.html) throws an error telling me so.
## Steps to reproduce the bug
```python
import datasets
f1 = datasets.load_metric("f1", keep_in_memory=True, average="min")
f1.add_batch(predictions=[0,2,3], references=[1, 2, 3])
f1.compute()
```
## Expected results
No error, because `average="min"` should be passed correctly to f1_score in sklearn.
## Actual results
```
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "C:\Users\bramv\.virtualenvs\pipeline-TpEsXVex\lib\site-packages\datasets\metric.py", line 402, in compute
output = self._compute(predictions=predictions, references=references, **kwargs)
File "C:\Users\bramv\.cache\huggingface\modules\datasets_modules\metrics\f1\82177930a325d4c28342bba0f116d73f6d92fb0c44cd67be32a07c1262b61cfe\f1.py", line 97, in _compute
"f1": f1_score(
File "C:\Users\bramv\.virtualenvs\pipeline-TpEsXVex\lib\site-packages\sklearn\utils\validation.py", line 63, in inner_f
return f(*args, **kwargs)
File "C:\Users\bramv\.virtualenvs\pipeline-TpEsXVex\lib\site-packages\sklearn\metrics\_classification.py", line 1071, in f1_score
return fbeta_score(y_true, y_pred, beta=1, labels=labels,
File "C:\Users\bramv\.virtualenvs\pipeline-TpEsXVex\lib\site-packages\sklearn\utils\validation.py", line 63, in inner_f
return f(*args, **kwargs)
File "C:\Users\bramv\.virtualenvs\pipeline-TpEsXVex\lib\site-packages\sklearn\metrics\_classification.py", line 1195, in fbeta_score
_, _, f, _ = precision_recall_fscore_support(y_true, y_pred,
File "C:\Users\bramv\.virtualenvs\pipeline-TpEsXVex\lib\site-packages\sklearn\utils\validation.py", line 63, in inner_f
return f(*args, **kwargs)
File "C:\Users\bramv\.virtualenvs\pipeline-TpEsXVex\lib\site-packages\sklearn\metrics\_classification.py", line 1464, in precision_recall_fscore_support
labels = _check_set_wise_labels(y_true, y_pred, average, labels,
File "C:\Users\bramv\.virtualenvs\pipeline-TpEsXVex\lib\site-packages\sklearn\metrics\_classification.py", line 1294, in _check_set_wise_labels
raise ValueError("Target is %s but average='binary'. Please "
ValueError: Target is multiclass but average='binary'. Please choose another average setting, one of [None, 'micro', 'macro', 'weighted'].
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.9.0
- Platform: Windows-10-10.0.19041-SP0
- Python version: 3.9.2
- PyArrow version: 4.0.1 | 2,669 |
https://github.com/huggingface/datasets/issues/2663 | [`to_json`] add multi-proc sharding support | [
"Hi @stas00, \r\nI want to work on this issue and I was thinking why don't we use `imap` [in this loop](https://github.com/huggingface/datasets/blob/440b14d0dd428ae1b25881aa72ba7bbb8ad9ff84/src/datasets/io/json.py#L99)? This way, using offset (which is being used to slice the pyarrow table) we can convert pyarrow table to `json` using multiprocessing. I've a small code snippet for some clarity:\r\n```\r\nresult = list(\r\n pool.imap(self._apply_df, [(offset, batch_size) for offset in range(0, len(self.dataset), batch_size)])\r\n )\r\n```\r\n`_apply_df` is a function which will return `batch.to_pandas().to_json(path_or_buf=None, orient=\"records\", lines=True)` which is basically json version of the batched pyarrow table. Later on we can concatenate it to form json file? \r\n\r\nI think the only downside here is to write file from `imap` output (output would be a list and we'll need to iterate over it and write in a file) which might add a little overhead cost. What do you think about this?",
"Followed up in https://github.com/huggingface/datasets/pull/2747"
] | As discussed on slack it appears that `to_json` is quite slow on huge datasets like OSCAR.
I implemented sharded saving, which is much much faster - but the tqdm bars all overwrite each other, so it's hard to make sense of the progress, so if possible ideally this multi-proc support could be implemented internally in `to_json` via `num_proc` argument. I guess `num_proc` will be the number of shards?
I think the user will need to use this feature wisely, since too many processes writing to say normal style HD is likely to be slower than one process.
I'm not sure whether the user should be responsible to concatenate the shards at the end or `datasets`, either way works for my needs.
The code I was using:
```
from multiprocessing import cpu_count, Process, Queue
[...]
filtered_dataset = concat_dataset.map(filter_short_documents, batched=True, batch_size=256, num_proc=cpu_count())
DATASET_NAME = "oscar"
SHARDS = 10
def process_shard(idx):
print(f"Sharding {idx}")
ds_shard = filtered_dataset.shard(SHARDS, idx, contiguous=True)
# ds_shard = ds_shard.shuffle() # remove contiguous=True above if shuffling
print(f"Saving {DATASET_NAME}-{idx}.jsonl")
ds_shard.to_json(f"{DATASET_NAME}-{idx}.jsonl", orient="records", lines=True, force_ascii=False)
queue = Queue()
processes = [Process(target=process_shard, args=(idx,)) for idx in range(SHARDS)]
for p in processes:
p.start()
for p in processes:
p.join()
```
Thank you!
