html_url stringlengths 48 51 | title stringlengths 1 290 | comments listlengths 0 30 | body stringlengths 0 228k ⌀ | number int64 2 7.08k |
|---|---|---|---|---|
https://github.com/huggingface/datasets/issues/5265 | Get an IterableDataset from a map-style Dataset | [
"I think `stream` could be misleading since the data is not being streamed from remote endpoints (one could think that's the case when they see `load_dataset` followed by `stream`). Hence, I prefer the second option.\r\n\r\nPS: When we resolve https://github.com/huggingface/datasets/issues/4542, we could add `as_tf... | This is useful to leverage iterable datasets specific features like:
- fast approximate shuffling
- lazy map, filter etc.
Iterating over the resulting iterable dataset should be at least as fast at iterating over the map-style dataset.
Here are some ideas regarding the API:
```python
# 1.
# - consistency with load_dataset(..., streaming=True)
# - gives intuition that map/filter/etc. are done on-the-fly
ids = ds.stream()
# 2.
# - more explicit on the output type
# - but maybe sounds like a conversion tool rather than a step in a processing pipeline
ids = ds.as_iterable_dataset()
``` | 5,265 |
https://github.com/huggingface/datasets/issues/5264 | `datasets` can't read a Parquet file in Python 3.9.13 | [
"Could you share the full stack trace please ?\r\n\r\n\r\nCan you also try running this code ? It can be useful to determine if the issue comes from `datasets` or `fsspec` (streaming) or `pyarrow` (parquet reading):\r\n```python\r\nds = load_dataset(\"parquet\", data_files=a_parquet_file_url, use_auth_token=True)\r... | ### Describe the bug
I have an error when trying to load this [dataset](https://huggingface.co/datasets/bigcode/the-stack-dedup-pjj) (it's private but I can add you to the bigcode org). `datasets` can't read one of the parquet files in the Java subset
```python
from datasets import load_dataset
ds = load_dataset("bigcode/the-stack-dedup-pjj", data_dir="data/java", split="train", revision="v1.1.a1", use_auth_token=True)
````
```
File "pyarrow/error.pxi", line 100, in pyarrow.lib.check_status
pyarrow.lib.ArrowInvalid: Parquet magic bytes not found in footer. Either the file is corrupted or this is not a parquet file.
```
It seems to be an issue with new Python versions, Because it works in these two environements:
```
- `datasets` version: 2.6.1
- Platform: Linux-5.4.0-131-generic-x86_64-with-glibc2.31
- Python version: 3.9.7
- PyArrow version: 9.0.0
- Pandas version: 1.3.4
```
```
- `datasets` version: 2.6.1
- Platform: Linux-4.19.0-22-cloud-amd64-x86_64-with-debian-10.13
- Python version: 3.7.12
- PyArrow version: 9.0.0
- Pandas version: 1.3.4
```
But not in this:
```
- `datasets` version: 2.6.1
- Platform: Linux-4.19.0-22-cloud-amd64-x86_64-with-glibc2.28
- Python version: 3.9.13
- PyArrow version: 9.0.0
- Pandas version: 1.3.4
```
### Steps to reproduce the bug
Load the dataset in python 3.9.13
### Expected behavior
Load the dataset without the pyarrow error.
### Environment info
```
- `datasets` version: 2.6.1
- Platform: Linux-4.19.0-22-cloud-amd64-x86_64-with-glibc2.28
- Python version: 3.9.13
- PyArrow version: 9.0.0
- Pandas version: 1.3.4
``` | 5,264 |
https://github.com/huggingface/datasets/issues/5263 | Save a dataset in a determined number of shards | [] | This is useful to distribute the shards to training nodes.
This can be implemented in `save_to_disk` and can also leverage multiprocessing to speed up the process | 5,263 |
https://github.com/huggingface/datasets/issues/5262 | AttributeError: 'Value' object has no attribute 'names' | [
"Hi ! It looks like your \"isDif\" column is a Sequence of Value(\"string\"), not a Sequence of ClassLabel.\r\n\r\nYou can convert your Value(\"string\") feature type to a ClassLabel feature type this way:\r\n```python\r\nfrom datasets import ClassLabel, Sequence\r\n\r\n# provide the label_names yourself\r\nlabel_n... | Hello
I'm trying to build a model for custom token classification
I already followed the token classification course on huggingface
while adapting the code to my work, this message occures :
'Value' object has no attribute 'names'
Here's my code:
`raw_datasets`
generates
DatasetDict({
train: Dataset({
features: ['isDisf', 'pos', 'tokens', 'id'],
num_rows: 14
})
})
`raw_datasets["train"][3]["isDisf"]`
generates
['B_RM', 'I_RM', 'I_RM', 'B_RP', 'I_RP', 'O', 'O']
`dis_feature = raw_datasets["train"].features["isDisf"]
dis_feature`
generates
Sequence(feature=Value(dtype='string', id=None), length=-1, id=None)
and
`label_names = dis_feature.feature.names
label_names`
generates
AttributeError Traceback (most recent call last)
[<ipython-input-28-972fd54a869a>](https://localhost:8080/#) in <module>
----> 1 label_names = dis_feature.feature.names
2 label_names
AttributeError: 'Value' object has
AttributeError: 'Value' object has no attribute 'names'
Thank you for your help | 5,262 |
https://github.com/huggingface/datasets/issues/5261 | Add PubTables-1M | [
"cc @albertvillanova the author would like to add this dataset to the hub: https://github.com/microsoft/table-transformer/issues/68#issuecomment-1319114621. Could you help him out?"
] | ### Name
PubTables-1M
### Paper
https://openaccess.thecvf.com/content/CVPR2022/html/Smock_PubTables-1M_Towards_Comprehensive_Table_Extraction_From_Unstructured_Documents_CVPR_2022_paper.html
### Data
https://github.com/microsoft/table-transformer
### Motivation
Table Transformer is now available in 🤗 Transformer, and it was trained on PubTables-1M. It's a large dataset for table extraction and structure recognition in unstructured documents. | 5,261 |
https://github.com/huggingface/datasets/issues/5260 | consumer-finance-complaints dataset not loading | [
"Thanks for reporting, @adiprasad.\r\n\r\nWe are having a look at it.",
"I have opened an issue in that dataset Community tab on the Hub: https://huggingface.co/datasets/consumer-finance-complaints/discussions/1\r\n\r\nPlease note that in the meantime, you can load the dataset by passing `ignore_verifications=Tru... | ### Describe the bug
Error during dataset loading
### Steps to reproduce the bug
```
>>> import datasets
>>> cf_raw = datasets.load_dataset("consumer-finance-complaints")
Downloading builder script: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 8.42k/8.42k [00:00<00:00, 3.33MB/s]
Downloading metadata: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5.60k/5.60k [00:00<00:00, 2.90MB/s]
Downloading readme: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 16.6k/16.6k [00:00<00:00, 510kB/s]
Downloading and preparing dataset consumer-finance-complaints/default to /root/.cache/huggingface/datasets/consumer-finance-complaints/default/0.0.0/30e483d37fb4b25bb98cad1bfd2dc48f6ed6d1f3371eb4568c625a61d1a79b69...
Downloading data: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 511M/511M [00:04<00:00, 103MB/s]
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/skunk-pod-storage-lee-2emartie-40ibm-2ecom-pvc/anaconda3/envs/datasets/lib/python3.8/site-packages/datasets/load.py", line 1741, in load_dataset
builder_instance.download_and_prepare(
File "/skunk-pod-storage-lee-2emartie-40ibm-2ecom-pvc/anaconda3/envs/datasets/lib/python3.8/site-packages/datasets/builder.py", line 822, in download_and_prepare
self._download_and_prepare(
File "/skunk-pod-storage-lee-2emartie-40ibm-2ecom-pvc/anaconda3/envs/datasets/lib/python3.8/site-packages/datasets/builder.py", line 1555, in _download_and_prepare
super()._download_and_prepare(
File "/skunk-pod-storage-lee-2emartie-40ibm-2ecom-pvc/anaconda3/envs/datasets/lib/python3.8/site-packages/datasets/builder.py", line 931, in _download_and_prepare
verify_splits(self.info.splits, split_dict)
File "/skunk-pod-storage-lee-2emartie-40ibm-2ecom-pvc/anaconda3/envs/datasets/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=1605177353, num_examples=2455765, shard_lengths=None, dataset_name=None), 'recorded': SplitInfo(name='train', num_bytes=2043641693, num_examples=3079747, shard_lengths=[721000, 656000, 788000, 846000, 68747], dataset_name='consumer-finance-complaints')}]
```
### Expected behavior
dataset should load
### Environment info
>>> datasets.__version__
'2.7.0'
Python 3.8.10
"Ubuntu 20.04.4 LTS" | 5,260 |
https://github.com/huggingface/datasets/issues/5259 | datasets 2.7 introduces sharding error | [
"I notice a comment in the code says:\r\n`Having lists of different sizes makes sharding ambigious, raise an error in this case until we decide how to define sharding without ambiguity for users` \r\n \r\n ... which suggests this update was pushed knowing that it might break some things. But, it didn't seem to h... | ### Describe the bug
dataset fails to load with runtime error
`RuntimeError: Sharding is ambiguous for this dataset: we found several data sources lists of different lengths, and we don't know over which list we should parallelize:
- key audio_files has length 46
- key data has length 0
To fix this, check the 'gen_kwargs' and make sure to use lists only for data sources, and use tuples otherwise. In the end there should only be one single list, or several lists with the same length.`
### Steps to reproduce the bug
With datasets[audio] 2.7 loaded, and logged into hugging face,
`data = datasets.load_dataset('sil-ai/bloom-speech', 'bis', use_auth_token=True)`
creates the error.
Full stack trace:
```---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
[<ipython-input-7-8cb9ca0f79f0>](https://localhost:8080/#) in <module>
----> 1 data = datasets.load_dataset('sil-ai/bloom-speech', 'bis', use_auth_token=True)
5 frames
[/usr/local/lib/python3.7/dist-packages/datasets/load.py](https://localhost:8080/#) in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, ignore_verifications, keep_in_memory, save_infos, revision, use_auth_token, task, streaming, num_proc, **config_kwargs)
1745 try_from_hf_gcs=try_from_hf_gcs,
1746 use_auth_token=use_auth_token,
-> 1747 num_proc=num_proc,
1748 )
1749
[/usr/local/lib/python3.7/dist-packages/datasets/builder.py](https://localhost:8080/#) in download_and_prepare(self, output_dir, download_config, download_mode, ignore_verifications, try_from_hf_gcs, dl_manager, base_path, use_auth_token, file_format, max_shard_size, num_proc, storage_options, **download_and_prepare_kwargs)
824 verify_infos=verify_infos,
825 **prepare_split_kwargs,
--> 826 **download_and_prepare_kwargs,
827 )
828 # Sync info
[/usr/local/lib/python3.7/dist-packages/datasets/builder.py](https://localhost:8080/#) in _download_and_prepare(self, dl_manager, verify_infos, **prepare_splits_kwargs)
1554 def _download_and_prepare(self, dl_manager, verify_infos, **prepare_splits_kwargs):
1555 super()._download_and_prepare(
-> 1556 dl_manager, verify_infos, check_duplicate_keys=verify_infos, **prepare_splits_kwargs
1557 )
1558
[/usr/local/lib/python3.7/dist-packages/datasets/builder.py](https://localhost:8080/#) in _download_and_prepare(self, dl_manager, verify_infos, **prepare_split_kwargs)
911 try:
912 # Prepare split will record examples associated to the split
--> 913 self._prepare_split(split_generator, **prepare_split_kwargs)
914 except OSError as e:
915 raise OSError(
[/usr/local/lib/python3.7/dist-packages/datasets/builder.py](https://localhost:8080/#) in _prepare_split(self, split_generator, check_duplicate_keys, file_format, num_proc, max_shard_size)
1362 fpath = path_join(self._output_dir, fname)
1363
-> 1364 num_input_shards = _number_of_shards_in_gen_kwargs(split_generator.gen_kwargs)
1365 if num_input_shards <= 1 and num_proc is not None:
1366 logger.warning(
[/usr/local/lib/python3.7/dist-packages/datasets/utils/sharding.py](https://localhost:8080/#) in _number_of_shards_in_gen_kwargs(gen_kwargs)
16 + "\n".join(f"\t- key {key} has length {length}" for key, length in lists_lengths.items())
17 + "\nTo fix this, check the 'gen_kwargs' and make sure to use lists only for data sources, "
---> 18 + "and use tuples otherwise. In the end there should only be one single list, or several lists with the same length."
19 )
20 )
RuntimeError: Sharding is ambiguous for this dataset: we found several data sources lists of different lengths, and we don't know over which list we should parallelize:
- key audio_files has length 46
- key data has length 0
To fix this, check the 'gen_kwargs' and make sure to use lists only for data sources, and use tuples otherwise. In the end there should only be one single list, or several lists with the same length.```
### Expected behavior
the dataset loads in datasets version 2.6.1 and should load with datasets 2.7
### Environment info
- `datasets` version: 2.7.0
- Platform: Linux-5.10.133+-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.15
- PyArrow version: 6.0.1
- Pandas version: 1.3.5 | 5,259 |
https://github.com/huggingface/datasets/issues/5258 | Restore order of split names in dataset_info for canonical datasets | [
"The bulk edit is running...\r\n\r\nSee for example: \r\n- A single config: https://huggingface.co/datasets/acronym_identification/discussions/2\r\n- Multiple configs: https://huggingface.co/datasets/babi_qa/discussions/1",
"TODO: Add \"dataset_info\" YAML metadata to:\r\n- [x] \"chr_en\" has no metadata JSON fil... | After a bulk edit of canonical datasets to create the YAML `dataset_info` metadata, the split names were accidentally sorted alphabetically. See for example:
- https://huggingface.co/datasets/bc2gm_corpus/commit/2384629484401ecf4bb77cd808816719c424e57c
Note that this order is the one appearing in the preview of the datasets.
I'm making a bulk edit to align the order of the splits appearing in the metadata info with the order appearing in the loading script.
Related to:
- #5202 | 5,258 |
https://github.com/huggingface/datasets/issues/5255 | Add a Depth Estimation dataset - DIODE / NYUDepth / KITTI | [
"Also cc @mariosasko and @lhoestq ",
"Cool ! Let us know if you have questions or if we can help :)\r\n\r\nI guess we'll also have to create the NYU CS Department on the Hub ?",
"> I guess we'll also have to create the NYU CS Department on the Hub ?\r\n\r\nYes, you're right! Let me add it to my profile first, a... | ### Name
NYUDepth
### Paper
http://cs.nyu.edu/~silberman/papers/indoor_seg_support.pdf
### Data
https://cs.nyu.edu/~silberman/datasets/nyu_depth_v2.html
### Motivation
Depth estimation is an important problem in computer vision. We have a couple of Depth Estimation models on Hub as well:
* [GLPN](https://huggingface.co/docs/transformers/model_doc/glpn)
* [DPT](https://huggingface.co/docs/transformers/model_doc/dpt)
Would be nice to have a dataset for depth estimation. These datasets usually have three things: input image, depth map image, and depth mask (validity mask to indicate if a reading for a pixel is valid or not). Since we already have [semantic segmentation datasets on the Hub](https://huggingface.co/datasets?task_categories=task_categories:image-segmentation&sort=downloads), I don't think we need any extended utilities to support this addition.
Having this dataset would also allow us to author data preprocessing guides for depth estimation, particularly like the ones we have for other tasks ([example](https://huggingface.co/docs/datasets/image_classification)).
Ccing @osanseviero @nateraw @NielsRogge
Happy to work on adding it. | 5,255 |
https://github.com/huggingface/datasets/issues/5251 | Docs are not generated after latest release | [
"After a discussion with @mishig25:\r\n- He said that this action should be triggered if we call our release branch according to the regex `v*-release`, as transformers does\r\n- I said that our procedure is different: our release branch is *temporary* and it is deleted just after the release PR is merged to main\r... | After the latest `datasets` release version 0.7.0, the docs were not generated.
As we have changed the release procedure (so that now we do not push directly to main branch), maybe we should also change the corresponding GitHub action:
https://github.com/huggingface/datasets/blob/edf1902f954c5568daadebcd8754bdad44b02a85/.github/workflows/build_documentation.yml#L3-L8
Related to:
- #5250
CC: @mishig25 | 5,251 |
https://github.com/huggingface/datasets/issues/5249 | Protect the main branch from inadvertent direct pushes | [
"It seems all the tasks have been addressed, meaning this issue can be closed, no?"
] | We have decided to implement a protection mechanism in this repository, so that nobody (not even administrators) can inadvertently push accidentally directly to the main branch.
See context here:
- d7c942228b8dcf4de64b00a3053dce59b335f618
To do:
- [x] Protect main branch
- Settings > Branches > Branch protection rules > main > Edit
- [x] Check: Do not allow bypassing the above settings
- The above settings will apply to administrators and custom roles with the "bypass branch protections" permission.
- [x] Additionally, uncheck: Require approvals [under "Require a pull request before merging", which was already checked]
- Before, we could exceptionally merge a non-approved PR, using Administrator bypass
- Now that Administrator bypass is no longer possible, we would always need an approval to be able to merge; and pull request authors cannot approve their own pull requests. This could be an inconvenient in some exceptional circumstances when an urgent fix is needed
- Nevertheless, although it is no longer enforced, it is strongly recommended to merge PRs only if they have at least one approval
- [x] #5250
- So that direct pushes to main branch are no longer necessary | 5,249 |
https://github.com/huggingface/datasets/issues/5245 | Unable to rename columns in streaming dataset | [
"Hi @peregilk this bug is directly related to https://github.com/huggingface/datasets/issues/3888, and still not fixed... But I'll try to have a look!",
"Thanks @alvarobartt. It is great if you are able to fix it, but when reading the explanation it seems like it is possible to work around it.\r\n\r\nWe also trie... | ### Describe the bug
Trying to rename column in a streaming datasets, destroys the features object.