@lhoestq | 2,663 |
https://github.com/huggingface/datasets/issues/2658 | Can't pass `sep=None` to load_dataset("csv", ...) to infer the separator via pandas.read_csv | [] | When doing `load_dataset("csv", sep=None)`, the `sep` passed to `pd.read_csv` is still the default `sep=","` instead, which makes it impossible to make the csv loader infer the separator.
Related to https://github.com/huggingface/datasets/pull/2656
cc @SBrandeis | 2,658 |
https://github.com/huggingface/datasets/issues/2657 | `to_json` reporting enhancements | [] | While using `to_json` 2 things came to mind that would have made the experience easier on the user:
1. Could we have a `desc` arg for the tqdm use and a fallback to just `to_json` so that it'd be clear to the user what's happening? Surely, one can just print the description before calling json, but I thought perhaps it'd help to have it self-identify like you did for other progress bars recently.
2. It took me a while to make sense of the reported numbers:
```
22%|██▏ | 1536/7076 [12:30:57<44:09:42, 28.70s/it]
```
So iteration here happens to be 10K samples, and the total is 70M records. But the user does't know that, so the progress bar is perfect, but the numbers it reports are meaningless until one discovers that 1it=10K samples. And one still has to convert these in the head - so it's not quick. Not exactly sure what's the best way to approach this, perhaps it can be part of `desc`? or report M or K, so it'd be built-in if it were to print, e.g.:
```
22%|██▏ | 15360K/70760K [12:30:57<44:09:42, 28.70s/it]
```
or
```
22%|██▏ | 15.36M/70.76M [12:30:57<44:09:42, 28.70s/it]
```
(while of course remaining friendly to small datasets)
I forget if tqdm lets you add a magnitude identifier to the running count.
Thank you! | 2,657 |
https://github.com/huggingface/datasets/issues/2655 | Allow the selection of multiple columns at once | [
"Hi! I was looking into this and hope you can clarify a point. Your my_dataset variable would be of type DatasetDict which means the alternative you've described (dict comprehension) is what makes sense. \r\nIs there a reason why you wouldn't want to convert my_dataset to a pandas df if you'd like to use it like one? Please let me know if I'm missing something.",
"Hi! Sorry for the delay.\r\n\r\nIn this case, the dataset would be a `datasets.Dataset` and we want to select multiple columns, the `idx` and `label` columns for example.\r\n\r\nMy issue is that my dataset is too big for memory if I load everything into pandas.",
"Hello - any update on this? Thank you.",
"Please note that you could use the method `Dataset.select_columns`: https://huggingface.co/docs/datasets/v2.15.0/en/package_reference/main_classes#datasets.Dataset.select_columns",
"Thank you!"
] | **Is your feature request related to a problem? Please describe.**
Similar to pandas, it would be great if we could select multiple columns at once.
**Describe the solution you'd like**
```python
my_dataset = ... # Has columns ['idx', 'sentence', 'label']
idx, label = my_dataset[['idx', 'label']]
```
**Describe alternatives you've considered**
we can do `[dataset[col] for col in ('idx', 'label')]`
**Additional context**
This is of course very minor.
| 2,655 |
https://github.com/huggingface/datasets/issues/2654 | Give a user feedback if the dataset he loads is streamable or not | [
"#self-assign",
"I understand it already raises a `NotImplementedError` exception, eg:\r\n\r\n```\r\n>>> dataset = load_dataset(\"journalists_questions\", name=\"plain_text\", split=\"train\", streaming=True)\r\n\r\n[...]\r\nNotImplementedError: Extraction protocol for file at https://drive.google.com/uc?export=download&id=1CBrh-9OrSpKmPQBxTK_ji6mq6WTN_U9U is not implemented yet\r\n```\r\n"
] | **Is your feature request related to a problem? Please describe.**
I would love to know if a `dataset` is with the current implementation streamable or not.
**Describe the solution you'd like**
We could show a warning when a dataset is loaded with `load_dataset('...',streaming=True)` when its lot streamable, e.g. if it is an archive.
**Describe alternatives you've considered**
Add a new metadata tag for "streaming"
| 2,654 |
https://github.com/huggingface/datasets/issues/2653 | Add SD task for SUPERB | [
"Note that this subset requires us to:\r\n\r\n* generate the LibriMix corpus from LibriSpeech\r\n* prepare the corpus for diarization\r\n\r\nAs suggested by @lhoestq we should perform these steps locally and add the prepared data to this public repo on the Hub: https://huggingface.co/datasets/superb/superb-data\r\n\r\nThen we can use the URLs for the files to load the data in `superb`'s dataset loading script.\r\n\r\nFor consistency, I suggest we name the folders in `superb-data` in the same way as the configs in the dataset loading script - e.g. use `sd` for speech diarization in both places :)",
"@lewtun @lhoestq: \r\n\r\nI have already generated the LibriMix corpus and prepared the corpus for diarization. The output is 3 dirs (train, dev, test), each one containing 6 files: reco2dur rttm segments spk2utt utt2spk wav.scp\r\n\r\nNext steps:\r\n- Upload these files to the superb-data repo\r\n- Transcribe the corresponding s3prl processing of these files into our superb loading script\r\n\r\nNote that processing of these files is a bit more intricate than usual datasets: https://github.com/s3prl/s3prl/blob/master/s3prl/downstream/diarization/dataset.py#L233\r\n\r\n"
] | Include the SD (Speaker Diarization) task as described in the [SUPERB paper](https://arxiv.org/abs/2105.01051) and `s3prl` [instructions](https://github.com/s3prl/s3prl/tree/master/s3prl/downstream#sd-speaker-diarization).
Steps:
- [x] Generate the LibriMix corpus
- [x] Prepare the corpus for diarization
- [x] Upload these files to the superb-data repo
- [x] Transcribe the corresponding s3prl processing of these files into our superb loading script
- [ ] README: tags + description sections
Related to #2619.
cc: @lewtun
| 2,653 |
https://github.com/huggingface/datasets/issues/2651 | Setting log level higher than warning does not suppress progress bar | [
"Hi,\r\n\r\nyou can suppress progress bars by patching logging as follows:\r\n```python\r\nimport datasets\r\nimport logging\r\ndatasets.logging.get_verbosity = lambda: logging.NOTSET\r\n# map call ...\r\n```\r\nEDIT: now you have to use `disable_progress_bar `",
"Thank you, it worked :)",
"See https://github.com/huggingface/datasets/issues/2528 for reference",
"Note also that you can disable the progress bar with\r\n\r\n```python\r\nfrom datasets.utils import disable_progress_bar\r\ndisable_progress_bar()\r\n```\r\n\r\nSee https://github.com/huggingface/datasets/blob/8814b393984c1c2e1800ba370de2a9f7c8644908/src/datasets/utils/tqdm_utils.py#L84",
"Now the library officially recommends `set_progress_bar_enabled(False)`\r\n\r\n```py\r\nfrom datasets.utils import set_progress_bar_enabled\r\n\r\nset_progress_bar_enabled(False)\r\n```\r\n\r\nsource:\r\n\r\nhttps://github.com/huggingface/datasets/blob/1fd47120ace13626c528367787ffa13e1a26e6c0/src/datasets/utils/tqdm_utils.py#L83-L88\r\n\r\n",
"From https://github.com/huggingface/datasets/pull/3897, `disable_progress_bar` is the function you should use",
"Now ``disable_progress_bar`` function is in ``datasets/src/datasets/utils/logging.py``.\r\nhttps://github.com/huggingface/datasets/blob/aa555a299ad73c65e3f997a764e9d211675ab05d/src/datasets/utils/logging.py#L233-L236\r\nAnd the method mentioned in https://github.com/huggingface/datasets/issues/2651#issuecomment-880270774 is not working now."
] | ## Describe the bug
I would like to disable progress bars for `.map` method (and other methods like `.filter` and `load_dataset` as well).
According to #1627 one can suppress it by setting log level higher than `warning`, however doing so doesn't suppress it with version 1.9.0.
I also tried to set `DATASETS_VERBOSITY` environment variable to `error` or `critical` but it also didn't work.
## Steps to reproduce the bug
```python
import datasets
from datasets.utils.logging import set_verbosity_error
set_verbosity_error()
def dummy_map(batch):
return batch
common_voice_train = datasets.load_dataset("common_voice", "de", split="train")
common_voice_test = datasets.load_dataset("common_voice", "de", split="test")
common_voice_train.map(dummy_map)
```
## Expected results
- The progress bar for `.map` call won't be shown
## Actual results
- The progress bar for `.map` is still shown
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.9.0
- Platform: Linux-5.4.0-1045-aws-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.5
- PyArrow version: 4.0.1
| 2,651 |
https://github.com/huggingface/datasets/issues/2650 | [load_dataset] shard and parallelize the process | [
"I need the same feature for distributed training",
"I think @TevenLeScao is exploring adding multiprocessing in `GeneratorBasedBuilder._prepare_split` - feel free to post updates here :)",
"Posted a PR to address the building side, still needs something to load sharded arrow files + tests",
"Closing as this feature has been implemented in #5107"
] | - Some huge datasets take forever to build the first time. (e.g. oscar/en) as it's done in a single cpu core.
- If the build crashes, everything done up to that point gets lost
Request: Shard the build over multiple arrow files, which would enable:
- much faster build by parallelizing the build process
- if the process crashed, the completed arrow files don't need to be re-built again
Thank you!
@lhoestq | 2,650 |
https://github.com/huggingface/datasets/issues/2649 | adding progress bar / ETA for `load_dataset` | [
"Is this done now? I see progress bars when using `load_dataset`.",
"There are progress bars when downloading data and when preparing them as Arrow files.\r\n\r\nThe \"total silence\" part mentioned in OP refer to checksums verifications which have had some changes in the latest release 2.10:\r\n- they're disabled by default (other less costly verifications are still done like checking the generated dataset size)\r\n- they're using a tqdm bar as well"
] | Please consider:
```
Downloading and preparing dataset oscar/unshuffled_deduplicated_en (download: 462.40 GiB, generated: 1.18 TiB, post-processed: Unknown size, total: 1.63 TiB) to cache/oscar/unshuffled_deduplicated_en/1.0.0/84838bd49d2295f62008383b05620571535451d84545037bb94d6f3501651df2...
HF google storage unreachable. Downloading and preparing it from source
```
and no indication whatsoever of whether things work well or when it'll be done. It's important to have an estimated completion time for when doing slurm jobs since some instances have a cap on run-time.