### Steps to reproduce the bug
The following code illustrates the error:
```
from datasets import load_dataset
dataset = load_dataset('mc4', 'en', streaming=True, split='train')
dataset.info.features
# {'text': Value(dtype='string', id=None), 'timestamp': Value(dtype='string', id=None), 'url': Value(dtype='string', id=None)}
dataset = dataset.rename_column("text", "content")
dataset.info.features
# This returned object is now None!
```
### Expected behavior
This should just alter the renamed column.
### Environment info
datasets 2.6.1 | 5,245 |
https://github.com/huggingface/datasets/issues/5244 | Allow dataset streaming from private a private source when loading a dataset with a dataset loading script | [
"Hi ! What kind of private source ? We're exploring adding support for cloud storage and URIs like s3://, gs:// etc. with authentication in the download manager",
"Hello! It's a google cloud storage, so gs://, but I'm using it with https.\r\nBeing able to provide a file system like [here](https://huggingface.co/d... | ### Feature request
Add arguments to the function _get_authentication_headers_for_url_ like custom_endpoint and custom_token in order to add flexibility when downloading files from a private source.
It should also be possible to provide these arguments from the dataset loading script, maybe giving them to the dl_manager
### Motivation
It is possible to share a dataset hosted on another platform by writing a dataset loading script. It works perfectly for publicly available resources.
For resources that require authentication, you can provide a [download_custom](https://huggingface.co/docs/datasets/package_reference/builder_classes#datasets.DownloadManager) method to the download_manager.
Unfortunately, this function doesn't work with **dataset streaming**.
A solution so as to allow dataset streaming from private sources would be a more flexible _get_authentication_headers_for_url_ function.
### Your contribution
Would you be interested in this improvement ?
If so I could provide a PR. I've got something working locally, but it's not very clean, I'd need some guidance regarding integration. | 5,244 |
https://github.com/huggingface/datasets/issues/5243 | Download only split data | [
"Hi @capsabogdan! Unfortunately, it's hard to implement because quite often datasets data is being hosted in a single archive for all splits :( So we have to download the whole archive to split it into splits. This is the case for CommonVoice too. \r\n\r\nHowever, for cases when data is distributed in separate arch... | ### Feature request
Is it possible to download only the data that I am requesting and not the entire dataset? I run out of disk spaceas it seems to download the entire dataset, instead of only the part needed.
common_voice["test"] = load_dataset("mozilla-foundation/common_voice_11_0", "en", split="test",
cache_dir="cache/path...",
use_auth_token=True,
download_config=DownloadConfig(delete_extracted='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')
)
### Motivation
efficiency improvement
### Your contribution
n/a | 5,243 |
https://github.com/huggingface/datasets/issues/5242 | Failed Data Processing upon upload with zip file full of images | [
"cc @abhishekkrthakur @SBrandeis "
] | I went to autotrain and under image classification arrived where it was time to prepare my dataset. Screenshot below
I chose the method 2 option. I have a csv file with two columns. ~23,000 files.
I uploaded this and chose the image_relpath, and target columns.
The image uploader said that I could only upload 10,000 singular images at a time so the 2nd option was to zip the images up and upload a zip archive which I did.
That all uploaded.
Now I have the message below. It appears the zip archive does just uncompress on the Hugging Face end?
What am I missing here?
| 5,242 |
https://github.com/huggingface/datasets/issues/5232 | Incompatible dill versions in datasets 2.6.1 | [
"Thanks for reporting, @vinaykakade.\r\n\r\nWe are discussing about making a release early this week.\r\n\r\nPlease note that in the meantime, in your specific case (as we also pointed out here: https://github.com/huggingface/datasets/issues/5162#issuecomment-1291720293), you can circumvent the issue by pinning `mu... | ### Describe the bug
datasets version 2.6.1 has a dependency on dill<0.3.6. This causes a conflict with dill>=0.3.6 used by multiprocess dependency in datasets 2.6.1
This issue is already fixed in https://github.com/huggingface/datasets/pull/5166/files, but not yet been released. Please release a new version of the datasets library to fix this.
### Steps to reproduce the bug
1. Create requirements.in with only dependency being datasets (or datasets[s3])
2. Run pip-compile
3. The output is as follows:
```
Could not find a version that matches dill<0.3.6,>=0.3.6 (from datasets[s3]==2.6.1->-r requirements.in (line 1))
Tried: 0.2, 0.2, 0.2.1, 0.2.1, 0.2.2, 0.2.2, 0.2.3, 0.2.3, 0.2.4, 0.2.4, 0.2.5, 0.2.5, 0.2.6, 0.2.7, 0.2.7.1, 0.2.8, 0.2.8.1, 0.2.8.2, 0.2.9, 0.3.0, 0.3.1, 0.3.1.1, 0.3.2, 0.3.3, 0.3.3, 0.3.4, 0.3.4, 0.3.5, 0.3.5, 0.3.5.1, 0.3.5.1, 0.3.6, 0.3.6
Skipped pre-versions: 0.1a1, 0.2a1, 0.2a1, 0.2b1, 0.2b1
There are incompatible versions in the resolved dependencies:
dill<0.3.6 (from datasets[s3]==2.6.1->-r requirements.in (line 1))
dill>=0.3.6 (from multiprocess==0.70.14->datasets[s3]==2.6.1->-r requirements.in (line 1))
```
### Expected behavior
pip-compile produces requirements.txt without any conflicts
### Environment info
datasets version 2.6.1 | 5,232 |
https://github.com/huggingface/datasets/issues/5231 | Using `set_format(type='torch', columns=columns)` makes Array2D/3D columns stop formatting correctly | [
"In case others find this, the problem was not with set_format, but my usages of `to_pandas()` and `from_pandas()` which I was using during dataset splitting; somewhere in the chain of converting to and from pandas the `Array2D/Array3D` types get converted to series of `Sequence()` types"
] | I have a Dataset with two Features defined as follows:
```
'image': Array3D(dtype="int64", shape=(3, 224, 224)),
'bbox': Array2D(dtype="int64", shape=(512, 4)),
```
On said dataset, if I `dataset.set_format(type='torch')` and then use the dataset in a dataloader, these columns are correctly cast to Tensors of (batch_size, 3, 224, 244) for example.
However, if I `dataset.set_format(type='torch', columns=['image', 'bbox'])` these columns are cast to Lists of tensors and miss the batch size completely (the 3 dimension is the list length).
I'm currently digging through datasets formatting code to try and find out why, but was curious if someone knew an immediate solution for this. | 5,231 |
https://github.com/huggingface/datasets/issues/5230 | dataclasses error when importing the library in python 3.11 | [
"I opened [this issue](https://github.com/python/cpython/issues/99401).\r\nPython's maintainers say that the issue is caused by [this change](https://docs.python.org/3.11/whatsnew/3.11.html#dataclasses).\r\nI believe adding a `__hash__` method to `datasets.utils.version.Version` should solve (at least partially) th... | ### Describe the bug
When I import datasets using python 3.11 the dataclasses standard library raises the following error:
`ValueError: mutable default <class 'datasets.utils.version.Version'> for field version is not allowed: use default_factory`
When I tried to import the library using the following jupyter notebook:
```
%%bash
# create python 3.11 conda env
conda create --yes --quiet -n myenv -c conda-forge python=3.11
# activate is
source activate myenv
# install pyarrow
/opt/conda/envs/myenv/bin/python -m pip install --quiet --extra-index-url https://pypi.fury.io/arrow-nightlies/ \
--prefer-binary --pre pyarrow
# install datasets
/opt/conda/envs/myenv/bin/python -m pip install --quiet datasets
```
```
# create a python file that only imports datasets
with open("import_datasets.py", 'w') as f:
f.write("import datasets")
# run it with the env
!/opt/conda/envs/myenv/bin/python import_datasets.py
```
I get the following error:
```
Traceback (most recent call last):
File "/kaggle/working/import_datasets.py", line 1, in <module>
import datasets
File "/opt/conda/envs/myenv/lib/python3.11/site-packages/datasets/__init__.py", line 45, in <module>
from .builder import ArrowBasedBuilder, BeamBasedBuilder, BuilderConfig, DatasetBuilder, GeneratorBasedBuilder
File "/opt/conda/envs/myenv/lib/python3.11/site-packages/datasets/builder.py", line 91, in <module>
@dataclass
^^^^^^^^^
File "/opt/conda/envs/myenv/lib/python3.11/dataclasses.py", line 1221, in dataclass
return wrap(cls)
^^^^^^^^^
File "/opt/conda/envs/myenv/lib/python3.11/dataclasses.py", line 1211, in wrap
return _process_class(cls, init, repr, eq, order, unsafe_hash,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/conda/envs/myenv/lib/python3.11/dataclasses.py", line 959, in _process_class
cls_fields.append(_get_field(cls, name, type, kw_only))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/conda/envs/myenv/lib/python3.11/dataclasses.py", line 816, in _get_field
raise ValueError(f'mutable default {type(f.default)} for field '
ValueError: mutable default <class 'datasets.utils.version.Version'> for field version is not allowed: use default_factory
```
This is probably due to one of the following changes in the [dataclasses standard library](https://docs.python.org/3/library/dataclasses.html) in version 3.11:
1. Changed in version 3.11: Instead of looking for and disallowing objects of type list, dict, or set, unhashable objects are now not allowed as default values. Unhashability is used to approximate mutability.
2. fields may optionally specify a default value, using normal Python syntax:
```
@dataclass
class C:
a: int # 'a' has no default value
b: int = 0 # assign a default value for 'b'
In this example, both a and b will be included in the added __init__() method, which will be defined as:
def __init__(self, a: int, b: int = 0):
```
3. Changed in version 3.11: If a field name is already included in the __slots__ of a base class, it will not be included in the generated __slots__ to prevent [overriding them](https://docs.python.org/3/reference/datamodel.html#datamodel-note-slots). Therefore, do not use __slots__ to retrieve the field names of a dataclass. Use [fields()](https://docs.python.org/3/library/dataclasses.html#dataclasses.fields) instead. To be able to determine inherited slots, base class __slots__ may be any iterable, but not an iterator.
4. weakref_slot: If true (the default is False), add a slot named “__weakref__”, which is required to make an instance weakref-able. It is an error to specify weakref_slot=True without also specifying slots=True.
[TypeError](https://docs.python.org/3/library/exceptions.html#TypeError) will be raised if a field without a default value follows a field with a default value. This is true whether this occurs in a single class, or as a result of class inheritance.
### Steps to reproduce the bug
Steps to reproduce the behavior:
1. go to [the notebook in kaggle](https://www.kaggle.com/yonikremer/repreducing-issue)
2. rub both of the cells
### Expected behavior
I'm expecting no issues.
This error should not occur.
### Environment info
kaggle kernels, with default settings:
pin to original environment, no accelerator. | 5,230 |
https://github.com/huggingface/datasets/issues/5229 | Type error when calling `map` over dataset containing 0-d tensors | [
"Hi! \r\n\r\nWe could address this by calling `.item()` on such tensors to extract the value, but this would lose us the type, which could lead to storing the generated dataset in a suboptimal format. Considering this, I think the only proper fix would be implementing support for 0-D tensors on Apache Arrow's side ... | ### Describe the bug
0-dimensional tensors in a dataset lead to `TypeError: iteration over a 0-d array` when calling `map`. It is easy to generate such tensors by using `.with_format("...")` on the whole dataset.
### Steps to reproduce the bug
```
ds = datasets.Dataset.from_list([{"a": 1}, {"a": 1}]).with_format("torch")
ds.map(None)
```
### Expected behavior
Getting back `ds` without errors.
### Environment info
Python 3.10.8
datasets 2.6.
torch 1.13.0 | 5,229 |
https://github.com/huggingface/datasets/issues/5228 | Loading a dataset from the hub fails if you happen to have a folder of the same name | [
"`load_dataset` first checks for a local directory before checking for the Hub.\r\n\r\nTo make it explicit that it has to fetch the Hub, we could support the `hffs` syntax:\r\n```python\r\nload_dataset(\"hf://datasets/glue\")\r\n```\r\n\r\nwould that work for you ? Also cc @mariosasko who's leading the `hffs` proje... | ### Describe the bug
I'm not 100% sure this should be considered a bug, but it was certainly annoying to figure out the cause of. And perhaps I am just missing a specific argument needed to avoid this conflict. Basically I had a situation where multiple workers were downloading different parts of the glue dataset and then training on them. Additionally, they were writing their checkpoints to a folder called `glue`. This meant that once one worker had created the `glue` folder to write checkpoints to, the next worker to try to load a glue dataset would fail as shown in the minimal repro below. I'm not sure what the solution would be since I'm not super familiar with the `datasets` code, but I would expect `load_dataset` to not crash just because i have a local folder with the same name as a dataset from the hub.
### Steps to reproduce the bug
```
In [1]: import datasets
In [2]: rte = datasets.load_dataset('glue', 'rte')
Downloading and preparing dataset glue/rte to /Users/danielking/.cache/huggingface/datasets/glue/rte/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96da6053ad...
Downloading data: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 697k/697k [00:00<00:00, 6.08MB/s]
Dataset glue downloaded and prepared to /Users/danielking/.cache/huggingface/datasets/glue/rte/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96da6053ad. Subsequent calls will reuse this data.
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00, 773.81it/s]
In [3]: import os
In [4]: os.mkdir('glue')
In [5]: rte = datasets.load_dataset('glue', 'rte')
---------------------------------------------------------------------------
EmptyDatasetError Traceback (most recent call last)
<ipython-input-5-0d6b9ad8bbd0> in <cell line: 1>()
----> 1 rte = datasets.load_dataset('glue', 'rte')
~/miniconda3/envs/composer/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, revision, use_auth_token, task, streaming, **config_kwargs)
1717
1718 # Create a dataset builder
-> 1719 builder_instance = load_dataset_builder(
1720 path=path,
1721 name=name,
~/miniconda3/envs/composer/lib/python3.9/site-packages/datasets/load.py in load_dataset_builder(path, name, data_dir, data_files, cache_dir, features, download_config, download_mode, revision, use_auth_token, **config_kwargs)
1495 download_config = download_config.copy() if download_config else DownloadConfig()
1496 download_config.use_auth_token = use_auth_token
-> 1497 dataset_module = dataset_module_factory(
1498 path,
1499 revision=revision,
~/miniconda3/envs/composer/lib/python3.9/site-packages/datasets/load.py in dataset_module_factory(path, revision, download_config, download_mode, dynamic_modules_path, data_dir, data_files, **download_kwargs)
1152 ).get_module()
1153 elif os.path.isdir(path):
-> 1154 return LocalDatasetModuleFactoryWithoutScript(
1155 path, data_dir=data_dir, data_files=data_files, download_mode=download_mode
1156 ).get_module()
~/miniconda3/envs/composer/lib/python3.9/site-packages/datasets/load.py in get_module(self)
624 base_path = os.path.join(self.path, self.data_dir) if self.data_dir else self.path
625 patterns = (
--> 626 sanitize_patterns(self.data_files) if self.data_files is not None else get_data_patterns_locally(base_path)
627 )
628 data_files = DataFilesDict.from_local_or_remote(
~/miniconda3/envs/composer/lib/python3.9/site-packages/datasets/data_files.py in get_data_patterns_locally(base_path)
458 return _get_data_files_patterns(resolver)
459 except FileNotFoundError:
--> 460 raise EmptyDatasetError(f"The directory at {base_path} doesn't contain any data files") from None
461
462
EmptyDatasetError: The directory at glue doesn't contain any data files
```
### Expected behavior
Dataset is still able to be loaded from the hub even if I have a local folder with the same name.
### Environment info
datasets version: 2.6.1 | 5,228 |
https://github.com/huggingface/datasets/issues/5227 | datasets.data_files.EmptyDatasetError: The directory at wikisql doesn't contain any data files | [
"Fixed. Please close.",
"how to fix?i need your help"
] | ### Describe the bug
From these lines:
from datasets import list_datasets, load_dataset
dataset = load_dataset("wikisql","binary")
I get error message:
datasets.data_files.EmptyDatasetError: The directory at wikisql doesn't contain any data files
And yet the 'wikisql' is reported to exist via the list_datasets().
Any help appreciated.
### Steps to reproduce the bug
From these lines:
from datasets import list_datasets, load_dataset
dataset = load_dataset("wikisql","binary")
I get error message:
datasets.data_files.EmptyDatasetError: The directory at wikisql doesn't contain any data files
And yet the 'wikisql' is reported to exist via the list_datasets().
Any help appreciated.
### Expected behavior
Dataset should load. This same code used to work.
### Environment info
Mac OS | 5,227 |
https://github.com/huggingface/datasets/issues/5226 | Q: Memory release when removing the column? | [
"Hi ! Datasets are memory mapped from your disk, i.e. they're not loaded in RAM. This is possible thanks to the Arrow data format.\r\n\r\nTherefore the column you remove is not in RAM, so removing it doesn't cause the RAM to decrease.",
"Thanks for the explanation! @lhoestq \r\nI wonder since it is memory mapped,... | ### Describe the bug
How do I release memory when I use methods like `.remove_columns()` or `clear()` in notebooks?
```python
from datasets import load_dataset
common_voice = load_dataset("mozilla-foundation/common_voice_11_0", "ja", use_auth_token=True)
# check memory -> RAM Used (GB): 0.704 / Total (GB) 33.670
common_voice = common_voice.remove_columns(column_names=common_voice.column_names['train'])
common_voice.clear()
# check memory -> RAM Used (GB): 0.705 / Total (GB) 33.670
```
I tried `gc.collect()` but did not help
### Steps to reproduce the bug
1. load dataset
2. remove all the columns
3. check memory is reduced or not
[link to reproduce](https://www.kaggle.com/code/bayartsogtya/huggingface-dataset-memory-issue/notebook?scriptVersionId=110630567)
### Expected behavior
Memory released when I remove the column
### Environment info
- `datasets` version: 2.1.0
- Platform: Linux-5.15.65+-x86_64-with-debian-bullseye-sid
- Python version: 3.7.12
- PyArrow version: 8.0.0
- Pandas version: 1.3.5 | 5,226 |
https://github.com/huggingface/datasets/issues/5225 | Add video feature | [
"@NielsRogge @rwightman may have additional requirements regarding this feature.\r\n\r\nWhen adding a new (decodable) type, the hardest part is choosing the right decoding library. What I mean by \"right\" here is that it has all the features we need and is easy to install (with GPU support?).\r\n\r\nSome candidate... | ### Feature request
Add a `Video` feature to the library so folks can include videos in their datasets.