I think for this particular job it sat for 30min in total silence and then after 30min it started generating:
```
897850 examples [07:24, 10286.71 examples/s]
```
which is already great!
Request:
1. ETA - knowing how many hours to allocate for a slurm job
2. progress bar - helps to know things are working and aren't stuck and where we are at.
Thank you!
@lhoestq
| 2,649 |
https://github.com/huggingface/datasets/issues/2648 | Add web_split dataset for Paraphase and Rephrase benchmark | [
"#take"
] | ## Describe:
For getting simple sentences from complex sentence there are dataset and task like wiki_split that is available in hugging face datasets. This web_split is a very similar dataset. There some research paper which states that by combining these two datasets we if we train the model it will yield better results on both tests data.
This dataset is made from web NLG data.
All the dataset related details are provided in the below repository
Github link: https://github.com/shashiongithub/Split-and-Rephrase
| 2,648 |
https://github.com/huggingface/datasets/issues/2646 | downloading of yahoo_answers_topics dataset failed | [
"Hi ! I just tested and it worked fine today for me.\r\n\r\nI think this is because the dataset is stored on Google Drive which has a quota limit for the number of downloads per day, see this similar issue https://github.com/huggingface/datasets/issues/996 \r\n\r\nFeel free to try again today, now that the quota was reset",
"Fixed once data URL was replaced:\r\n- #4023"
] | ## Describe the bug
I get an error datasets.utils.info_utils.NonMatchingChecksumError: Checksums didn't match for dataset source files when I try to download the yahoo_answers_topics dataset
## Steps to reproduce the bug
self.dataset = load_dataset(
'yahoo_answers_topics', cache_dir=self.config['yahoo_cache_dir'], split='train[:90%]')
# Sample code to reproduce the bug
self.dataset = load_dataset(
'yahoo_answers_topics', cache_dir=self.config['yahoo_cache_dir'], split='train[:90%]')
## Expected results
A clear and concise description of the expected results.
## Actual results
Specify the actual results or traceback.
datasets.utils.info_utils.NonMatchingChecksumError: Checksums didn't match for dataset source files
| 2,646 |
https://github.com/huggingface/datasets/issues/2645 | load_dataset processing failed with OS error after downloading a dataset | [
"Hi ! It looks like an issue with pytorch.\r\n\r\nCould you try to run `import torch` and see if it raises an error ?",
"> Hi ! It looks like an issue with pytorch.\r\n> \r\n> Could you try to run `import torch` and see if it raises an error ?\r\n\r\nIt works. Thank you!"
] | ## Describe the bug
After downloading a dataset like opus100, there is a bug that
OSError: Cannot find data file.
Original error:
dlopen: cannot load any more object with static TLS
## Steps to reproduce the bug
```python
from datasets import load_dataset
this_dataset = load_dataset('opus100', 'af-en')
```
## Expected results
there is no error when running load_dataset.
## Actual results
Specify the actual results or traceback.
Traceback (most recent call last):
File "/home/anaconda3/lib/python3.6/site-packages/datasets/builder.py", line 652, in _download_and_prep
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/home/anaconda3/lib/python3.6/site-packages/datasets/builder.py", line 989, in _prepare_split
example = self.info.features.encode_example(record)
File "/home/anaconda3/lib/python3.6/site-packages/datasets/features.py", line 952, in encode_example
example = cast_to_python_objects(example)
File "/home/anaconda3/lib/python3.6/site-packages/datasets/features.py", line 219, in cast_to_python_ob
return _cast_to_python_objects(obj)[0]
File "/home/anaconda3/lib/python3.6/site-packages/datasets/features.py", line 165, in _cast_to_python_o
import torch
File "/home/anaconda3/lib/python3.6/site-packages/torch/__init__.py", line 188, in <module>
_load_global_deps()
File "/home/anaconda3/lib/python3.6/site-packages/torch/__init__.py", line 141, in _load_global_deps
ctypes.CDLL(lib_path, mode=ctypes.RTLD_GLOBAL)
File "/home/anaconda3/lib/python3.6/ctypes/__init__.py", line 348, in __init__
self._handle = _dlopen(self._name, mode)
OSError: dlopen: cannot load any more object with static TLS
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "download_hub_opus100.py", line 9, in <module>
this_dataset = load_dataset('opus100', language_pair)
File "/home/anaconda3/lib/python3.6/site-packages/datasets/load.py", line 748, in load_dataset
use_auth_token=use_auth_token,
File "/home/anaconda3/lib/python3.6/site-packages/datasets/builder.py", line 575, in download_and_prepa
dl_manager=dl_manager, verify_infos=verify_infos, **download_and_prepare_kwargs
File "/home/anaconda3/lib/python3.6/site-packages/datasets/builder.py", line 658, in _download_and_prep
+ str(e)
OSError: Cannot find data file.
Original error:
dlopen: cannot load any more object with static TLS
## Environment info
- `datasets` version: 1.8.0
- Platform: Linux-3.13.0-32-generic-x86_64-with-debian-jessie-sid
- Python version: 3.6.6
- PyArrow version: 3.0.0
| 2,645 |
https://github.com/huggingface/datasets/issues/2644 | Batched `map` not allowed to return 0 items | [
"Hi ! Thanks for reporting. Indeed it looks like type inference makes it fail. We should probably just ignore this step until a non-empty batch is passed.",
"Sounds good! Do you want me to propose a PR? I'm quite busy right now, but if it's not too urgent I could take a look next week.",
"Sure if you're interested feel free to open a PR :)\r\n\r\nYou can also ping me anytime if you have questions or if I can help !",
"Sorry to ping you, @lhoestq, did you have a chance to take a look at the proposed PR? Thank you!",
"Yes and it's all good, thank you :)\r\n\r\nFeel free to close this issue if it's good for you",
"Everything's good, thanks!"