### Motivation
Being able to load Video data would be quite helpful. However, there are some challenges when it comes to videos:
1. Videos, unlike images, can end up being extremely large files
2. Often times when training video models, you need to do some very specific sampling. Videos might end up needing to be broken down into X number of clips used for training/inference
3. Videos have an additional audio stream, which must be accounted for
4. The feature needs to be able to encode/decode videos (with right video settings) from bytes.
### Your contribution
I did work on this a while back in [this (now closed) PR](https://github.com/huggingface/datasets/pull/4532). It used a library I made called [encoded_video](https://github.com/nateraw/encoded-video), which is basically the utils from [pytorchvideo](https://github.com/facebookresearch/pytorchvideo), but without the `torch` dep. It included the ability to read/write from bytes, as we need to do here. We don't want to be using a sketchy library that I made as a dependency in this repo, though.
Would love to use this issue as a place to:
- brainstorm ideas on how to do this right
- list ways/examples to work around it for now
CC @sayakpaul @mariosasko @fcakyon | 5,225 |
https://github.com/huggingface/datasets/issues/5224 | Seems to freeze when loading audio dataset with wav files from local folder | [
"I just tried to do the same but changing the `.wav` files to `.mp3` files and that doesn't fix it.",
"I don't know if anyone will ever read this but I've tried to upload the same dataset with google colab and the output seems more clarifying. I didn't specify the train/test split so the dataset wasn't fully uplo... | ### Describe the bug
I'm following the instructions in [https://huggingface.co/docs/datasets/audio_load#audiofolder-with-metadata](url) to be able to load a dataset from a local folder.
I have everything into a folder, into a train folder and then the audios and csv. When I try to load the dataset and run from terminal, seems to work but then freezes with no apparent reason.
The metadata.csv file contains a few columns but the important ones, `file_name` with the filename and `transcription` with the transcription are okay.
The audios are `.wav` files, I don't know if that might be the problem (I will proceed to try to change them all to `.mp3` and try again).
### Steps to reproduce the bug
The code I'm using:
```python
from datasets import load_dataset
dataset = load_dataset("audiofolder", data_dir="../archive/Dataset")
dataset[0]["audio"]
```
The output I obtain:
```
Resolving data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 311135.43it/s]
Using custom data configuration default-38d4546ffd010f3e
Downloading and preparing dataset audiofolder/default to /Users/mine/.cache/huggingface/datasets/audiofolder/default-38d4546ffd010f3e/0.0.0/6cbdd16f8688354c63b4e2a36e1585d05de285023ee6443ffd71c4182055c0fc...
Resolving data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 166467.72it/s]
Using custom data configuration default-38d4546ffd010f3e
Resolving data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 187772.74it/s]
Using custom data configuration default-38d4546ffd010f3e
Resolving data files: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 59623.71it/s]
Resolving data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 138090.55it/s]
Resolving data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 106065.64it/s]
Resolving data files: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 56036.38it/s]
Resolving data files: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 74004.24it/s]
Resolving data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 162343.45it/s]
Resolving data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 101881.23it/s]
Using custom data configuration default-38d4546ffd010f3e
Resolving data files: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 60145.67it/s]
Resolving data files: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 80890.02it/s]
Resolving data files: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 54036.67it/s]
Resolving data files: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 95851.09it/s]
Resolving data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 155897.00it/s]
Resolving data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 137656.96it/s]
Resolving data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 439/439 [00:00<00:00, 131230.81it/s]
Using custom data configuration default-38d4546ffd010f3e
Using custom data configuration default-38d4546ffd010f3e
Using custom data configuration default-38d4546ffd010f3e
Using custom data configuration default-38d4546ffd010f3e
Using custom data configuration default-38d4546ffd010f3e
Using custom data configuration default-38d4546ffd010f3e
Using custom data configuration default-38d4546ffd010f3e
Using custom data configuration default-38d4546ffd010f3e
Using custom data configuration default-38d4546ffd010f3e
Using custom data configuration default-38d4546ffd010f3e
Using custom data configuration default-38d4546ffd010f3e
Using custom data configuration default-38d4546ffd010f3e
Using custom data configuration default-38d4546ffd010f3e
```
And then here it just freezes and nothing more happens.
### Expected behavior
Load the dataset.
### Environment info
Datasets version:
datasets 2.6.1 pypi_0 pypi
| 5,224 |
https://github.com/huggingface/datasets/issues/5222 | HuggingFace website is incorrectly reporting that my datasets are pickled | [
"cc @McPatate maybe you know what's happening ?",
"Yes I think I know what is happening. We check in zips for pickles, and the UI must display the pickle jar when a scan has an associated list of imports, even when empty.\r\n~I'll fix ASAP !~",
"> I'll fix ASAP !\r\n\r\nActually I'd rather leave it like that f... | ### Describe the bug
HuggingFace is incorrectly reporting that my datasets are pickled. They are not picked, they are simple ZIP files containing PNG images.
Hopefully this is the right location to report this bug.
### Steps to reproduce the bug
Inspect my dataset respository here: https://huggingface.co/datasets/ProGamerGov/StableDiffusion-v1-5-Regularization-Images
### Expected behavior
They should not be reported as being pickled.
### Environment info
N/A | 5,222 |
https://github.com/huggingface/datasets/issues/5221 | Cannot push | [
"Did you run `huggingface-cli lfs-enable-largefiles` before committing or before adding ? Maybe you can try before adding\r\n\r\nAnyway I'd encourage you to split your data into several TAR archives if possible, this way the dataset can loaded faster using multiprocessing (by giving each process a subset of shards ... | ### Describe the bug
I am facing the issue when I try to push the tar.gz file around 11G to HUB.
```
(venv) ╭─laptop@laptop ~/PersonalProjects/data/ulaanbal_v0 ‹main●›
╰─$ du -sh *
4.0K README.md
13G data
516K test.jsonl
18M train.jsonl
4.0K ulaanbal_v0.py
11G ulaanbal_v0.tar.gz
452K validation.jsonl
(venv) ╭─laptop@laptop~/PersonalProjects/data/ulaanbal_v0 ‹main●›
╰─$ git add ulaanbal_v0.tar.gz && git commit -m 'large version'
(venv) ╭─laptop@laptop ~/PersonalProjects/data/ulaanbal_v0 ‹main●›
╰─$ git push
EOFoading LFS objects: 0% (0/1), 0 B | 0 B/s
Uploading LFS objects: 0% (0/1), 0 B | 0 B/s, done.
error: failed to push some refs to 'https://huggingface.co/datasets/bayartsogt/ulaanbal_v0'
```
I have already tried pushing a small version of this and it was working fine. So my guess it is probably because of the big file.
Following I run before the commit:
```
╰─$ git lfs install
╰─$ huggingface-cli lfs-enable-largefiles .
```
### Steps to reproduce the bug
Create a private dataset on huggingface and push 12G tar.gz file
### Expected behavior
To be pushed with no issue
### Environment info
- `datasets` version: 2.6.1
- Platform: Darwin-21.6.0-x86_64-i386-64bit
- Python version: 3.7.11
- PyArrow version: 10.0.0
- Pandas version: 1.3.5
| 5,221 |
https://github.com/huggingface/datasets/issues/5220 | Implicit type conversion of lists in to_pandas | [
"I think this behavior comes from PyArrow:\r\n```python\r\nimport pyarrow as pa\r\nt = pa.table({\"a\": [[0]]})\r\nt.to_pandas().a.values[0]\r\n# array([0])\r\n```\r\n\r\nI believe this has to do with zero-copy: you can get a pandas DataFrame without copying the buffers from arrow, and therefore end up with numpy a... | ### Describe the bug
```
ds = Dataset.from_list([{'a':[1,2,3]}])
ds.to_pandas().a.values[0]
```
Results in `array([1, 2, 3])` -- a rather unexpected conversion of types which made downstream tools expecting lists not happy.
### Steps to reproduce the bug
See snippet
### Expected behavior
Keep the original type
### Environment info
datasets 2.6.1
python 3.8.10 | 5,220 |
https://github.com/huggingface/datasets/issues/5219 | Delta Tables usage using Datasets Library | [
"Hi ! Interesting :) Can you provide concrete examples of cases where it can be useful ?",
"Few example blogs and posts that might help on this - \r\n\r\n1. https://hevodata.com/learn/databricks-delta-tables/\r\n2. https://docs.databricks.com/delta/index.html\r\n\r\nBasically, we are looking at utility of Dataset... | ### Feature request
Adding compatibility of Datasets library with Delta Format. Elevating the utilities of Datasets library from Machine Learning Scope to Data Engineering Scope as well.
### Motivation
We know datasets library can absorb csv, json, parquet, etc. file formats but it would be great if Datasets library could work with Delta Tables (with delta format) as it has different features such as time travelling, layout optimization, query performance, aids in Data Engineering.
This will help and enhance Datasets library from Machine Learning utility to Data Engineering utilities and expand horizons thereafter. I am totally using Datasets library in all my usecases and as my role expands so does the work, compatibility with Datasets library is something I don't want to lose.
### Your contribution
Would love to work on this feature, even if this has to picked up from scratch, including design paradigms and patterns.
I have basic idea about Delta Live Tables, would brush it easily for this feature. | 5,219 |
https://github.com/huggingface/datasets/issues/5218 | Delta Tables usage using Datasets Library | [] | ### Feature request
Adding compatibility of Datasets library with Delta Format. Elevating the utilities of Datasets library from Machine Learning Scope to Data Engineering Scope as well.
### Motivation
We know datasets library can absorb csv, json, parquet, etc. file formats but it would be great if Datasets library could work with Delta Tables (with delta format) as it has different features such as time travelling, layout optimization, query performance, aids in Data Engineering.
This will help and enhance Datasets library from Machine Learning utility to Data Engineering utilities and expand horizons thereafter. I am totally using Datasets library in all my usecases and as my role expands so does the work, compatibility with Datasets library is something I don't want to lose.
### Your contribution
Would love to work on this feature, even if this has to picked up from scratch, including design paradigms and patterns.
I have basic idea about Delta Live Tables, would brush it easily for this feature. | 5,218 |
https://github.com/huggingface/datasets/issues/5216 | save_elasticsearch_index | [
"Hi ! I think there exist tools to dump and reload an index in your elastic search but I'm not super familiar with it.\r\n\r\nAnyway after reloading an index in elastic search you can call `ds.load_elasticsearch_index` which will connect the index to the dataset without re-indexing"
] | Hi,
I am new to Dataset and elasticsearch. I was wondering is there any equivalent approach to save elasticsearch index as of save_faiss_index locally for later use, to remove the need to re-index a dataset? | 5,216 |
https://github.com/huggingface/datasets/issues/5209 | Implement ability to define splits in metadata section of dataset card | [
"@merveenoyan Do you want different files to be splits or configurations?\r\n\r\nFrom [what you specified in `Readme.md`](https://huggingface.co/datasets/inria-soda/tabular-benchmark/commit/fb4575853772c62a20203bdd6cc0202f5db4ce4e) I hypothesize that you want to have 4 **configs** corresponding to directories: `\"c... | ### Feature request
If you go here: https://huggingface.co/datasets/inria-soda/tabular-benchmark/tree/main you will see bunch of folders that has various CSV files. I’d like dataset viewer to show these files instead of only one dataset like it currently does. (and also people to be able to load them as splits instead of loading through `data_files`)
e.g GLUE has various splits on viewer but it’s too overkill to ask people to implement loading script, so it would be better to let them define these in the README file instead.
Also pinging @polinaeterna @lhoestq @adrinjalali
| 5,209 |
https://github.com/huggingface/datasets/issues/5207 | Connection error of the HuggingFace's dataset Hub due to SSLError with proxy | [
"Hi ! It looks like an issue with your python environment, can you make sure you're able to run GET requests to https://huggingface.co using `requests` in python ?",
"Thanks for your reply. Does this mean that I have to use the `do_dataset `function and the `requests `function to download the dataset from the com... | ### Describe the bug
It's weird. I could not normally connect the dataset Hub of HuggingFace due to a SSLError in my office.
Even when I try to connect using my company's proxy address (e.g., http_proxy and https_proxy),
I'm getting the SSLError issue. What should I do to download the datanet stored in HuggingFace normally?
I welcome any comments. I think those comments will be helpful to me.
* Dataset address - https://huggingface.co/datasets/moyix/debian_csrc/viewer/moyix--debian_csrc
* Log message
```
............ OMISSION ..............
Traceback (most recent call last):
File "/data/home/geunsik-lim/qtlab/./transformers/examples/pytorch/language-modeling/run_clm.py", line 587, in <module>
main()
File "/data/home/geunsik-lim/qtlab/./transformers/examples/pytorch/language-modeling/run_clm.py", line 278, in main
raw_datasets = load_dataset(
File "/home/geunsik-lim/anaconda3/envs/deepspeed/lib/python3.10/site-packages/datasets/load.py", line 1719, in load_dataset
builder_instance = load_dataset_builder(
File "/home/geunsik-lim/anaconda3/envs/deepspeed/lib/python3.10/site-packages/datasets/load.py", line 1497, in load_dataset_builder
dataset_module = dataset_module_factory(
File "/home/geunsik-lim/anaconda3/envs/deepspeed/lib/python3.10/site-packages/datasets/load.py", line 1222, in dataset_module_factory
raise e1 from None
File "/home/geunsik-lim/anaconda3/envs/deepspeed/lib/python3.10/site-packages/datasets/load.py", line 1179, in dataset_module_factory
raise ConnectionError(f"Couldn't reach '{path}' on the Hub ({type(e).__name__})")
ConnectionError: Couldn't reach 'moyix/debian_csrc' on the Hub (SSLError)
[2022-11-07 15:23:38,476] [INFO] [launch.py:318:sigkill_handler] Killing subprocess 6760
[2022-11-07 15:23:38,476] [ERROR] [launch.py:324:sigkill_handler] ['/home/geunsik-lim/anaconda3/envs/deepspeed/bin/python', '-u', './transformers/examples/pytorch/language-modeling/run_clm.py', '--local_rank=0', '--model_name_or_path=Salesforce/codegen-350M-multi', '--per_device_train_batch_size=1', '--learning_rate', '2e-5', '--num_train_epochs', '1', '--output_dir=./codegen-350M-finetuned', '--overwrite_output_dir', '--dataset_name', 'moyix/debian_csrc', '--cache_dir', '/data/home/geunsik-lim/.cache', '--tokenizer_name', 'Salesforce/codegen-350M-multi', '--block_size', '2048', '--gradient_accumulation_steps', '32', '--do_train', '--fp16', '--deepspeed', 'ds_config_zero2.json'] exits with return code = 1
real 0m7.742s
user 0m4.930s
```
### Steps to reproduce the bug
Steps to reproduce this behavior.
```
(deepspeed) geunsik-lim@ai02:~/qtlab$ ./test_debian_csrc_dataset.py
Traceback (most recent call last):
File "/data/home/geunsik-lim/qtlab/./test_debian_csrc_dataset.py", line 6, in <module>
dataset = load_dataset("moyix/debian_csrc")
File "/home/geunsik-lim/anaconda3/envs/deepspeed/lib/python3.10/site-packages/datasets/load.py", line 1719, in load_dataset
builder_instance = load_dataset_builder(
File "/home/geunsik-lim/anaconda3/envs/deepspeed/lib/python3.10/site-packages/datasets/load.py", line 1497, in load_dataset_builder
dataset_module = dataset_module_factory(
File "/home/geunsik-lim/anaconda3/envs/deepspeed/lib/python3.10/site-packages/datasets/load.py", line 1222, in dataset_module_factory
raise e1 from None
File "/home/geunsik-lim/anaconda3/envs/deepspeed/lib/python3.10/site-packages/datasets/load.py", line 1179, in dataset_module_factory
raise ConnectionError(f"Couldn't reach '{path}' on the Hub ({type(e).__name__})")
ConnectionError: Couldn't reach 'moyix/debian_csrc' on the Hub (SSLError)
(deepspeed) geunsik-lim@ai02:~/qtlab$
(deepspeed) geunsik-lim@ai02:~/qtlab$
(deepspeed) geunsik-lim@ai02:~/qtlab$
(deepspeed) geunsik-lim@ai02:~/qtlab$ cat ./test_debian_csrc_dataset.py
#!/usr/bin/env python
from datasets import load_dataset
dataset = load_dataset("moyix/debian_csrc")
```
1. Adde proxy address of a company in /etc/profile
2. Download dataset with load_dataset() function of datasets package that is provided by HuggingFace.
3. In this case, the address would be "moyix--debian_csrc".
4. I get the "`ConnectionError: Couldn't reach 'moyix/debian_csrc' on the Hub (SSLError`)" error message.
### Expected behavior
* error message:
ConnectionError: Couldn't reach 'moyix/debian_csrc' on the Hub (SSLError)
### Environment info
* software version information:
```
(deepspeed) geunsik-lim@ai02:~$
(deepspeed) geunsik-lim@ai02:~$ conda list -f pytorch
# packages in environment at /home/geunsik-lim/anaconda3/envs/deepspeed:
#
# Name Version Build Channel
pytorch 1.13.0 py3.10_cuda11.7_cudnn8.5.0_0 pytorch
(deepspeed) geunsik-lim@ai02:~$ conda list -f python
# packages in environment at /home/geunsik-lim/anaconda3/envs/deepspeed:
#
# Name Version Build Channel
python 3.10.6 haa1d7c7_1
(deepspeed) geunsik-lim@ai02:~$ conda list -f datasets
# packages in environment at /home/geunsik-lim/anaconda3/envs/deepspeed:
#
# Name Version Build Channel
datasets 2.6.1 py_0 huggingface
(deepspeed) geunsik-lim@ai02:~$ uname -a
Linux ai02 5.4.0-131-generic #147-Ubuntu SMP Fri Oct 14 17:07:22 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux
(deepspeed) geunsik-lim@ai02:~$ cat /etc/lsb-release
DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=20.04
DISTRIB_CODENAME=focal
DISTRIB_DESCRIPTION="Ubuntu 20.04.5 LTS"
``` | 5,207 |
https://github.com/huggingface/datasets/issues/5206 | Use logging instead of printing to console | [
"Actually upon closer inspection, it is documented in the code that this behavior is intentional, so I'll close this."