] | ## Describe the bug
I'm trying to use `map` to filter a large dataset by selecting rows that match an expensive condition (files referenced by one of the columns need to exist in the filesystem, so we have to `stat` them). According to [the documentation](https://huggingface.co/docs/datasets/processing.html#augmenting-the-dataset), `a batch mapped function can take as input a batch of size N and return a batch of size M where M can be greater or less than N and can even be zero`.
However, when the returned batch has a size of zero (neither item in the batch fulfilled the condition), we get an `index out of bounds` error. I think that `arrow_writer.py` is [trying to infer the returned types using the first element returned](https://github.com/huggingface/datasets/blob/master/src/datasets/arrow_writer.py#L100), but no elements were returned in this case.
For this error to happen, I'm returning a dictionary that contains empty lists for the keys I want to keep, see below. If I return an empty dictionary instead (no keys), then a different error eventually occurs.
## Steps to reproduce the bug
```python
def select_rows(examples):
# `key` is a column name that exists in the original dataset
# The following line simulates no matches found, so we return an empty batch
result = {'key': []}
return result
filtered_dataset = dataset.map(
select_rows,
remove_columns = dataset.column_names,
batched = True,
num_proc = 1,
desc = "Selecting rows with images that exist"
)
```
The code above immediately triggers the exception. If we use the following instead:
```python
def select_rows(examples):
# `key` is a column name that exists in the original dataset
result = {'key': []} # or defaultdict or whatever
# code to check for condition and append elements to result
# some_items_found will be set to True if there were any matching elements in the batch
return result if some_items_found else {}
```
Then it _seems_ to work, but it eventually fails with some sort of schema error. I believe it may happen when an empty batch is followed by a non-empty one, but haven't set up a test to verify it.
In my opinion, returning a dictionary with empty lists and valid column names should be accepted as a valid result with zero items.
## Expected results
The dataset would be filtered and only the matching fields would be returned.
## Actual results
An exception is encountered, as described. Using a workaround makes it fail further along the line.
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.9.1.dev0
- Platform: Linux-5.4.0-53-generic-x86_64-with-glibc2.17
- Python version: 3.8.10
- PyArrow version: 4.0.1
| 2,644 |
https://github.com/huggingface/datasets/issues/2643 | Enum used in map functions will raise a RecursionError with dill. | [
"I'm running into this as well. (Thank you so much for reporting @jorgeecardona — was staring at this massive stack trace and unsure what exactly was wrong!)",
"Hi ! Thanks for reporting :)\r\n\r\nUntil this is fixed on `dill`'s side, we could implement a custom saving in our Pickler indefined in utils.py_utils.py\r\nThere is already a suggestion in this message about how to do it:\r\nhttps://github.com/uqfoundation/dill/issues/250#issuecomment-852566284\r\n\r\nLet me know if such a workaround could help, and feel free to open a PR if you want to contribute !",
"I have the same bug.\r\nthe code is as follows:\r\n\r\nthe error is: \r\n\r\n\r\nLook for the solution for this bug.",
"Hi ! I think your RecursionError comes from a different issue @BitcoinNLPer , could you open a separate issue please ?\r\n\r\nAlso which dataset are you using ? I tried loading `CodedotAI/code_clippy` but I get a different error\r\n```python\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/Users/quentinlhoest/Desktop/hf/datasets/src/datasets/load.py\", line 1615, in load_dataset\r\n **config_kwargs,\r\n File \"/Users/quentinlhoest/Desktop/hf/datasets/src/datasets/load.py\", line 1446, in load_dataset_builder\r\n builder_cls = import_main_class(dataset_module.module_path)\r\n File \"/Users/quentinlhoest/Desktop/hf/datasets/src/datasets/load.py\", line 101, in import_main_class\r\n module = importlib.import_module(module_path)\r\n File \"/Users/quentinlhoest/.virtualenvs/hf-datasets/lib/python3.7/importlib/__init__.py\", line 127, in import_module\r\n return _bootstrap._gcd_import(name[level:], package, level)\r\n File \"<frozen importlib._bootstrap>\", line 1006, in _gcd_import\r\n File \"<frozen importlib._bootstrap>\", line 983, in _find_and_load\r\n File \"<frozen importlib._bootstrap>\", line 967, in _find_and_load_unlocked\r\n File \"<frozen importlib._bootstrap>\", line 677, in _load_unlocked\r\n File \"<frozen importlib._bootstrap_external>\", line 728, in exec_module\r\n File \"<frozen importlib._bootstrap>\", line 219, in _call_with_frames_removed\r\n File \"/Users/quentinlhoest/.cache/huggingface/modules/datasets_modules/datasets/CodedotAI___code_clippy/d332f69d036e8c80f47bc9a96d676c3fa30cb50af7bb81e2d4d12e80b83efc4d/code_clippy.py\", line 66, in <module>\r\n url_elements = results.find_all(\"a\")\r\nAttributeError: 'NoneType' object has no attribute 'find_all'\r\n```"
] | ## Describe the bug
Enums used in functions pass to `map` will fail at pickling with a maximum recursion exception as described here: https://github.com/uqfoundation/dill/issues/250#issuecomment-852566284
In my particular case, I use an enum to define an argument with fixed options using the `TraininigArguments` dataclass as base class and the `HfArgumentParser`. In the same file I use a `ds.map` that tries to pickle the content of the module including the definition of the enum that runs into the dill bug described above.
## Steps to reproduce the bug
```python
from datasets import load_dataset
from enum import Enum
class A(Enum):
a = 'a'
def main():
a = A.a
def f(x):
return {} if a == a.a else x
ds = load_dataset('cnn_dailymail', '3.0.0')['test']
ds = ds.map(f, num_proc=15)
if __name__ == "__main__":
main()
```
## Expected results
The known problem with dill could be prevented as explained in the link above (workaround.) Since `HFArgumentParser` nicely uses the enum class for choices it makes sense to also deal with this bug under the hood.
## Actual results
```python
File "/home/xxxx/miniconda3/lib/python3.8/site-packages/dill/_dill.py", line 1373, in save_type
pickler.save_reduce(_create_type, (type(obj), obj.__name__,
File "/home/xxxx/miniconda3/lib/python3.8/pickle.py", line 690, in save_reduce
save(args)
File "/home/xxxx/miniconda3/lib/python3.8/pickle.py", line 558, in save
f(self, obj) # Call unbound method with explicit self
File "/home/xxxx/miniconda3/lib/python3.8/pickle.py", line 899, in save_tuple
save(element)
File "/home/xxxx/miniconda3/lib/python3.8/pickle.py", line 534, in save
self.framer.commit_frame()
File "/home/xxxx/miniconda3/lib/python3.8/pickle.py", line 220, in commit_frame
if f.tell() >= self._FRAME_SIZE_TARGET or force:
RecursionError: maximum recursion depth exceeded while calling a Python object
```
## Environment info
- `datasets` version: 1.8.0
- Platform: Linux-5.9.0-4-amd64-x86_64-with-glibc2.10
- Python version: 3.8.5
- PyArrow version: 3.0.0
| 2,643 |
https://github.com/huggingface/datasets/issues/2642 | Support multi-worker with streaming dataset (IterableDataset). | [
"Hi ! This is a great idea :)\r\nI think we could have something similar to what we have in `datasets.Dataset.map`, i.e. a `num_proc` parameter that tells how many processes to spawn to parallelize the data processing. \r\n\r\nRegarding AUTOTUNE, this could be a nice feature as well, we could see how to add it in a second step",
"Any update on this feature request?",
"Not yet, I'm happy to provide some guidance if someone wants to give it a shot though.\r\n\r\nThe code that applies the `map` function is in `iterable_dataset.py`, in `MappedExamplesIterable.__iter__`"
] | **Is your feature request related to a problem? Please describe.**
The current `.map` does not support multi-process, CPU can become bottleneck if the pre-processing is complex (e.g. t5 span masking).
**Describe the solution you'd like**
Ideally `.map` should support multi-worker like tfds, with `AUTOTUNE`.
**Describe alternatives you've considered**
A simpler solution is to shard the dataset and process it in parallel with pytorch dataloader. The shard does not need to be of equal size.
* https://pytorch.org/docs/stable/data.html#torch.utils.data.IterableDataset
**Additional context**
| 2,642 |
https://github.com/huggingface/datasets/issues/2641 | load_dataset("financial_phrasebank") NonMatchingChecksumError | [
"Hi! It's probably because this dataset is stored on google drive and it has a per day quota limit. It should work if you retry, I was able to initiate the download.\r\n\r\nSimilar issue [here](https://github.com/huggingface/datasets/issues/2646)",
"Hi ! Loading the dataset works on my side as well.\r\nFeel free to try again and let us know if it works for you know",
"Thank you! I've been trying periodically for the past month, and no luck yet with this particular dataset. Just tried again and still hitting the checksum error.\r\n\r\nCode:\r\n\r\n`dataset = load_dataset(\"financial_phrasebank\", \"sentences_allagree\") `\r\n\r\nTraceback:\r\n\r\n```\r\n---------------------------------------------------------------------------\r\nNonMatchingChecksumError Traceback (most recent call last)\r\n<ipython-input-2-55cc2144f31e> in <module>\r\n----> 1 dataset = load_dataset(\"financial_phrasebank\", \"sentences_allagree\")\r\n\r\n/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, script_version, use_auth_token, task, streaming, **config_kwargs)\r\n 859 ignore_verifications=ignore_verifications,\r\n 860 try_from_hf_gcs=try_from_hf_gcs,\r\n--> 861 use_auth_token=use_auth_token,\r\n 862 )\r\n 863 \r\n\r\n/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)\r\n 582 if not downloaded_from_gcs:\r\n 583 self._download_and_prepare(\r\n--> 584 dl_manager=dl_manager, verify_infos=verify_infos, **download_and_prepare_kwargs\r\n 585 )\r\n 586 # Sync info\r\n\r\n/opt/conda/lib/python3.7/site-packages/datasets/builder.py in _download_and_prepare(self, dl_manager, verify_infos, **prepare_split_kwargs)\r\n 642 if verify_infos:\r\n 643 verify_checksums(\r\n--> 644 self.info.download_checksums, dl_manager.get_recorded_sizes_checksums(), \"dataset source files\"\r\n 645 )\r\n 646 \r\n\r\n/opt/conda/lib/python3.7/site-packages/datasets/utils/info_utils.py in verify_checksums(expected_checksums, recorded_checksums, verification_name)\r\n 38 if len(bad_urls) > 0:\r\n 39 error_msg = \"Checksums didn't match\" + for_verification_name + \":\\n\"\r\n---> 40 raise NonMatchingChecksumError(error_msg + str(bad_urls))\r\n 41 logger.info(\"All the checksums matched successfully\" + for_verification_name)\r\n 42 \r\n\r\nNonMatchingChecksumError: Checksums didn't match for dataset source files:\r\n['https://www.researchgate.net/profile/Pekka_Malo/publication/251231364_FinancialPhraseBank-v10/data/0c96051eee4fb1d56e000000/FinancialPhraseBank-v10.zip']\r\n```",