] | ### Describe the bug
Some logs ([here](https://github.com/huggingface/datasets/blob/4a6e1fe2735505efc7e3a3dbd3e1835da0702575/src/datasets/builder.py#L778), [here](https://github.com/huggingface/datasets/blob/4a6e1fe2735505efc7e3a3dbd3e1835da0702575/src/datasets/builder.py#L786), and [here](https://github.com/huggingface/datasets/blob/4a6e1fe2735505efc7e3a3dbd3e1835da0702575/src/datasets/builder.py#L830)) generated by the `DatasetBuilder` are printed to the console instead of passed to `datasets` logger.
### Steps to reproduce the bug
```python
>> import datasets
>> datasets.load_dataset("some-dataset")
Downloading and preparing dataset csv/data to <path>...
Downloading data files: 100%|██████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00, 7729.06it/s]
Extracting data files: 100%|████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00, 527.23it/s]
Dataset csv downloaded and prepared to <path>. Subsequent calls will reuse this data.
```
### Expected behavior
The logs should not be printed to the console directly but passed to the logger so that the user can redirect them wherever he wants.
### Environment info
- `datasets` version: 2.6.1
- Platform: macOS-13.0-x86_64-i386-64bit
- Python version: 3.9.15
- PyArrow version: 10.0.0
- Pandas version: 1.5.1 | 5,206 |
https://github.com/huggingface/datasets/issues/5204 | `push_to_hub` not propagating `token` through `DownloadConfig` | [
"#self-assign",
"@lhoestq can you close this issue as part of the recent #5205 merge? Thanks 🤗 ",
"Thank you :)"
] | ### Describe the bug
When trying to upload a new 🤗 Dataset to the Hub via Python, and providing the `token` as a parameter to the `Dataset.push_to_hub` function, it just works for the first time, assuming that the dataset didn't exist before.
But when trying to run `Dataset.push_to_hub` again over the same dataset, instead of updating it, it throws a `ConnectionError` when trying to retrieve the `README.md` that may contain some metadata about the dataset, so as to also update it, but since the `token` is not propagated, the `DownloadConfig` provided to the `datasets.utils.file_utils.get_from_cache` function doesn't contain the `use_auth_token` value set to `token`, it's just using the default one which is None/False.
So on, when uploading a dataset via Python with `push_to_hub` with the `token` as a parameter with the HuggingFace API Token as value, it can just be uploaded when the dataset is new, otherwise it fails with to `ConnectionError` due to the `token` not being propagated as `use_auth_token`.
### Steps to reproduce the bug
Let's create a new dataset in our HF account via Python as:
```python
from datasets import Dataset
data = {"a": [1, 2, 3], "b": [4, 5, 6]}
ds = Dataset.from_dict(data)
ds.push_to_hub(repo_id=<HF_USERNAME>/<HF_DATASET>, private=private, token=<HF_TOKEN_HERE>)
```
When we create the `Dataset` for the first time it works and there are no issues, but when trying to actually upload a new version of the same dataset (same name under the same username), we encounter the following issue:
```python
from datasets import Dataset
data = {"a": [1, 2, 3], "b": [4, 5, 6]}
ds = Dataset.from_dict(data)
ds.push_to_hub(repo_id=<HF_USERNAME>/<HF_DATASET>, private=private, token=<HF_TOKEN_HERE>)
>>> ConnectionError: Couldn't reach https://huggingface.co/datasets/alvarobartt/demo/resolve/main/README.md (ConnectionError('Unauthorized for URL https://huggingface.co/datasets/<HF_USERNAME>/<HF_DATASET>/resolve/main/README.md. Please use the parameter `use_auth_token=True` after logging in with `huggingface-cli login`'))
```
### Expected behavior
Ideally, the `token` parameter provided to `push_to_hub` should be propagated and used to download the `README.md` when trying to update a `Dataset`, instead of throwing that exception, so that the authentication can be done directly through code without running `huggingface-cli login`as mentioned at https://huggingface.co/docs/datasets/upload_dataset#upload-with-python.
### Environment info
- `datasets` version: 2.6.1
- Platform: macOS-13.0-arm64-arm-64bit
- Python version: 3.10.8
- PyArrow version: 10.0.0
- Pandas version: 1.5.1 | 5,204 |
https://github.com/huggingface/datasets/issues/5202 | CI fails after bulk edit of canonical datasets | [
"Fixed by: https://huggingface.co/datasets/paws/discussions/1"
] | ```
______ test_get_dataset_config_info[paws-labeled_final-expected_splits2] _______
[gw0] linux -- Python 3.7.15 /opt/hostedtoolcache/Python/3.7.15/x64/bin/python
path = 'paws', config_name = 'labeled_final'
expected_splits = ['train', 'test', 'validation']
@pytest.mark.parametrize(
"path, config_name, expected_splits",
[
("squad", "plain_text", ["train", "validation"]),
("dalle-mini/wit", "dalle-mini--wit", ["train"]),
("paws", "labeled_final", ["train", "test", "validation"]),
],
)
def test_get_dataset_config_info(path, config_name, expected_splits):
info = get_dataset_config_info(path, config_name=config_name)
assert info.config_name == config_name
> assert list(info.splits.keys()) == expected_splits
E AssertionError: assert ['test', 'tra... 'validation'] == ['train', 'te... 'validation']
E At index 0 diff: 'test' != 'train'
E Full diff:
E - ['train', 'test', 'validation']
E + ['test', 'train', 'validation']
tests/test_inspect.py:45: AssertionError
_ test_get_dataset_info[paws-expected_configs2-expected_splits_in_first_config2] _
[gw0] linux -- Python 3.7.15 /opt/hostedtoolcache/Python/3.7.15/x64/bin/python
path = 'paws'
expected_configs = ['labeled_final', 'labeled_swap', 'unlabeled_final']
expected_splits_in_first_config = ['train', 'test', 'validation']
@pytest.mark.parametrize(
"path, expected_configs, expected_splits_in_first_config",
[
("squad", ["plain_text"], ["train", "validation"]),
("dalle-mini/wit", ["dalle-mini--wit"], ["train"]),
("paws", ["labeled_final", "labeled_swap", "unlabeled_final"], ["train", "test", "validation"]),
],
)
def test_get_dataset_info(path, expected_configs, expected_splits_in_first_config):
infos = get_dataset_infos(path)
assert list(infos.keys()) == expected_configs
expected_config = expected_configs[0]
assert expected_config in infos
info = infos[expected_config]
assert info.config_name == expected_config
> assert list(info.splits.keys()) == expected_splits_in_first_config
E AssertionError: assert ['test', 'tra... 'validation'] == ['train', 'te... 'validation']
E At index 0 diff: 'test' != 'train'
E Full diff:
E - ['train', 'test', 'validation']
E + ['test', 'train', 'validation']
tests/test_inspect.py:90: AssertionError
______ test_get_dataset_split_names[paws-labeled_final-expected_splits2] _______
[gw0] linux -- Python 3.7.15 /opt/hostedtoolcache/Python/3.7.15/x64/bin/python
path = 'paws', expected_config = 'labeled_final'
expected_splits = ['train', 'test', 'validation']
@pytest.mark.parametrize(
"path, expected_config, expected_splits",
[
("squad", "plain_text", ["train", "validation"]),
("dalle-mini/wit", "dalle-mini--wit", ["train"]),
("paws", "labeled_final", ["train", "test", "validation"]),
],
)
def test_get_dataset_split_names(path, expected_config, expected_splits):
infos = get_dataset_infos(path)
assert expected_config in infos
info = infos[expected_config]
assert info.config_name == expected_config
> assert list(info.splits.keys()) == expected_splits
E AssertionError: assert ['test', 'tra... 'validation'] == ['train', 'te... 'validation']
E At index 0 diff: 'test' != 'train'
E Full diff:
E - ['train', 'test', 'validation']
E + ['test', 'train', 'validation']
``` | 5,202 |
https://github.com/huggingface/datasets/issues/5200 | Some links to canonical datasets in the docs are outdated | [
"Thanks for catching this, I can go through the docs and replace the links to their corresponding datasets on the Hub!"
] | As we don't have canonical datasets in the github repo anymore, some old links to them doesn't work. I don't know how many of them are there, I found link to SuperGlue here: https://huggingface.co/docs/datasets/dataset_script#multiple-configurations, probably there are more of them. These links should be replaced by links to the corresponding datasets on the Hub. | 5,200 |
https://github.com/huggingface/datasets/issues/5193 | "One or several metadata. were found, but not in the same directory or in a parent directory" | [
"Also unrelated but still: https://huggingface.co/docs/datasets/image_dataset#generate-the-dataset\r\n```If your loading script passed the test, you should now have a dataset_infos.json file in your dataset folder.```\r\nIt's not the case anymore as it's now in the readme.md, it was confusing to me",
"And here is... | ### Describe the bug
When loading my own dataset, on loading it I get an error.
Here is my dataset link: https://huggingface.co/datasets/corentinm7/MyoQuant-SDH-Data
And the error after loading with:
```python
from datasets import load_dataset
load_dataset("corentinm7/MyoQuant-SDH-Data")
```
```python
Downloading readme: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3.34k/3.34k [00:00<00:00, 4.45MB/s]
Using custom data configuration SDH_16k-53e7301a92ab0025
Downloading and preparing dataset None/SDH_16k to /home/corentin/.cache/huggingface/datasets/corentinm7___imagefolder/SDH_16k-53e7301a92ab0025/0.0.0/37fbb85cc714a338bea574ac6c7d0b5be5aff46c1862c1989b20e0771199e93f...
Downloading data: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3.28M/3.28M [00:00<00:00, 4.31MB/s]
Downloading data files: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:01<00:00, 1.75s/it]
Downloading data: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1.13G/1.13G [00:15<00:00, 74.3MB/s]
Downloading data files: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:16<00:00, 16.09s/it]
Extracting data files: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:13<00:00, 13.16s/it]
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/corentin/code-project/hugging_face_play/.venv/lib/python3.10/site-packages/datasets/load.py", line 1742, in load_dataset
builder_instance.download_and_prepare(
File "/home/corentin/code-project/hugging_face_play/.venv/lib/python3.10/site-packages/datasets/builder.py", line 814, in download_and_prepare
self._download_and_prepare(
File "/home/corentin/code-project/hugging_face_play/.venv/lib/python3.10/site-packages/datasets/builder.py", line 1423, in _download_and_prepare
super()._download_and_prepare(
File "/home/corentin/code-project/hugging_face_play/.venv/lib/python3.10/site-packages/datasets/builder.py", line 905, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/home/corentin/code-project/hugging_face_play/.venv/lib/python3.10/site-packages/datasets/builder.py", line 1374, in _prepare_split
for key, record in logging.tqdm(
File "/home/corentin/code-project/hugging_face_play/.venv/lib/python3.10/site-packages/tqdm/std.py", line 1195, in __iter__
for obj in iterable:
File "/home/corentin/code-project/hugging_face_play/.venv/lib/python3.10/site-packages/datasets/packaged_modules/folder_based_builder/folder_based_builder.py", line 394, in _generate_examples
raise ValueError(
ValueError: One or several metadata. were found, but not in the same directory or in a parent directory of /home/corentin/.cache/huggingface/datasets/downloads/extracted/60c4aa8d4da3065bb3d310de4373dffd73bd4dc331aedcb4ee867febe4fdb7cd/validation/sick/2_CG_SDH_TAM_Bin1cKO_ko_pla_4_1640.tif.
```
However the test command is working fine. ```datasets-cli test hugging_face_play/ds_test/SDH_16k.py --save_info --all_configs --force_redownload```
```
Using custom data configuration SDH_16k
Testing builder 'SDH_16k' (1/1)
Downloading and preparing dataset sdh_16k/SDH_16k to /home/corentin/.cache/huggingface/datasets/sdh_16k/SDH_16k/1.0.0/21b584239a638aeeda33cba1ac2ca4869d48e4b4f20fb22274d5a5ddc487659d...
Downloading data: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1.13G/1.13G [00:14<00:00, 76.5MB/s]
Downloading data files: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:15<00:00, 15.66s/it]
Downloading data: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3.28M/3.28M [00:02<00:00, 1.44MB/s]
Downloading data files: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:03<00:00, 3.21s/it]
Downloading data files: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 11586.48it/s]
Extracting data files: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:13<00:00, 13.42s/it]
Dataset sdh_16k downloaded and prepared to /home/corentin/.cache/huggingface/datasets/sdh_16k/SDH_16k/1.0.0/21b584239a638aeeda33cba1ac2ca4869d48e4b4f20fb22274d5a5ddc487659d. Subsequent calls will reuse this data.
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00, 605.27it/s]
Dataset card saved at hugging_face_play/ds_test/README.md
Test successful.
```
### Steps to reproduce the bug
Simply run on python
```python
from datasets import load_dataset
load_dataset("corentinm7/MyoQuant-SDH-Data")
```
### Expected behavior
As the test command worked, this error should not appear
### Environment info
- `datasets` version: 2.6.1
- Platform: Linux-5.10.16.3-microsoft-standard-WSL2-x86_64-with-glibc2.31
- Python version: 3.10.6
- PyArrow version: 10.0.0
- Pandas version: 1.5.1
| 5,193 |
https://github.com/huggingface/datasets/issues/5190 | `path` is `None` when downloading a custom audio dataset from the Hub | [
"Hi! Yes, this is expected behavior - we do this as a security measure to not leak local paths (this info would be useless on other users' machines anyways) and only push audio bytes. \r\n"
] | ### Describe the bug
I've created an [audio dataset](https://huggingface.co/datasets/lewtun/audio-test-push) using the `audiofolder` feature desribed in the [docs](https://huggingface.co/docs/datasets/audio_dataset#audiofolder) and then pushed it to the Hub.
Locally, I can see the `audio.path` feature is of the expected form `path/to/data_dir`, but when I download the dataset from the Hub, I see `audio.path` is `None`
Here's an example:
```python
from datasets import load_dataset
ds = load_dataset("lewtun/audio-test-push")
ds["train"][0]
# {
# "audio": {
# "path": None, <-- Is this expected?
# "array": array(
# [
# 3.97140226e-07,
# 7.30310290e-07,
# 7.56406735e-07,
# ...,
# -1.19636677e-01,
# -1.16811886e-01,
# -1.12441722e-01,
# ]
# ),
# "sampling_rate": 44100,
# },
# "song_id": 0,
# "genre_id": 0,
# "genre": "Electronic",
# }
```
Is this expected behaviour? If yes, feel free to close this issue as it's not a true bug then :)
### Steps to reproduce the bug
1. Create an audio dataset with the `audiofolder` feature
2. Push the dataset to the Hub with `push_to_hub()`
3. Download the Hub dataset and inspect the `audio.path` feature
### Expected behavior
`audio.path` points to the file associated with the audio data
### Environment info
- `datasets` version: 2.6.2.dev0
- Platform: macOS-10.16-x86_64-i386-64bit
- Python version: 3.8.13
- PyArrow version: 9.0.0
- Pandas version: 1.5.1 | 5,190 |
https://github.com/huggingface/datasets/issues/5189 | Reduce friction in tabular dataset workflow by eliminating having splits when dataset is loaded | [
"I have to admit I'm not a fan of this idea, as this would result in a non-consistent behavior between tabular and non-tabular datasets, which is confusing if done without the context you provided. Instead, we could consider returning a `Dataset` object rather than `DatasetDict` if there is only one split in the ge... | ### Feature request
Sorry for cryptic name but I'd like to explain using code itself. When I want to load a specific dataset from a repository (for instance, this: https://huggingface.co/datasets/inria-soda/tabular-benchmark)
```python
from datasets import load_dataset
dataset = load_dataset("inria-soda/tabular-benchmark", data_files=["reg_cat/house_sales.csv"], streaming=True)
print(next(iter(dataset["train"])))
```
`datasets` library is essentially designed for people who'd like to use benchmark datasets on various modalities to fine-tune their models, and these benchmark datasets usually have pre-defined train and test splits. However, for tabular workflows, having train and test splits usually ends up model overfitting to validation split so usually the users would like to do validation techniques like `StratifiedKFoldCrossValidation` or when they tune for hyperparameters they do `GridSearchCrossValidation` so often the behavior is to create their own splits. Even [in this paper](https://hal.archives-ouvertes.fr/hal-03723551) a benchmark is introduced but the split is done by authors.
It's a bit confusing for average tabular user to try and load a dataset and see `"train"` so it would be nice if we would not load dataset into a split called `train `by default.
```diff
from datasets import load_dataset
dataset = load_dataset("inria-soda/tabular-benchmark", data_files=["reg_cat/house_sales.csv"], streaming=True)
-print(next(iter(dataset["train"])))
+print(next(iter(dataset)))
```
### Motivation
I explained it above 😅
### Your contribution
I think this is quite a big change that seems small (e.g. how to determine datasets that will not be load to train split?), it's best if we discuss first! | 5,189 |
https://github.com/huggingface/datasets/issues/5186 | Incorrect error message when Dataset.from_sql fails and sqlalchemy not installed | [
"Hi! The first `Dataset.from_sql` call also outputs the \"ImportError: Using URI string without sqlalchemy installed.\" message, but you also get \"During handling of the above exception another exception occurred: ...\" after which the ValueError is printed. I agree that this behavior makes it easy to miss the ori... | ### Describe the bug
When calling `Dataset.from_sql` (in my case, with sqlite3), it fails with a message ```ValueError: Please pass `features` or at least one example when writing data``` when I don't have `sqlalchemy` installed.
### Steps to reproduce the bug
Make a new sqlite db with `sqlite3` and `pandas` from a remote [URL](https://raw.githubusercontent.com/nytimes/covid-19-data/master/us-states.csv).
```python
import sqlite3
import pandas as pd
from datasets import Dataset
conn = sqlite3.connect('us_covid_data.db')
df = pd.read_csv('https://raw.githubusercontent.com/nytimes/covid-19-data/master/us-states.csv')
df.to_sql('states', conn, if_exists='replace')
```
Then if you try to query this DB like this:
```python
ds = Dataset.from_sql('''SELECT * from states WHERE state=="New York";''', "sqlite:///us_covid_data.db")
```
You run into the error I described above:
```ValueError: Please pass `features` or at least one example when writing data```
However, if you try to pass features, as the error suggests, then you get an error that tells you the underlying problem...
```python
from datasets import Dataset, Features, Value
features = Features({
'date': Value('date32'),
'label': Value('string'),
'fips': Value('int32'),
'cases': Value('int32'),
'deaths': Value('int32')
})
ds = Dataset.from_sql(
'''SELECT * from states WHERE state=="New York";''',
"sqlite:///us_covid_data.db",
features=features
)
```
Which results in the actual underlying error: `ImportError: Using URI string without sqlalchemy installed.`
### Expected behavior
Instead of `ValueError` about needing to pass features, we should provide the actual underlying error about not having SQLAlchemy installed when it isn't found in the environment.
### Environment info
- `datasets` version: 2.6.1
- Platform: macOS-10.16-x86_64-i386-64bit
- Python version: 3.8.10
- PyArrow version: 10.0.0
- Pandas version: 1.2.5 | 5,186 |
https://github.com/huggingface/datasets/issues/5185 | Allow passing a subset of output features to Dataset.map | [] | ### Feature request
Currently, map does one of two things to the features (if I'm not mistaken):
* when you do not pass features, types are assumed to be equal to the input if they can be cast, and inferred otherwise
* when you pass a full specification of features, output features are set to this
However, sometimes you want to just pass some of the output types, particularly when the first of these modes makes an incorrect type. This currently crashes.
### Motivation
To give a little background: this problem appears in converting labels to ids, where the labels happen to be floats rather than strings
Consider the following use of map to convert from float to int
```python
data = Dataset.from_dict({'y':[1.0,2.0,3.0]})
mapped = data.map(lambda r: {'y': int(r['y'])})
mapped['y'] # is floats, not ints
```
The result is a float again, since after the mapping operation it forces the old datatypes back on the data.
Passing `features=Features({"y": Value(dtype="int64")})` to map works in principle, but then extending it a little to e.g.
```python
def format_data(r):
return {**tokenizer(r["text"]), "y": int(r["y"])}
data = Dataset.from_dict({"y": [1.0, 2.0, 3.0], "text": ["one", "two", "three"]})
mapped = data.map(
format_data,
features=Features({'y': Value(dtype="int64")}),
remove_columns=["text"],
)
```
Results in a crash in dataset internals, as it expects either all or no output features to be specified.
Of course one can pass a full feature specification, but this becomes tokenizer specific and very awkward.
### Your contribution
I've looked at `write_batch` and particularly `col_type = features[col] if features else None`, but checking for `col in features` here makes it fail elsewhere, but the structure makes it hard to understand how and why. I do not think I would have the time myself to get to the bottom of this anytime soon. | 5,185 |
https://github.com/huggingface/datasets/issues/5183 | Loading an external dataset in a format similar to conll2003 | [] | I'm trying to load a custom dataset in a Dataset object, it's similar to conll2003 but with 2 columns only (word entity), I used the following script:
features = datasets.Features(
{"tokens": datasets.Sequence(datasets.Value("string")),
"ner_tags": datasets.Sequence(
datasets.features.ClassLabel(
names=["B-PER", .... etc.]))}
)
from datasets import Dataset
INPUT_COLUMNS = "tokens ner_tags".split(" ")
def read_conll(file):
#all_labels = []
example = {col: [] for col in INPUT_COLUMNS}
idx = 0
with open(file) as f:
for line in f:
if line:
if line.startswith("-DOCSTART-") and example["tokens"] != []:
print(idx, example)
yield idx, example
idx += 1
example = {col: [] for col in INPUT_COLUMNS}
elif line == "\n" or (line.startswith("-DOCSTART-") and example["tokens"] == []):
continue
else:
row_cols = line.split(" ")
for i, col in enumerate(example):
example[col] = row_cols[i].rstrip()
dset = Dataset.from_generator(read_conll, gen_kwargs={"file": "/content/new_train.txt"}, features = features)
The following error happened:
[/usr/local/lib/python3.7/dist-packages/datasets/utils/py_utils.py](https://localhost:8080/#) in <genexpr>(.0)
285 for key in unique_values(itertools.chain(*dicts)): # set merge all keys
286 # Will raise KeyError if the dict don't have the same keys
--> 287 yield key, tuple(d[key] for d in dicts)
288
TypeError: tuple indices must be integers or slices, not str
What does this mean and what should I modify? | 5,183 |
https://github.com/huggingface/datasets/issues/5182 | Add notebook / other resource links to the task-specific data loading guides | [
"Yea this would be great! We would need an object detection tutorial notebook too if it doesn't already exist there. ",
"There is one: https://huggingface.co/docs/datasets/object_detection.\r\n\r\nI will start the work. "
] | Does it make sense to include links to notebooks / scripts that show how to use a dataset for training / fine-tuning a model?
For example, here in [https://huggingface.co/docs/datasets/image_classification] we could include a mention of https://github.com/huggingface/notebooks/blob/main/examples/image_classification.ipynb.
Applies to https://huggingface.co/docs/datasets/object_detection as well.
Cc: @osanseviero @nateraw | 5,182 |
https://github.com/huggingface/datasets/issues/5181 | Add a guide for semantic segmentation | [
"Sure this sounds great! Would this be pure torchvision, albumentations, or something else?",
"I am considering `torchvision` and `albumentations`. Also [works with TensorFlow](https://github.com/deep-diver/segformer-tf-transformers/blob/main/notebooks/TFSegFormer_Finetune.ipynb). \r\n\r\nI am assigning the issue... | Currently, we have these guides for object detection and image classification:
* https://huggingface.co/docs/datasets/object_detection
* https://huggingface.co/docs/datasets/image_classification
I am proposing adding a similar guide for semantic segmentation.
I am happy to contribute a PR for it.
Cc: @osanseviero @nateraw | 5,181 |
https://github.com/huggingface/datasets/issues/5180 | An example or recommendations for creating large image datasets? | [
"The beam utilities allow to prepare a dataset as parquet in your cloud storage. From my perspective this CLI is not super easy to use, but we've been working on a new python API to prepare a dataset in your cloud storage:\r\n```python\r\nfrom datasets import load_dataset_builder\r\n\r\nbuilder = load_dataset_build... | I know that Apache Beam and `datasets` have [some connector utilities](https://huggingface.co/docs/datasets/beam). But it's a little unclear what we mean by "But if you want to run your own Beam pipeline with Dataflow, here is how:". What does that pipeline do?
As a user, I was wondering if we have this support for creating large image datasets. If so, we should mention that [here](https://huggingface.co/docs/datasets/image_dataset).
Cc @lhoestq | 5,180 |
https://github.com/huggingface/datasets/issues/5179 | `map()` fails midway due to format incompatibility | [
"Cc: @lhoestq ",
"You can end up with a list instead of a tensor if all the tensors inside the list can't be stacked together - can you make sure all your inputs are tensors with the same shape ?",
"Is there an easy way to ensure it?",
"You can make sure your `tokenize` function always return tensors of the s... | ### Describe the bug
I am using the `emotion` dataset from Hub for sequence classification. After training the model, I am using it to generate predictions for all the entries present in the `validation` split of the dataset.
```py
def get_test_accuracy(model):
def fn(batch):
inputs = {k:v.to(device) for k,v in batch.items()
if k in tokenizer.model_input_names}
with torch.no_grad():
output = model(**inputs)
pred_label = torch.argmax(output.logits, axis=-1)
return {"predicted_label": pred_label.cpu().numpy()}
return fn
```
This is how the `get_test_accuracy()` is being used:
```py
emotions = load_dataset("emotion")
def tokenize(batch):
return tokenizer(batch["text"], padding=True, truncation=True)
emotions_encoded = emotions.map(tokenize, batched=True)
emotions_encoded.set_format("torch",
columns=["input_ids", "attention_mask", "label"])
new_dataset = emotions_encoded["validation"].map(
accuracy_fn, batched=True, batch_size=128
)
```
Complete code is available in the Colab Notebook provided below.
The `map()` process fails midway giving:
```shell
AttributeError Traceback (most recent call last)
<ipython-input-8-ad24ac288eb4> in <module>
2
3 new_dataset = emotions_encoded["validation"].map(
----> 4 accuracy_fn, batched=True, batch_size=128
5 )
7 frames
/usr/local/lib/python3.7/dist-packages/datasets/arrow_dataset.py in map(self, function, with_indices, with_rank, input_columns, batched, batch_size, drop_last_batch, remove_columns, keep_in_memory, load_from_cache_file, cache_file_name, writer_batch_size, features, disable_nullable, fn_kwargs, num_proc, suffix_template, new_fingerprint, desc)
2588 new_fingerprint=new_fingerprint,
2589 disable_tqdm=disable_tqdm,
-> 2590 desc=desc,
2591 )
2592 else:
/usr/local/lib/python3.7/dist-packages/datasets/arrow_dataset.py in wrapper(*args, **kwargs)
582 self: "Dataset" = kwargs.pop("self")
583 # apply actual function
--> 584 out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
585 datasets: List["Dataset"] = list(out.values()) if isinstance(out, dict) else [out]
586 for dataset in datasets:
/usr/local/lib/python3.7/dist-packages/datasets/arrow_dataset.py in wrapper(*args, **kwargs)
549 }
550 # apply actual function
--> 551 out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
552 datasets: List["Dataset"] = list(out.values()) if isinstance(out, dict) else [out]
553 # re-apply format to the output
/usr/local/lib/python3.7/dist-packages/datasets/fingerprint.py in wrapper(*args, **kwargs)
478 # Call actual function
479
--> 480 out = func(self, *args, **kwargs)
481
482 # Update fingerprint of in-place transforms + update in-place history of transforms
/usr/local/lib/python3.7/dist-packages/datasets/arrow_dataset.py in _map_single(self, function, with_indices, with_rank, input_columns, batched, batch_size, drop_last_batch, remove_columns, keep_in_memory, load_from_cache_file, cache_file_name, writer_batch_size, features, disable_nullable, fn_kwargs, new_fingerprint, rank, offset, disable_tqdm, desc, cache_only)
2970 indices,
2971 check_same_num_examples=len(input_dataset.list_indexes()) > 0,
-> 2972 offset=offset,
2973 )
2974 except NumExamplesMismatchError:
/usr/local/lib/python3.7/dist-packages/datasets/arrow_dataset.py in apply_function_on_filtered_inputs(inputs, indices, check_same_num_examples, offset)
2850 if with_rank:
2851 additional_args += (rank,)
-> 2852 processed_inputs = function(*fn_args, *additional_args, **fn_kwargs)
2853 if update_data is None:
2854 # Check if the function returns updated examples
<ipython-input-6-4e0d280426f6> in fn(batch)
1 def get_test_accuracy(model):
2 def fn(batch):
----> 3 inputs = {k:v.to(device) for k,v in batch.items()
4 if k in tokenizer.model_input_names}
5 with torch.no_grad():
<ipython-input-6-4e0d280426f6> in <dictcomp>(.0)
2 def fn(batch):
3 inputs = {k:v.to(device) for k,v in batch.items()
----> 4 if k in tokenizer.model_input_names}
5 with torch.no_grad():
6 output = model(**inputs)
AttributeError: 'list' object has no attribute 'to'
```
As you'd notice in the notebook, the process fails _midway_ and not at the beginning.
Is this expected?
### Steps to reproduce the bug
Colab Notebook:
https://colab.research.google.com/gist/sayakpaul/d1570d537faf39040d02d77b1ed7de07/scratchpad.ipynb
### Expected behavior
The mapping process should complete as is. If you switch the `split` to `test` it works as expected.
### Environment info
Colab | 5,179 |
https://github.com/huggingface/datasets/issues/5178 | Unable to download the Chinese `wikipedia`, the dumpstatus.json not found! | [
"In the dumps page of the wiki (https://dumps.wikimedia.org/zhwiki/), I found the following dumps:\r\n```\r\nIndex of /zhwiki/\r\n[../](https://dumps.wikimedia.org/)\r\n[20220701/](https://dumps.wikimedia.org/zhwiki/20220701/) 21-Aug-2022 01:48 -\r\n[202207... | ### Describe the bug
I tried:
`data = load_dataset('wikipedia', '20220301.zh', beam_runner='DirectRunner')`
and
`data = load_dataset("wikipedia", language="zh", date="20220301", beam_runner='DirectRunner')`
but both got:
`FileNotFoundError: Couldn't find file at https://dumps.wikimedia.org/zhwiki/20220301/dumpstatus.json`
the full report is:
```
FileNotFoundError Traceback (most recent call last)
<ipython-input-13-d07c5021090c> in <module>
1 from datasets import load_dataset
2
----> 3 data = load_dataset("wikipedia", language="zh", date="20220301", beam_runner='DirectRunner')<?, ?it/s]
/opt/conda/lib/python3.8/site-packages/datasets/load.py in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, ignore_verifications, keep_in_memory, save_infos, revision, use_auth_token, task, streaming, **config_kwargs)
1740
1741 # Download and prepare data
-> 1742 builder_instance.download_and_prepare(
1743 download_config=download_config,
1744 download_mode=download_mode,
/opt/conda/lib/python3.8/site-packages/datasets/builder.py in download_and_prepare(self, output_dir, download_config, download_mode, ignore_verifications, try_from_hf_gcs, dl_manager, base_path, use_auth_token, file_format, max_shard_size, storage_options, **download_and_prepare_kwargs)
812 **download_and_prepare_kwargs,
813 }
--> 814 self._download_and_prepare(
815 dl_manager=dl_manager,
816 verify_infos=verify_infos,
/opt/conda/lib/python3.8/site-packages/datasets/builder.py in _download_and_prepare(self, dl_manager, verify_infos, **prepare_splits_kwargs)
1645 options=beam_options,
1646 )
-> 1647 super()._download_and_prepare(
1648 dl_manager, verify_infos=False, pipeline=pipeline, **prepare_splits_kwargs
1649 ) # TODO handle verify_infos in beam datasets
/opt/conda/lib/python3.8/site-packages/datasets/builder.py in _download_and_prepare(self, dl_manager, verify_infos, **prepare_split_kwargs)
881 split_dict = SplitDict(dataset_name=self.name)
882 split_generators_kwargs = self._make_split_generators_kwargs(prepare_split_kwargs)
--> 883 split_generators = self._split_generators(dl_manager, **split_generators_kwargs)
884
885 # Checksums verification
~/.cache/huggingface/modules/datasets_modules/datasets/wikipedia/aa542ed919df55cc5d3347f42dd4521d05ca68751f50dbc32bae2a7f1e167559/wikipedia.py in _split_generators(self, dl_manager, pipeline)
943 info_url = _base_url(lang) + _INFO_FILE
944 # Use dictionary since testing mock always returns the same result.
--> 945 downloaded_files = dl_manager.download_and_extract({"info": info_url})
946
947 xml_urls = []
/opt/conda/lib/python3.8/site-packages/datasets/download/download_manager.py in download_and_extract(self, url_or_urls)
431 extracted_path(s): `str`, extracted paths of given URL(s).
432 """
--> 433 return self.extract(self.download(url_or_urls))
434
435 def get_recorded_sizes_checksums(self):
/opt/conda/lib/python3.8/site-packages/datasets/download/download_manager.py in download(self, url_or_urls)
308
309 start_time = datetime.now()
--> 310 downloaded_path_or_paths = map_nested(
311 download_func,
312 url_or_urls,
/opt/conda/lib/python3.8/site-packages/datasets/utils/py_utils.py in map_nested(function, data_struct, dict_only, map_list, map_tuple, map_numpy, num_proc, parallel_min_length, types, disable_tqdm, desc)
427 num_proc = 1
428 if num_proc <= 1 or len(iterable) < parallel_min_length:
--> 429 mapped = [
430 _single_map_nested((function, obj, types, None, True, None))
431 for obj in logging.tqdm(iterable, disable=disable_tqdm, desc=desc)
/opt/conda/lib/python3.8/site-packages/datasets/utils/py_utils.py in <listcomp>(.0)
428 if num_proc <= 1 or len(iterable) < parallel_min_length:
429 mapped = [
--> 430 _single_map_nested((function, obj, types, None, True, None))
431 for obj in logging.tqdm(iterable, disable=disable_tqdm, desc=desc)
432 ]
/opt/conda/lib/python3.8/site-packages/datasets/utils/py_utils.py in _single_map_nested(args)
329 # Singleton first to spare some computation
330 if not isinstance(data_struct, dict) and not isinstance(data_struct, types):
--> 331 return function(data_struct)
332
333 # Reduce logging to keep things readable in multiprocessing with tqdm
/opt/conda/lib/python3.8/site-packages/datasets/download/download_manager.py in _download(self, url_or_filename, download_config)
335 # append the relative path to the base_path
336 url_or_filename = url_or_path_join(self._base_path, url_or_filename)
--> 337 return cached_path(url_or_filename, download_config=download_config)
338
339 def iter_archive(self, path_or_buf: Union[str, io.BufferedReader]):
/opt/conda/lib/python3.8/site-packages/datasets/utils/file_utils.py in cached_path(url_or_filename, download_config, **download_kwargs)
186 if is_remote_url(url_or_filename):
187 # URL, so get it from the cache (downloading if necessary)
--> 188 output_path = get_from_cache(
189 url_or_filename,
190 cache_dir=cache_dir,
/opt/conda/lib/python3.8/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, ignore_url_params, download_desc)
533 )
534 elif response is not None and response.status_code == 404:
--> 535 raise FileNotFoundError(f"Couldn't find file at {url}")
536 _raise_if_offline_mode_is_enabled(f"Tried to reach {url}")
537 if head_error is not None:
FileNotFoundError: Couldn't find file at https://dumps.wikimedia.org/zhwiki/20220301/dumpstatus.json
```
### Steps to reproduce the bug
`data = load_dataset('wikipedia', '20220301.zh', beam_runner='DirectRunner')`
### Expected behavior
download the data
### Environment info
python3.6
latest datasets/transformers version | 5,178 |
https://github.com/huggingface/datasets/issues/5176 | prepare dataset for cloud storage doesn't work | [
"It looks like an issue with `gcsfs`, are you able to instantiate a `GCSFileSystem` manually ?",
"closing since it was probably due to gcsfs"
] | ### Describe the bug
Following the [documentation](https://huggingface.co/docs/datasets/filesystems#load-and-save-your-datasets-using-your-cloud-storage-filesystem) and [this PR](https://github.com/huggingface/datasets/pull/4724), I was downloading and storing huggingface dataset to cloud storage.