"Fixed once data files are hosted on the Hub:\r\n- #4598"
] | ## Describe the bug
Attempting to download the financial_phrasebank dataset results in a NonMatchingChecksumError
## Steps to reproduce the bug
```python
from datasets import load_dataset
dataset = load_dataset("financial_phrasebank", 'sentences_allagree')
```
## Expected results
I expect to see the financial_phrasebank dataset downloaded successfully
## Actual results
NonMatchingChecksumError: Checksums didn't match for dataset source files:
['https://www.researchgate.net/profile/Pekka_Malo/publication/251231364_FinancialPhraseBank-v10/data/0c96051eee4fb1d56e000000/FinancialPhraseBank-v10.zip']
## Environment info
- `datasets` version: 1.9.0
- Platform: Linux-4.14.232-177.418.amzn2.x86_64-x86_64-with-debian-10.6
- Python version: 3.7.10
- PyArrow version: 4.0.1
| 2,641 |
https://github.com/huggingface/datasets/issues/2630 | Progress bars are not properly rendered in Jupyter notebook | [
"To add my experience when trying to debug this issue:\r\n\r\nSeems like previously the workaround given [here](https://github.com/tqdm/tqdm/issues/485#issuecomment-473338308) worked around this issue. But with the latest version of jupyter/tqdm I still get terminal warnings that IPython tried to send a message from a forked process.",
"Hi @mludv, thanks for the hint!!! :) \r\n\r\nWe will definitely take it into account to try to fix this issue... It seems somehow related to `multiprocessing` and `tqdm`..."
] | ## Describe the bug
The progress bars are not Jupyter widgets; regular progress bars appear (like in a terminal).
## Steps to reproduce the bug
```python
ds.map(tokenize, num_proc=10)
```
## Expected results
Jupyter widgets displaying the progress bars.
## Actual results
Simple plane progress bars.
cc: Reported by @thomwolf | 2,630 |
https://github.com/huggingface/datasets/issues/2629 | Load datasets from the Hub without requiring a dataset script | [
"This is so cool, let us know if we can help with anything on the hub side (@Pierrci @elishowk) 🎉 "
] | As a user I would like to be able to upload my csv/json/text/parquet/etc. files in a dataset repository on the Hugging Face Hub and be able to load this dataset with `load_dataset` without having to implement a dataset script.
Moreover I would like to be able to specify which file goes into which split using the `data_files` argument.
This feature should be compatible with private repositories and dataset streaming.
This can be implemented by checking the extension of the files in the dataset repository and then by using the right dataset builder that is already packaged in the library (csv/json/text/parquet/etc.) | 2,629 |
https://github.com/huggingface/datasets/issues/2625 | ⚛️😇⚙️🔑 | [] | 2,625 |
|
https://github.com/huggingface/datasets/issues/2624 | can't set verbosity for `metric.py` | [
"Thanks @thomas-happify for reporting and thanks @mariosasko for the fix."
] | ## Describe the bug
```
[2021-07-10 20:13:11,528][datasets.utils.filelock][INFO] - Lock 139705371374976 acquired on /root/.cache/huggingface/metrics/seqeval/default/default_experiment-1-0.arrow.lock
[2021-07-10 20:13:11,529][datasets.arrow_writer][INFO] - Done writing 32 examples in 6100 bytes /root/.cache/huggingface/metrics/seqeval/default/default_experiment-1-0.arrow.
[2021-07-10 20:13:11,531][datasets.arrow_dataset][INFO] - Set __getitem__(key) output type to python objects for no columns (when key is int or slice) and don't output other (un-formatted) columns.
[2021-07-10 20:13:11,543][/conda/envs/myenv/lib/python3.8/site-packages/datasets/metric.py][INFO] - Removing /root/.cache/huggingface/metrics/seqeval/default/default_experiment-1-0.arrow
```
As you can see, `datasets` logging come from different places.
`filelock`, `arrow_writer` & `arrow_dataset` comes from `datasets.*` which are expected
However, `metric.py` logging comes from `/conda/envs/myenv/lib/python3.8/site-packages/datasets/`
So when setting `datasets.utils.logging.set_verbosity_error()`, it still logs the last message which is annoying during evaluation.