```
from datasets import load_dataset, load_dataset_builder
dataset = load_dataset_builder("wikipedia", "20220301.en", cache_dir='LOCAL_PATH')
dataset.download_and_prepare("gs://Bucket_NAME", file_format="parquet")
```
The above code successfully downloaded dataset, however, it returns error from `download_and_prepare`.
> Traceback (most recent call last):
> File "/shared/zhuiai/research/wiki/wiki/gcsfs.py", line 12, in <module>
> dataset.download_and_prepare("gs://upgen/dataset/wiki", file_format="parquet")
> File "/shared/zhuiai/.conda/envs/wiki/lib/python3.9/site-packages/datasets/builder.py", line 671, in download_and_prepare
> fs_token_paths = fsspec.get_fs_token_paths(output_dir, storage_options=storage_options)
> File "/shared/zhuiai/.conda/envs/wiki/lib/python3.9/site-packages/fsspec/core.py", line 635, in get_fs_token_paths
> cls = get_filesystem_class(protocol)
> File "/shared/zhuiai/.conda/envs/wiki/lib/python3.9/site-packages/fsspec/registry.py", line 234, in get_filesystem_class
> register_implementation(protocol, _import_class(bit["class"]))
> File "/shared/zhuiai/.conda/envs/wiki/lib/python3.9/site-packages/fsspec/registry.py", line 257, in _import_class
> mod = importlib.import_module(mod)
> File "/shared/zhuiai/.conda/envs/wiki/lib/python3.9/importlib/__init__.py", line 127, in import_module
> return _bootstrap._gcd_import(name[level:], package, level)
> File "<frozen importlib._bootstrap>", line 1030, in _gcd_import
> File "<frozen importlib._bootstrap>", line 1007, in _find_and_load
> File "<frozen importlib._bootstrap>", line 986, in _find_and_load_unlocked
> File "<frozen importlib._bootstrap>", line 680, in _load_unlocked
> File "<frozen importlib._bootstrap_external>", line 850, in exec_module
> File "<frozen importlib._bootstrap>", line 228, in _call_with_frames_removed
> File "/shared/zhuiai/research/wiki/wiki/gcsfs.py", line 12, in <module>
> dataset.download_and_prepare("gs://upgen/dataset/wiki", file_format="parquet")
> File "/shared/zhuiai/.conda/envs/wiki/lib/python3.9/site-packages/datasets/builder.py", line 671, in download_and_prepare
> fs_token_paths = fsspec.get_fs_token_paths(output_dir, storage_options=storage_options)
> File "/shared/zhuiai/.conda/envs/wiki/lib/python3.9/site-packages/fsspec/core.py", line 635, in get_fs_token_paths
> cls = get_filesystem_class(protocol)
> File "/shared/zhuiai/.conda/envs/wiki/lib/python3.9/site-packages/fsspec/registry.py", line 234, in get_filesystem_class
> register_implementation(protocol, _import_class(bit["class"]))
> File "/shared/zhuiai/.conda/envs/wiki/lib/python3.9/site-packages/fsspec/registry.py", line 258, in _import_class
> return getattr(mod, name)
> AttributeError: partially initialized module 'gcsfs' has no attribute 'GCSFileSystem' (most likely due to a circular import)
### Steps to reproduce the bug
1. pip install datasets==2.6.1 gcsfs==2022.8.2
2. Run the following code will reproduce the issue (change `LOCAL_PATH` and `Bucket_NAME` accordingly)
```
from datasets import load_dataset, load_dataset_builder
dataset = load_dataset_builder("wikipedia", "20220301.en", cache_dir='LOCAL_PATH')
dataset.download_and_prepare("gs://Bucket_NAME", file_format="parquet")
```
### Expected behavior
Expecting successful downloading dataset and uploading it to cloud storage.
### Environment info
- `datasets` version: 2.6.1
- Platform: Linux-5.15.0-25-generic-x86_64-with-glibc2.35
- Python version: 3.9.12
- PyArrow version: 7.0.0
- Pandas version: 1.5.1 | 5,176 |
https://github.com/huggingface/datasets/issues/5175 | Loading an external NER dataset | [] | I need to use huggingface datasets to load a custom dataset similar to conll2003 but with more entities and each the files contain only two columns: word and ner tag.
I tried this code snnipet that I found here as an answer to a similar issue:
from datasets import Dataset
INPUT_COLUMNS = "ID Text NER".split()
def read_conll(file):
example = {col: [] for col in INPUT_COLUMNS}
idx = 0
with open(file) as f:
for line in f:
if line.startswith("-DOCSTART-") or line == "\n" or not line:
if example[next(iter(example))]:
yield idx, example
idx += 1
example = {col: [] for col in INPUT_COLUMNS}
else:
row_cols = line.split()
for i, col in enumerate(example):
example[col] = row_cols[i].rstrip()
train = Dataset.from_generator(read_conll, gen_kwargs={"file": "some_path"})
But the following error happened:
ValueError: Please pass `features` or at least one example when writing data | 5,175 |
https://github.com/huggingface/datasets/issues/5172 | Inconsistency behavior between handling local file protocol and other FS protocols | [] | ### Describe the bug
These lines us used during load_from_disk:
```
if is_remote_filesystem(fs):
dest_dataset_dict_path = extract_path_from_uri(dataset_dict_path)
else:
fs = fsspec.filesystem("file")
dest_dataset_dict_path = dataset_dict_path
```
If a local FS is given, then it will the URL as the path name. If a remote Fs is given, then it will use the path of the URL. This is an inconsistent behavior when handling a file: when using remote FS, you must write a URL, but for local FS, even if you passed LocalFileSystem as `fs` you still can't use a `file://` URL. It will be recognized as a directory named `file:`.
### Steps to reproduce the bug
```
import fsspec.core
url = "hdfs:///somewhere/MNIST"
# url = "file:///somewhere/MNIST"
fs, path = fsspec.core.url_to_fs(url)
fs.ls(path) # this will always work
load_from_disk(path, fs) # only works for local FS
load_from_disk(url, fs) # only works for remote FS
```
### Expected behavior
one of `url` or `path` should always work
I think we extract path from given URL by using `fsspec.core.url_to_fs` instead of using `is_remote_filesystem` and `extract_path_from_uri` will fix this, since:
```
fsspec.core.url_to_fs("/somewhere/MNIST") -> LocalFs, '/somewhere/MNIST'
fsspec.core.url_to_fs("file:///somewhere/MNIST") -> LocalFs, '/somewhere/MNIST'
fsspec.core.url_to_fs("hdfs:///somewhere/MNIST") -> HDFS, '/somewhere/MNIST'
```
and
```
fsspec.core.url_to_fs("file:///somewhere/MNIST") == fsspec.core.url_to_fs("/somewhere/MNIST")
```
In theory, this wouldn't break anything, since giving local path and remote uri still works. It will only affect local URI (make it works too)
### Environment info
- `datasets` version: 2.5.1
- Platform: Linux-5.4.205.1**HIDDEN**
- Python version: 3.7.10
- PyArrow version: 8.0.0
- Pandas version: 1.2.4
| 5,172 |
https://github.com/huggingface/datasets/issues/5170 | [Caching] Deterministic hashing of torch tensors | [] | Currently this fails
```python
import torch
from datasets.fingerprint import Hasher
t = torch.tensor([1.])
def func(x):
return t + x
hash1 = Hasher.hash(func)
t = torch.tensor([1.])
hash2 = Hasher.hash(func)
assert hash1 == hash2
```
Also as noticed in https://discuss.huggingface.co/t/dataset-cant-cache-models-outputs/24945, using a model in a `map` function doesn't work well with caching. Indeed the `bert-base-uncased` model has a different hash every time you reload it. Supporting torch tensors may also help in this case.
This can be fixed by registering a custom pickling functions for torch tensors - as we did for other objects such as CodeType, FunctionType and Regex in `py_utils.py` | 5,170 |
https://github.com/huggingface/datasets/issues/5165 | Memory explosion when trying to access 4d tensors in datasets cast to torch or np | [] | ### Describe the bug
When trying to access an item by index, in a datasets.Dataset cast to torch/np using `set_format` or `with_format`, we get a memory explosion if the item contains 4d (or above) tensors.
### Steps to reproduce the bug
MWE:
```python
from datasets import load_dataset
import numpy as np
def create_4d_tensor(item):
i = item["num_nodes"]
item["x_big"] = np.random.rand(i, 2*i, int(i/2), 1) + 1 # we create a big 4d tensor
return item
if __name__ == "__main__":
dataset = load_dataset(path=f"graphs-datasets/PROTEINS")
# This works
print(dataset["train"].format)
print(dataset["train"][0].keys())
dataset = dataset.map(
create_4d_tensor,
batched=False,
writer_batch_size=100,
)
# This works
print(dataset["train"].format)
print(dataset["train"][0].keys())
dataset.set_format("torch")
print(dataset["train"].format)
# This gets killed :(
print(dataset["train"][0].keys())
```
The problem likely comes from `format_table` [here](https://cs.github.com/huggingface/datasets/blob/f09f781be3278156ce3aa6ec90c1926b1846a78f/src/datasets/arrow_dataset.py#L2328)
### Expected behavior
No memory explosion when trying to access dataset items after cast.
### Environment info
- `datasets` version: 2.3.2
- Platform: Linux-5.14.0-1054-oem-x86_64-with-glibc2.29
- Python version: 3.8.10
- PyArrow version: 8.0.0
- Pandas version: 1.4.3 | 5,165 |
https://github.com/huggingface/datasets/issues/5162 | Pip-compile: Could not find a version that matches dill<0.3.6,>=0.3.6 | [
"Thanks for reporting, @Rijgersberg.\r\n\r\nWe were waiting for the release of `dill` 0.3.6, that happened 2 days ago (24 Oct 2022): https://github.com/uqfoundation/dill/releases/tag/dill-0.3.6\r\n- See comment: https://github.com/huggingface/datasets/pull/4397#discussion_r880629543\r\n\r\nAlso `multiprocess` 0.70.... | ### Describe the bug
When using `pip-compile` (part of `pip-tools`) to generate a pinned requirements file that includes `datasets`, a version conflict of `dill` appears.
It is caused by a transitive dependency conflict between `datasets` and `multiprocess`.
### Steps to reproduce the bug
```bash
$ echo "datasets" > requirements.in
$ pip install pip-tools
$ pip-compile requirements.in
Could not find a version that matches dill<0.3.6,>=0.3.6 (from datasets==2.6.1->-r requirements.in (line 1))
Tried: 0.2, 0.2, 0.2.1, 0.2.1, 0.2.2, 0.2.2, 0.2.3, 0.2.3, 0.2.4, 0.2.4, 0.2.5, 0.2.5, 0.2.6, 0.2.7, 0.2.7.1, 0.2.8, 0.2.8.1, 0.2.8.2, 0.2.9, 0.3.0, 0.3.1, 0.3.1.1, 0.3.2, 0.3.3, 0.3.3, 0.3.4, 0.3.4, 0.3.5, 0.3.5, 0.3.5.1, 0.3.5.1, 0.3.6, 0.3.6
Skipped pre-versions: 0.1a1, 0.2a1, 0.2a1, 0.2b1, 0.2b1
There are incompatible versions in the resolved dependencies:
dill<0.3.6 (from datasets==2.6.1->-r requirements.in (line 1))
dill>=0.3.6 (from multiprocess==0.70.14->datasets==2.6.1->-r requirements.in (line 1))
```
### Expected behavior
A correctly generated file `requirements.txt` with pinned dependencies
### Environment info
Tested with versions `2.6.1, 2.6.0, 2.5.2` on Python 3.8 and 3.10 on Ubuntu 20.04LTS and Python 3.10 on MacOS 12.6 (M1). | 5,162 |
https://github.com/huggingface/datasets/issues/5161 | Dataset can’t cache model’s outputs | [
"Addressed in https://github.com/huggingface/datasets/pull/5191 (torch.Tensor objects now produce deterministic hashes)"
] | ### Describe the bug
Hi,
I try to cache some outputs of teacher model( Knowledge Distillation ) by using map function of Dataset library, while every time I run my code, I still recompute all the sequences. I tested Bert Model like this, I got different hash every single run, so any idea to deal with this?
### Steps to reproduce the bug
1. run below code
2. get different hash
```
from transformers import BertModel
from transformers import AutoTokenizer
import torch
token = ['hello']
model = BertModel.from_pretrained("bert-base-uncased").eval()
tok = AutoTokenizer.from_pretrained("bert-base-uncased")
def abcd():
with torch.no_grad():
out = model(**tok(token,return_tensors='pt'))[0]
# out = tok(token)
return out
from datasets.fingerprint import Hasher
my_func = abcd
print(Hasher.hash(my_func))
print(abcd())
```
### Expected behavior
I wanna cache all the model output
### Environment info
datasets:2.5.0 | 5,161 |
https://github.com/huggingface/datasets/issues/5160 | Automatically add filename for image/audio folder | [
"Also cc @anton-l ",
"BTW the exact same holds true for the audio folder",
"I'm fine with adding a new column with the file name personally. Not sure how breaking this is though",
"@patrickvonplaten do you mean just filename or full relative path inside the repo?\r\nI think it shouldn't be breaking, at least ... | ### Feature request
When creating a custom audio of image dataset, it would be great to automatically have access to the filename. It should be both:
a) Automatically displayed in the viewer
b) Automatically added as a column to the dataset when doing `load_dataset`
In `diffusers` our test rely quite heavily on images and audio files now and it's a bit tedious at the moment to download specific images from a datasets repo.
E.g. we have a dataset of images for tests in `diffusers`: https://huggingface.co/datasets/hf-internal-testing/diffusers-images
where it would be extremely nice to have direct access to the filename both visually on the datasets page (@severo ) as well as via the `load_datasets` function. We currently have some akward functionality to download images by path name: https://github.com/huggingface/diffusers/blob/2fb8fafa4b761f6fc144cf75a6f6f0ea6af3a1c1/src/diffusers/utils/testing_utils.py#L131
It would be much nicer to just go over `load_dataset(...)`
### Motivation
Intuitively the filename is something people understand directly. E.g if you upload a folder of images online, it's nice if you recognize the image as well as the filename next to it directly and that you're able to use it right away.
The label on the other hand is less intuitive to understand as you haven't added it yourself.
### Your contribution
Not sure if I have the time to add it myself anytime soon, but it would help us a lot for `diffusers`. | 5,160 |
https://github.com/huggingface/datasets/issues/5158 | Fix language and license tag names in all Hub datasets | [
"There are currently 402 datasets with deprecated \"languages\" or \"licenses\".",
"hey @albertvillanova ,i would love to work on this issue if you like.",
"Hi @ayushthe1, thanks for your offer.\r\n\r\nBut as you can see, I self-assigned this issue.\r\n\r\nI have already fixed 200 out of the 402 datasets. My sc... | While working on this:
- #5137
we realized there are still many datasets with deprecated "languages" and "licenses" tag names (instead of "language" and "license").
This is a blocking issue: no subsequent PR can be opened to modify their metadata: a ValueError will be thrown.
We should fix the "language" and "license" tag names in all Hub datasets.
TODO:
- [x] Fix language and license tag names in 402 Hub datasets
CC: @julien-c | 5,158 |
https://github.com/huggingface/datasets/issues/5157 | Consistent caching between python and jupyter | [
"Hi ! Maybe it's possible to have a consistent hash for a function defined in `__main__` and a function define in a notebook.\r\n\r\nHowever for functions imported from another location, pickle uses the location to identify the code, so in that case we can't do much I believe.\r\n\r\nWould it be ok for you if we on... | ### Feature request
I hope this is not my mistake, currently if I use `load_dataset` from a python session on a custom dataset to do the preprocessing, it will be saved in the cache and in other python sessions it will be loaded from the cache, however calling the same from a jupyter notebook does not work, meaning the preprocessing starts from scratch.
If adjusting the hashes is impossible, is there a way to manually set dataset fingerprint to "force" this behaviour?
### Motivation
If this is not already the case and I am doing something wrong, it would be useful to have the two fingerprints consistent so one can create the dataset once and then try small things on jupyter without preprocessing everything again.
### Your contribution
I am happy to try a PR if you give me some pointers where the changes should happen | 5,157 |
https://github.com/huggingface/datasets/issues/5156 | Unable to download dataset using Azure Data Lake Gen 2 | [
"Hi ! From the `adlfs` docs, there are two filesystems you can use:\r\n> To use the Gen1 filesystem:\r\n> - known_implementations[‘adl’] = {‘class’: ‘adlfs.AzureDatalakeFileSystem’}\r\n> \r\n> To use the Gen2 filesystem:\r\n> - known_implementations[‘abfs’] = {‘class’: ‘adlfs.AzureBlobFileSystem’}\r\n\r\nIf I'm no... | ### Describe the bug
When using the DatasetBuilder method with the credentials for the cloud storage Azure Data Lake (adl) Gen2, the following error is showed:
```
Traceback (most recent call last):
File "download_hf_dataset.py", line 143, in <module>
main()
File "download_hf_dataset.py", line 102, in main
builder.download_and_prepare(save_dir, storage_options=storage_options, max_shard_size="250MB", file_format="parquet")
File "/home/clarisses/miniconda3/envs/hf_datasets_env/lib/python3.8/site-packages/datasets/builder.py", line 671, in download_and_prepare
fs_token_paths = fsspec.get_fs_token_paths(output_dir, storage_options=storage_options)
File "/home/clarisses/miniconda3/envs/hf_datasets_env/lib/python3.8/site-packages/fsspec/core.py", line 639, in get_fs_token_paths
fs = cls(**options)
File "/home/clarisses/miniconda3/envs/hf_datasets_env/lib/python3.8/site-packages/fsspec/spec.py", line 76, in __call__
obj = super().__call__(*args, **kwargs)
TypeError: __init__() got an unexpected keyword argument 'account_name'
```
If I don't pass the storage_options argument (leave it as None), it requires the credentials used in ADL Gen 1:
`TypeError: __init__() missing 3 required positional arguments: 'tenant_id', 'client_id', and 'client_secret'`
Thus, it is not possible to download a dataset from the cloud using Azure Data Lake (adl) Gen2.
### Steps to reproduce the bug
Assuming that you have an account on Azure and at Storage Account that can be used for reproduce:
1. Create a dict with the format to connect to Azure Data Lake Gen 2
```
storage_options = {"account_name": ACCOUNT_NAME, "account_key": ACCOUNT_KEY) # gen 2 filesystem
```
2. Create a dataset builder for any HF hosted dataset
```
builder = load_dataset_builder(dataset_name)
```
3. Try to download the dataset passing the storage_options as an argument
```
save_dir = 'adl://my_save_dir'
builder.download_and_prepare(save_dir, storage_options=storage_options, max_shard_size="250MB", file_format="parquet")
```
### Expected behavior
Not seeing the error mentioned above and being able to download the dataset to the provided path on ADL
### Environment info
- `datasets` version: 2.6.1
- Platform: Linux-5.15.0-46-generic-x86_64-with-glibc2.17
- Python version: 3.8.13
- PyArrow version: 9.0.0
- Pandas version: 1.5.1 | 5,156 |
https://github.com/huggingface/datasets/issues/5153 | default Image/AudioFolder infers labels when there is no metadata files even if there is only one dir | [
"Makes sense! For the last structure, we could count the path segments (delimited by \"/\" for URLs and `os.sep` for local paths) to ensure all inferred labels are on the same level. Otherwise, I think it's safe to assume they are meaningless and ignore them.\r\n"
] | ### Describe the bug
By default FolderBasedBuilder infers labels if there is not metadata files, even if it's meaningless (for example, they are in a single directory or in the root folder, see this repo as an example: https://huggingface.co/datasets/patrickvonplaten/audios
As this is a corner case for quick exploration of images or audios on the Hub.
### Steps to reproduce the bug
If you have directory like this:
```
repo
image1.jpg
image2.jpg
image3.jpg
```
or
```
repo
data
image1.jpg
image2.jpg
image3.jpg
```
doing `ds = load_dataset(repo)` would create `label` feature:
```python
print(ds["train"][0])
>> {'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=500x375 at 0x7FB5326468E0>, 'label': 0}
```
Also, if you have the following structure:
```
repo
data
image1.jpg
image2.jpg
image3.jpg
image4.jpg
image5.jpg
image6.jpg
```
it will infer two labels:
```python
print(ds["train"][0])
print(ds["train"][-1])
>> {'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=500x375 at 0x7FB5326468E0>, 'label': 1}
>> {'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=500x415 at 0x7FB5326555B0>, 'label': 0}
```
### Expected behavior
We should have only one base feature (Image/Audio) in such cases.
### Environment info
all versions of `datasets` | 5,153 |
https://github.com/huggingface/datasets/issues/5152 | refactor FolderBasedBuilder and Image/AudioFolder tests | [] | Tests for FolderBasedBuilder, ImageFolder and AudioFolder are mostly duplicating each other. They need to be refactored and Audio/ImageFolder should have only tests specific to the loader. | 5,152 |
https://github.com/huggingface/datasets/issues/5151 | Add support to create different configs with `push_to_hub` (+ inferring configs from directories with package managers?) | [
"also asked in https://discuss.huggingface.co/t/create-multiple-dataset-configs-with-push-to-hub-method/25480"
] | Now one can push only different splits within one default config of a dataset.
Would be nice to allow something like:
```
ds.push_to_hub(repo_name, config=config_name)
```
I'm not sure, but this will probably require changes in `data_files.py` patterns. If so, it would also allow to create different configs for packaged modules datasets.
| 5,151 |
https://github.com/huggingface/datasets/issues/5150 | Problems after upgrading to 2.6.1 | [
"Hi! I can't reproduce the error following these steps. Can you please provide a reproducible example?",
"I faced the same issue:\r\n\r\n### Repro\r\n```\r\n!pip install datasets==2.6.1\r\nimport datasets as Dataset\r\ndataset = Dataset.from_pandas(dataframe)\r\ndataset.save_to_disk(local)\r\n\r\n!pip install dat... | ### Describe the bug
Loading a dataset_dict from disk with `load_from_disk` is now creating a `KeyError "length"` that was not occurring in v2.5.2.
Context:
- Each individual dataset in the dict is created with `Dataset.from_pandas`
- The dataset_dict is create from a dict of `Dataset`s, e.g., `DatasetDict({"train": train_ds, "validation": val_ds})
- The pandas dataframe, besides text columns, has a column with a dictionary inside and potentially different keys in each row. Correctly the `Dataset.from_pandas` function adds `key: None` to all dictionaries in each row so that the schema can be correctly inferred.
### Steps to reproduce the bug
Steps to reproduce:
- Upgrade to datasets==2.6.1
- Create a dataset from pandas dataframe with `Dataset.from_pandas`
- Create a dataset_dict from a dict of `Dataset`s, e.g., `DatasetDict({"train": train_ds, "validation": val_ds})
- Save to disk with the `save` function
### Expected behavior
Same as in v2.5.2, that is load from disk without errors
### Environment info
- `datasets` version: 2.6.1
- Platform: Linux-5.4.209-129.367.amzn2int.x86_64-x86_64-with-glibc2.26
- Python version: 3.9.13
- PyArrow version: 9.0.0
- Pandas version: 1.5.1 | 5,150 |
https://github.com/huggingface/datasets/issues/5148 | Cannot find the rvl_cdip dataset | [
"Hi, @santule.\r\n\r\nWe have transferred all dataset scripts from GitHub to the Hugging Face Hub: https://huggingface.co/datasets\r\n- Concretely, you have \"rvl_cdip\" here: https://huggingface.co/datasets/rvl_cdip\r\n\r\nTo be able to load them, you should update your `datasets` library:\r\n```\r\npip install -U... | Hi,
I am trying to use load_dataset to load the official "rvl_cdip" dataset but getting an error.
dataset = load_dataset("rvl_cdip")
Couldn't find 'rvl_cdip' on the Hugging Face Hub either: FileNotFoundError: Couldn't find the file at https://raw.githubusercontent.com/huggingface/datasets/master/datasets/rvl_cdip/rvl_cdip.py
Regards,
| 5,148 |
https://github.com/huggingface/datasets/issues/5147 | Allow ignoring kwargs inside fn_kwargs during dataset.map's fingerprinting | [
"Hi ! In the `transformers` issue the object to not hash is a `Pool` - I think you can instantiate it inside your function instead of passing it as a parameter. It's good practice that your function and all its fn_kwargs are picklable, in case you want to parallelize `map` using `num_proc>1`\r\n\r\nFor the other ca... | ### Feature request
`dataset.map` accepts a `fn_kwargs` that is passed to `fn`. Currently, the whole `fn_kwargs` is used by `fingerprint_transform` to calculate the new fingerprint.
I'd like to be able to inform `fingerprint_transform` which `fn_kwargs` shoud/shouldn't be taken into account during hashing.
Of course, users should be aware to properly use this new feature, just like the internal usages of `fingerprint_transform` [does](https://github.com/huggingface/datasets/blob/2699593b33ee63d17aad2a2bfddedd38a8df57b8/src/datasets/arrow_dataset.py#L2700).
### Motivation
This is originally motivated by https://github.com/huggingface/transformers/pull/18351#issuecomment-1263588680.
Nonetheless, consider a more general processing function that accepts a kwarg that does not influence it's output:
```python
def fn(example, verbose=False):
...
```
Then `dataset.map(fn, verbose=True)` would not benefit from dataset caching.
I'm not sure if other methods in the `Dataset` API could benefit from this feature.
### Your contribution
Based on `fingerprint_transform `'s `wrapper` function [here](https://github.com/huggingface/datasets/blob/c59cc34fcd2a369d27b77cc678017f5976a926a9/src/datasets/fingerprint.py#L443), it seems to me that it should be possible to make `.map`/`._map_single` accept something like `fn_use_fingerprint_kwargs`/`fn_ignore_fingerprint_kwargs` (probably another arg name). This would then be used by `fingerprint_transform.wrapper` to better/more flexibly hash the transformation.
I could contribute with a PR if this feature and approach look good to you. | 5,147 |
https://github.com/huggingface/datasets/issues/5145 | Dataset order is not deterministic with ZIP archives and `iter_files` | [
"Thanks for reporting ! The issue doesn't come from shuffling, but from `beans` row order not being deterministic:\r\n\r\nhttps://huggingface.co/datasets/beans/blob/main/beans.py uses `dl_manager.iter_files` on ZIP archives and the file order doesn't seen to be deterministic and changes across machines",
"Thank y... | ### Describe the bug
For the `beans` dataset (did not try on other), the order of samples is not the same on different machines. Tested on my local laptop, github actions machine, and ec2 instance. The three yield a different order.
### Steps to reproduce the bug
In a clean docker container or conda environment with datasets==2.6.1, run
```python
from datasets import load_dataset
from pprint import pprint
data = load_dataset("beans", split="validation")
pprint(data["image_file_path"])
```
### Expected behavior
The order of the images is the same on all machines.
### Environment info
On the EC2 instance:
```
- `datasets` version: 2.6.1
- Platform: Linux-4.14.291-218.527.amzn2.x86_64-x86_64-with-glibc2.2.5
- Python version: 3.7.10
- PyArrow version: 9.0.0
- Pandas version: 1.3.5
- Numpy version: not checked
```
On my local laptop:
```
- `datasets` version: 2.6.1
- Platform: Linux-5.15.0-50-generic-x86_64-with-glibc2.35
- Python version: 3.9.12
- PyArrow version: 7.0.0
- Pandas version: 1.3.5
- Numpy version: 1.23.1
```
On github actions:
```
- `datasets` version: 2.6.1
- Platform: Linux-5.15.0-1022-azure-x86_64-with-glibc2.2.5
- Python version: 3.8.14
- PyArrow version: 9.0.0
- Pandas version: 1.5.1
- Numpy version: 1.23.4
``` | 5,145 |
https://github.com/huggingface/datasets/issues/5144 | Inconsistent documentation on map remove_columns | [
"Thanks for reporting, @zhaowei-wang-nlp.\r\n\r\nYou are right, the documentation is confusing on the behavior of `remove_columns`. We should better explain it. ",
"This is a duplicate of https://github.com/huggingface/datasets/issues/2343.",
"I'm closing this issue because as @mariosasko pointed out, it is a d... | ### Describe the bug
The page [process](https://huggingface.co/docs/datasets/process) says this about the parameter `remove_columns` of the function `map`:
When you remove a column, it is only removed after the example has been provided to the mapped function.
So it seems that the `remove_columns` parameter removes after the mapped functions.
However, another page, [the documentation of the function map](https://huggingface.co/docs/datasets/v2.6.1/en/package_reference/main_classes#datasets.Dataset.map.remove_columns) says:
Columns will be removed before updating the examples with the output of `function`, i.e. if `function` is adding columns with names in remove_columns, these columns will be kept.
So one page says "after the mapped function" and another says "before the mapped function."
Is there something wrong?
### Steps to reproduce the bug
Not about code.
### Expected behavior
consistent about the descriptions of the behavior of the parameter `remove_columns` in the function `map`.
### Environment info
datasets V2.6.0 | 5,144 |
https://github.com/huggingface/datasets/issues/5143 | DownloadManager Git LFS support | [
"Hey ! Actually it works, just pass the right URL ;)\r\nThe URL must be the one with “/resolve/”\r\n\r\ne.g. https://huggingface.co/datasets/imagenet-1k/resolve/main/data/test_images.tar.gz\r\n\r\nYou can even pass a relative path to the dl_manager instead, like `dl_manager.download(\"data/test_images.tar.gz\")`",
... | ### Feature request
Maybe I'm mistaken but the `DownloadManager` does not support extracting git lfs files out of the box right?
Using `dl_manager.download()` or `dl_manager.download_and_extract()` still returns lfs files afaict.
Is there a good way to write a dataset loading script for a repo with lfs files?
### Motivation
/
### Your contribution
/ | 5,143 |
https://github.com/huggingface/datasets/issues/5137 | Align task tags in dataset metadata | [
"I removed all the invalid task_ids in datasts without namespace, based on the <s>(internal)</s> types.ts",
"(Types.ts is not internal it's public)",
"I have opened PRs to fix the task_ids in all datasets within a namespace as well.\r\n\r\nWorking on task_categories...",
"For future reference: this fix had so... | ## Describe
Once we have agreed on a common naming for task tags for all open source projects, we should align on them.
## Steps
- [x] Align task tags in canonical datasets
- [x] task_categories: 4 datasets
- [x] task_ids (by @lhoestq)
- [x] Open PRs in community datasets
- [x] task_categories: 451 datasets
- [x] task_ids: 556 datasets
| 5,137 |
https://github.com/huggingface/datasets/issues/5135 | Update docs once dataset scripts transferred to the Hub | [] | ## Describe the bug
As discussed in:
- https://github.com/huggingface/hub-docs/pull/423#pullrequestreview-1146083701
we should update our docs once dataset scripts have been transferred to the Hub (and removed from GitHub):
- #4974
Concretely:
- [x] Datasets on GitHub (legacy): https://huggingface.co/docs/datasets/main/en/share#datasets-on-github-legacy
- [x] ADD_NEW_DATASET: https://github.com/huggingface/datasets/blob/main/ADD_NEW_DATASET.md
- ...
This PR complements the work of:
- #5067
This PR is a follow-up of PRs:
- #3777
CC: @julien-c | 5,135 |
https://github.com/huggingface/datasets/issues/5134 | Raise ImportError instead of OSError if required extraction library is not installed | [
"hey ,i would like to work on this issue . Please assign it to me.",
"hey @mariosasko , i made a pr for this issue. Could you please review it.\r\nAlso i found multiple `OSError` in `extract.py` file which i thought could be replaced too but wasn't sure about them.\r\nPlease do tell if that also needs to be done.... | According to the official Python docs, `OSError` should be thrown in the following situations:
> This exception is raised when a system function returns a system-related error, including I/O failures such as “file not found” or “disk full” (not for illegal argument types or other incidental errors).
Hence, it makes more sense to raise `ImportError` instead of `OSError` when the required extraction/decompression library is not installed. | 5,134 |
https://github.com/huggingface/datasets/issues/5133 | Tensor operation not functioning in dataset mapping | [
"Hi! The Torch ops in your snippet are not equivalent to the NumPy ones, hence the difference. You can get the same behavior by replacing the line `feature = torch.mean(feature, dim=1)` with `feature = feature.squeeze().mean(1)` .",
"> Hi! The Torch ops in your snippet are not equivalent to the NumPy ones, hence ... | ## Describe the bug
I'm doing a torch.mean() operation in data preprocessing, and it's not working.
## Steps to reproduce the bug
```
from transformers import pipeline
import torch
import numpy as np
from datasets import load_dataset
device = 'cuda:0'
raw_dataset = load_dataset("glue", "sst2")
feature_extraction = pipeline('feature-extraction', 'bert-base-uncased', device=device)
def extracted_data(examples):
# feature = torch.tensor(feature_extraction(examples['sentence'], batch_size=16), device=device)
# feature = torch.mean(feature, dim=1)
feature = np.asarray(feature_extraction(examples['sentence'], batch_size=16)).squeeze().mean(1)
print(feature.shape)
return {'feature': feature}
extracted_dataset = raw_dataset.map(extracted_data, batched=True, batch_size=16)
```
## Results
When running with torch.mean(), the shape printed out is [16, seq_len, 768], which is exactly the same before the operation. While numpy works just fine, which gives [16, 768].
## Environment info
- `datasets` version: 2.6.1
- Platform: Linux-4.4.0-142-generic-x86_64-with-glibc2.31
- Python version: 3.10.6
- PyArrow version: 9.0.0
- Pandas version: 1.5.0
| 5,133 |
https://github.com/huggingface/datasets/issues/5132 | Depracate `num_proc` parameter in `DownloadManager.extract` | [
"I can take this! #self-assign",
"#self-assign",
"@lazarust i'm already working on this issue :smile: ",
"#self-assign",
"hey @mariosasko , i made a pr for this issue. Could you please review it."
] | The `num_proc` parameter is only present in `DownloadManager.extract` but not in `StreamingDownloadManager.extract`, making it impossible to support streaming in the dataset scripts that use it (`openwebtext` and `the_pile_stack_exchange`). We can avoid this situation by deprecating this parameter and passing `DownloadConfig`'s `num_proc` to `map_nested` instead, as it's done in `DownloadManager.download`. | 5,132 |
https://github.com/huggingface/datasets/issues/5131 | WikiText 103 tokenizer hangs | [
"any updates on this? It happens to me on [OpenWikiText-20%](https://huggingface.co/datasets/Bingsu/openwebtext_20p) dataset, but not on [OpenWebText-10k](https://huggingface.co/datasets/stas/openwebtext-10k). This is really strange because I don't change anything else in my running script.\r\n\r\ntransformers vers... | See issue here: https://github.com/huggingface/transformers/issues/19702 | 5,131 |
https://github.com/huggingface/datasets/issues/5129 | unexpected `cast` or `class_encode_column` result after `rename_column` | [
"Hi! Unfortunately, I can't reproduce this issue locally (in Python 3.7/3.10) or in Colab. I would assume this is due to a bug we fixed in the latest release, but your version is up-to-date, so I'm not sure if there is something we can do to help...",
"Hi, 方子东. I tried running the code with exact the same configu... | ## Describe the bug
When invoke `cast` or `class_encode_column` to a colunm renamed by `rename_column` , it will convert all the variables in this column into one variable. I also run this script in version 2.5.2, this bug does not appear. So I switched to the older version.