I had to do
```
logging.getLogger("/conda/envs/myenv/lib/python3.8/site-packages/datasets/metric").setLevel(logging.ERROR)
```
to fully mute these messages
## Expected results
it shouldn't log these messages when setting `datasets.utils.logging.set_verbosity_error()`
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: tried both 1.8.0 & 1.9.0
- Platform: Ubuntu 18.04.5 LTS
- Python version: 3.8.10
- PyArrow version: 3.0.0
| 2,624 |
https://github.com/huggingface/datasets/issues/2622 | Integration with AugLy | [
"Hi,\r\n\r\nyou can define your own custom formatting with `Dataset.set_transform()` and then run the tokenizer with the batches of augmented data as follows:\r\n```python\r\ndset = load_dataset(\"imdb\", split=\"train\") # Let's say we are working with the IMDB dataset\r\ndset.set_transform(lambda ex: {\"text\": augly_text_augmentation(ex[\"text\"])}, columns=\"text\", output_all_columns=True)\r\ndataloader = torch.utils.data.DataLoader(dset, batch_size=32)\r\nfor epoch in range(5):\r\n for batch in dataloader:\r\n tokenizer_output = tokenizer(batch.pop(\"text\"), padding=True, truncation=True, return_tensors=\"pt\")\r\n batch.update(tokenizer_output)\r\n output = model(**batch)\r\n ...\r\n```",
"Preprocessing functions/augmentations, unless super generic, should be defined in separate libraries, so I'm closing this issue."
] | **Is your feature request related to a problem? Please describe.**
Facebook recently launched a library, [AugLy](https://github.com/facebookresearch/AugLy) , that has a unified API for augmentations for image, video and text.
It would be pretty exciting to have it hooked up to HF libraries so that we can make NLP models robust to misspellings or to punctuation, or emojis etc. Plus, with Transformers supporting more CV use cases, having augmentations support becomes crucial.
**Describe the solution you'd like**
The biggest difference between augmentations and preprocessing is that preprocessing happens only once, but you are running augmentations once per epoch. AugLy operates on text directly, so this breaks the typical workflow where we would run the tokenizer once, set format to pt tensors and be ready for the Dataloader.
**Describe alternatives you've considered**
One possible way of implementing these is to make a custom Dataset class where getitem(i) runs the augmentation and the tokenizer every time, though this would slow training down considerably given we wouldn't even run the tokenizer in batches.
| 2,622 |
https://github.com/huggingface/datasets/issues/2618 | `filelock.py` Error | [
"Hi @liyucheng09, thanks for reporting.\r\n\r\nApparently this issue has to do with your environment setup. One question: is your data in an NFS share? Some people have reported this error when using `fcntl` to write to an NFS share... If this is the case, then it might be that your NFS just may not be set up to provide file locks. You should ask your system administrator, or try these commands in the terminal:\r\n```shell\r\nsudo systemctl enable rpc-statd\r\nsudo systemctl start rpc-statd\r\n```",
"We now use the `filelock` package for file locking, so please report this issue in their repo if it's still causing you problems."
] | ## Describe the bug
It seems that the `filelock.py` went error.
```
>>> ds=load_dataset('xsum')
^CTraceback (most recent call last):
File "/user/HS502/yl02706/.conda/envs/lyc/lib/python3.6/site-packages/datasets/utils/filelock.py", line 402, in _acquire
fcntl.flock(fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
OSError: [Errno 37] No locks available
```
According to error log, it is OSError, but there is an `except` in the `_acquire` function.
```
def _acquire(self):
open_mode = os.O_WRONLY | os.O_CREAT | os.O_EXCL | os.O_TRUNC
try:
fd = os.open(self._lock_file, open_mode)
except (IOError, OSError):
pass
else:
self._lock_file_fd = fd
return None
```
I don't know why it stucked rather than `pass` directly.
I am not quite familiar with filelock operation, so any help is highly appriciated.
## Steps to reproduce the bug
```python
ds = load_dataset('xsum')
```
## Expected results
A clear and concise description of the expected results.
## Actual results
```
>>> ds=load_dataset('xsum')
^CTraceback (most recent call last):
File "/user/HS502/yl02706/.conda/envs/lyc/lib/python3.6/site-packages/datasets/utils/filelock.py", line 402, in _acquire
fcntl.flock(fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
OSError: [Errno 37] No locks available
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/user/HS502/yl02706/.conda/envs/lyc/lib/python3.6/site-packages/datasets/load.py", line 818, in load_dataset
use_auth_token=use_auth_token,
File "/user/HS502/yl02706/.conda/envs/lyc/lib/python3.6/site-packages/datasets/load.py", line 470, in prepare_module
with FileLock(lock_path):
File "/user/HS502/yl02706/.conda/envs/lyc/lib/python3.6/site-packages/datasets/utils/filelock.py", line 323, in __enter__
self.acquire()
File "/user/HS502/yl02706/.conda/envs/lyc/lib/python3.6/site-packages/datasets/utils/filelock.py", line 272, in acquire
self._acquire()
File "/user/HS502/yl02706/.conda/envs/lyc/lib/python3.6/site-packages/datasets/utils/filelock.py", line 402, in _acquire
fcntl.flock(fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
KeyboardInterrupt
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.9.0
- Platform: Linux-4.15.0-135-generic-x86_64-with-debian-buster-sid
- Python version: 3.6.13
- PyArrow version: 4.0.1
| 2,618 |
Subsets and Splits