## Steps to reproduce the bug
```python
from datasets import load_dataset
dataset = load_dataset("amazon_reviews_multi", "en")
data = dataset['train']
data = data.remove_columns(
[
"review_id",
"product_id",
"reviewer_id",
"review_title",
"language",
"product_category",
]
)
data = data.rename_column("review_body", "text")
data1 = data.class_encode_column("stars")
print(set(data1.data.columns[0]))
# output: {<pyarrow.Int64Scalar: 4>, <pyarrow.Int64Scalar: 2>, <pyarrow.Int64Scalar: 3>, <pyarrow.Int64Scalar: 0>, <pyarrow.Int64Scalar: 1>}
data = data.rename_column("stars", "label")
print(set(data.data.columns[0]))
# output: {<pyarrow.Int32Scalar: 5>, <pyarrow.Int32Scalar: 4>, <pyarrow.Int32Scalar: 1>, <pyarrow.Int32Scalar: 3>, <pyarrow.Int32Scalar: 2>}
data2 = data.class_encode_column("label")
print(set(data2.data.columns[0]))
# output: {<pyarrow.Int64Scalar: 0>}
```
## Expected results
the last print should be:
{<pyarrow.Int64Scalar: 4>, <pyarrow.Int64Scalar: 2>, <pyarrow.Int64Scalar: 3>, <pyarrow.Int64Scalar: 0>, <pyarrow.Int64Scalar: 1>}
## Actual results
but it output:
{<pyarrow.Int64Scalar: 0>}
## Environment info
- `datasets` version: 2.6.1
- Platform: macOS-12.5.1-arm64-arm-64bit
- Python version: 3.10.6
- PyArrow version: 9.0.0
- Pandas version: 1.5.0
| 5,129 |
https://github.com/huggingface/datasets/issues/5123 | datasets freezes with streaming mode in multiple-gpu | [
"@lhoestq I tested the script without accelerator, and I confirm this is due to datasets part as this gets similar results without accelerator.",
"Hi ! You said it works on 1 GPU but doesn't wortk without accelerator - what's the difference between running on 1 GPU and running without accelerator in your case ?"... | ## Describe the bug
Hi. I am using this dataloader, which is for processing large datasets in streaming mode mentioned in one of examples of huggingface. I am using it to read c4: https://github.com/huggingface/transformers/blob/b48ac1a094e572d6076b46a9e4ed3e0ebe978afc/examples/research_projects/codeparrot/scripts/codeparrot_training.py#L22
During using multi-gpu in accelerator in one node, the code freezes, but works for 1 GPU:
```
10/16/2022 14:18:46 - INFO - datasets.info - Loading Dataset Infos from /home/jack/.cache/huggingface/modules/datasets_modules/datasets/c4/df532b158939272d032cc63ef19cd5b83e9b4d00c922b833e4cb18b2e9869b01
Steps: 0%| | 0/400000 [00:00<?, ?it/s]10/16/2022 14:18:47 - INFO - torch.utils.data.dataloader - Shared seed (135290893754684706) sent to store on rank 0
```
# Code to reproduce
please run this code with `accelerate launch code.py`
```
from accelerate import Accelerator
from accelerate.logging import get_logger
from datasets import load_dataset
from torch.utils.data.dataloader import DataLoader
import torch
from datasets import load_dataset
from transformers import AutoTokenizer
import torch
from accelerate.logging import get_logger
from torch.utils.data import IterableDataset
from torch.utils.data.datapipes.iter.combinatorics import ShufflerIterDataPipe
logger = get_logger(__name__)
class ConstantLengthDataset(IterableDataset):
"""
Iterable dataset that returns constant length chunks of tokens from stream of text files.
Args:
tokenizer (Tokenizer): The processor used for proccessing the data.
dataset (dataset.Dataset): Dataset with text files.
infinite (bool): If True the iterator is reset after dataset reaches end else stops.
max_seq_length (int): Length of token sequences to return.
num_of_sequences (int): Number of token sequences to keep in buffer.
chars_per_token (int): Number of characters per token used to estimate number of tokens in text buffer.
"""
def __init__(
self,
tokenizer,
dataset,
infinite=False,
max_seq_length=1024,
num_of_sequences=1024,
chars_per_token=3.6,
):
self.tokenizer = tokenizer
# self.concat_token_id = tokenizer.bos_token_id
self.dataset = dataset
self.max_seq_length = max_seq_length
self.epoch = 0
self.infinite = infinite
self.current_size = 0
self.max_buffer_size = max_seq_length * chars_per_token * num_of_sequences
self.content_field = "text"
def __iter__(self):
iterator = iter(self.dataset)
more_examples = True
while more_examples:
buffer, buffer_len = [], 0
while True:
if buffer_len >= self.max_buffer_size:
break
try:
buffer.append(next(iterator)[self.content_field])
buffer_len += len(buffer[-1])
except StopIteration:
if self.infinite:
iterator = iter(self.dataset)
self.epoch += 1
logger.info(f"Dataset epoch: {self.epoch}")
else:
more_examples = False
break
tokenized_inputs = self.tokenizer(buffer, truncation=False)["input_ids"]
all_token_ids = []
for tokenized_input in tokenized_inputs:
all_token_ids.extend(tokenized_input)
for i in range(0, len(all_token_ids), self.max_seq_length):
input_ids = all_token_ids[i : i + self.max_seq_length]
if len(input_ids) == self.max_seq_length:
self.current_size += 1
yield torch.tensor(input_ids)
def shuffle(self, buffer_size=1000):
return ShufflerIterDataPipe(self, buffer_size=buffer_size)
def create_dataloaders(tokenizer, accelerator):
ds_kwargs = {"streaming": True}
# In distributed training, the load_dataset function gaurantees that only one process
# can concurrently download the dataset.
datasets = load_dataset(
"c4",
"en",
cache_dir="cache_dir",
**ds_kwargs,
)
train_data, valid_data = datasets["train"], datasets["validation"]
with accelerator.main_process_first():
train_data = train_data.shuffle(buffer_size=10000, seed=None)
train_dataset = ConstantLengthDataset(
tokenizer,
train_data,
infinite=True,
max_seq_length=256,
)
valid_dataset = ConstantLengthDataset(
tokenizer,
valid_data,
infinite=False,
max_seq_length=256,
)
train_dataset = train_dataset.shuffle(buffer_size=10000)
train_dataloader = DataLoader(train_dataset, batch_size=160, shuffle=True)
eval_dataloader = DataLoader(valid_dataset, batch_size=160)
return train_dataloader, eval_dataloader
def main():
# Accelerator.
logging_dir = "data_save_dir/log"
accelerator = Accelerator(
gradient_accumulation_steps=1,
mixed_precision="bf16",
log_with="tensorboard",
logging_dir=logging_dir,
)
# We need to initialize the trackers we use, and also store our configuration.
# The trackers initializes automatically on the main process.
if accelerator.is_main_process:
accelerator.init_trackers("test")
tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
# Load datasets and create dataloaders.
train_dataloader, _ = create_dataloaders(tokenizer, accelerator)
train_dataloader = accelerator.prepare(train_dataloader)
for step, batch in enumerate(train_dataloader, start=1):
print(step)
accelerator.end_training()
if __name__ == "__main__":
main()
```
## Results expected
Being able to run the code for streamining datasets with multi-gpu
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.5.2
- Platform: linux
- Python version: 3.9.12
- PyArrow version: 9.0.0
@lhoestq I do not have any idea why this freezing happens, and I removed the streaming mode and this was working fine, so I know this is caused by streaming mode of the dataloader part not working well with multi-gpu setting. Since datasets are large, I hope to keep the streamining mode. I very much appreciate your help.
| 5,123 |
https://github.com/huggingface/datasets/issues/5118 | Installing `datasets` on M1 computers | [
"Thanks for reporting, @david1542."
] | ## Describe the bug
I wanted to install `datasets` dependencies on my M1 (in order to start contributing to the project). However, I got an error regarding `tensorflow`.
On M1, `tensorflow-macos` needs to be installed instead. Can we add a conditional requirement, so that `tensorflow-macos` would be installed on M1?
## Steps to reproduce the bug
Fresh clone this project (on m1), create a virtualenv and run this:
```python
pip install -e ".[dev]"
```
## Expected results
Installation should be smooth, and all the dependencies should be installed on M1.
## Actual results
You should receive an error, saying pip couldn't find a version that matches this pattern:
```
tensorflow>=2.3,!=2.6.0,!=2.6.1
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.6.2.dev0
- Platform: macOS-12.6-arm64-arm-64bit
- Python version: 3.9.6
- PyArrow version: 7.0.0
- Pandas version: 1.5.0
| 5,118 |
https://github.com/huggingface/datasets/issues/5117 | Progress bars have color red and never completed to 100% | [
"Hi @echatzikyriakidis, thanks for submitting the issue.\r\nWhich shell are you using exactly? I tried to run the command you sent, but I don't see colors at all 🧐\r\n\r\nI tried from bash and zsh as well.",
"Hi @david1542 ,\r\n\r\nI use Google Colab.\r\n",
"Got it. I [created a PR](https://github.com/huggingf... | ## Describe the bug
Progress bars after transformative operations turn in red and never be completed to 100%
## Steps to reproduce the bug
```python
from datasets import load_dataset
load_dataset('rotten_tomatoes', split='test').filter(lambda o: True)
```
## Expected results
Progress bar should be 100% and green
## Actual results
Progress bar turn in red and never completed to 100%
## Environment info
- `datasets` version: 2.6.1
- Platform: Linux-5.10.133+-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.14
- PyArrow version: 6.0.1
- Pandas version: 1.3.5 | 5,117 |
https://github.com/huggingface/datasets/issues/5114 | load_from_disk with remote filesystem fails due to a wrong temporary local folder path | [
"Hi Hubert! Could you please probably create a publicly available `gs://` dataset link? I think this would be easier for others to directly start to debug.",
"What seems to work is to change the line to:\r\n```\r\nfs.download(src_dataset_path, dataset_path.parent.as_posix(), recursive=True)\r\n```"
] | ## Describe the bug
The function load_from_disk fails when using a remote filesystem because of a wrong temporary path generation in the load_from_disk method of arrow_dataset.py:
```python
if is_remote_filesystem(fs):
src_dataset_path = extract_path_from_uri(dataset_path)
dataset_path = Dataset._build_local_temp_path(src_dataset_path)
fs.download(src_dataset_path, dataset_path.as_posix(), recursive=True)
```
If _dataset_path_ is `gs://speech/mydataset/train`, then _src_dataset_path_ will be `speech/mydataset/train` and _dataset_path_ will be something like `/var/folders/9s/gf0b/T/tmp6t/speech/mydataset/train`
Then, after downloading the **folder** _src_dataset_path_, you will get a path like `/var/folders/9s/gf0b/T/tmp6t/speech/mydataset/train/train/state.json` (notice we have train twice)
Instead of downloading the remote folder we should be downloading all the files in the folder for the path to be right:
```python
fs.download(os.path.join(src_dataset_path,*), dataset_path.as_posix(), recursive=True)
```
## Steps to reproduce the bug
```python
fs = gcsfs.GCSFileSystem(**storage_options)
dataset = load_from_disk("common_voice_processed") # loading local dataset previously saved locally, works fine
dataset.save_to_disk(output_dir, fs=fs) #works fine
dataset = load_from_disk(output_dir, fs=fs) # crashes
```
## Expected results
The dataset is loaded
## Actual results
FileNotFoundError: [Errno 2] No such file or directory: '/var/folders/9s/gf0b9jz15d517yrf7m3nvlxr0000gn/T/tmp6t5e221_/speech/datasets/tests/common_voice_processed/train/state.json'
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: datasets-2.6.1.dev0
- Platform: mac os monterey 12.5.1
- Python version: 3.8.13
- PyArrow version:pyarrow==9.0.0
| 5,114 |
https://github.com/huggingface/datasets/issues/5112 | Bug with filtered indices | [
"The issue is here:\r\nhttps://github.com/huggingface/datasets/blob/3ad9644b9a2e4558dd1d0f1e43c67658674e6228/src/datasets/arrow_dataset.py#L2964",
"@PartiallyTyped, @Muennighoff: the issue is fixed.\r\n\r\nWe are planning to make a patch release today.",
"Thanks a lot for the swift response! For a brief moment ... | ## Describe the bug
As reported by @PartiallyTyped (and by @Muennighoff):
- https://github.com/huggingface/datasets/issues/5111#issuecomment-1278652524
There is an issue with the indices of a filtered dataset.
## Steps to reproduce the bug
```python
ds = Dataset.from_dict({"num": [0, 1, 2, 3]})
ds = ds.filter(lambda num: num % 2 == 0, input_columns="num", batch_size=2)
assert all(item["num"] % 2 == 0 for item in ds)
```
## Expected results
The indices of the filtered dataset should correspond to the examples with "language" equals to "english".
## Actual results
Indices to items with other languages are included in the filtered dataset indices
## Preliminar investigation
It seems a bug introduced by:
- #5030
| 5,112 |
https://github.com/huggingface/datasets/issues/5111 | map and filter not working properly in multiprocessing with the new release 2.6.0 | [
"Same bug exists with `num_proc=1` on colab. `3.7.14 (default, Sep 8 2022, 00:06:44) [GCC 7.5.0]` ",
"Thanks for reporting, @loubnabnl and for the additional information, @PartiallyTyped.\r\n\r\nHowever, I'm not able to reproduce this issue, neither locally nor on Colab:\r\n```\r\nDataset({\r\n features: ['re... | ## Describe the bug
When mapping is used on a dataset with more than one process, there is a weird behavior when trying to use `filter` , it's like only the samples from one worker are retrieved, one needs to specify the same `num_proc` in filter for it to work properly. This doesn't happen with `datasets` version 2.5.2
In the code below the data is filtered differently when we increase `num_proc` used in `map` although the datsets before and after mapping have identical elements.
## Steps to reproduce the bug
```python
import datasets
from datasets import load_dataset
def preprocess(example):
return example
ds = load_dataset("codeparrot/codeparrot-clean-valid", split="train").select([i for i in range(10)])
ds1 = ds.map(preprocess, num_proc=2)
ds2 = ds.map(preprocess)
# the datasets elements are the same
for i in range(len(ds1)):
assert ds1[i]==ds2[i]
print(f'Target column before filtering {ds1["autogenerated"]}')
print(f'Target column before filtering {ds2["autogenerated"]}')
print(f"datasets version {datasets.__version__}")
ds_filtered_1 = ds1.filter(lambda x: not x["autogenerated"])
ds_filtered_2 = ds2.filter(lambda x: not x["autogenerated"])
# all elements in Target column are false so they should all be kept, but for ds2 only the first 5=num_samples/num_proc are kept
print(ds_filtered_1)
print(ds_filtered_2)
```
```
Target column before filtering [False, False, False, False, False, False, False, False, False, False]
Target column before filtering [False, False, False, False, False, False, False, False, False, False]
Dataset({
features: ['repo_name', 'path', 'copies', 'size', 'content', 'license', 'hash', 'line_mean', 'line_max', 'alpha_frac', 'autogenerated'],
num_rows: 5
})
Dataset({
features: ['repo_name', 'path', 'copies', 'size', 'content', 'license', 'hash', 'line_mean', 'line_max', 'alpha_frac', 'autogenerated'],
num_rows: 10
})
```
## Expected results
Increasing `num_proc` in mapping shouldn't alter filtering. With the previous version 2.5.2 this doesn't happen
## Actual results
Filtering doesn't work properly when we increase `num_proc` in mapping but not when calling `filter`
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.6.0
- Platform: Linux-4.19.0-22-cloud-amd64-x86_64-with-glibc2.28
- Python version: 3.9.13
- PyArrow version: 8.0.0
- Pandas version: 1.4.2 | 5,111 |
https://github.com/huggingface/datasets/issues/5109 | Map caching not working for some class methods | [
"The hash used for caching is computed by pickling recursively the function passed to `map`. Maybe some objects don't have the same hash across sessions. In particular you can check the hash of your model using\r\n```python\r\nfrom datasets.fingerprint import Hasher\r\nobj = AutoModel.from_config(config=config, ad... | ## Describe the bug
The cache loading is not working as expected for some class methods with a model stored in an attribute.
The new fingerprint for `_map_single` is not the same at each run. The hasher generate a different hash for the class method.
This comes from `dumps` function in `datasets.utils.py_utils` which generates a different dump at each run.
## Steps to reproduce the bug
```python
from datasets import load_dataset
from transformers import AutoConfig, AutoModel, AutoTokenizer
dataset = load_dataset("ethos", "binary")
BASE_MODELNAME = "sentence-transformers/all-MiniLM-L6-v2"
class Object:
def __init__(self):
config = AutoConfig.from_pretrained(BASE_MODELNAME)
self.bert = AutoModel.from_config(config=config, add_pooling_layer=False)
self.tok = AutoTokenizer.from_pretrained(BASE_MODELNAME)
def tokenize(self, examples):
tokenized_texts = self.tok(
examples["text"],
padding="max_length",
truncation=True,
max_length=256,
)
return tokenized_texts
instance = Object()
result = dict()
for phase in ["train"]:
result[phase] = dataset[phase].map(instance.tokenize, batched=True, load_from_cache_file=True, num_proc=2)
```
## Expected results
Load cache instead of recompute result.
## Actual results
Result recomputed from scratch at each run.
The cache works fine when deleting `bert` attribute.
## Environment info
- `datasets` version: 2.5.3.dev0
- Platform: macOS-10.16-x86_64-i386-64bit
- Python version: 3.9.13
- PyArrow version: 7.0.0
- Pandas version: 1.5.0
| 5,109 |
https://github.com/huggingface/datasets/issues/5105 | Specifying an exisiting folder in download_and_prepare deletes everything in it | [
"cc @lhoestq ",
"Thanks for reporting, @cakiki.\r\n\r\nI would say the deletion of the dir is an expected behavior though...",
"`dask.to_parquet` has an \"overwrite\" parameter and default is `False`, we could also have something similar",
"Thank you both for your feedback!\r\n\r\n@albertvillanova I think I m... | ## Describe the bug
The builder correctly creates the `output_dir` folder if it doesn't exist, but if the folder exists everything within it is deleted. Specifying `"."` as the `output_dir` deletes everything in your current dir but also leads to **another bug** whose traceback is the following:
```
Traceback (most recent call last)
Input In [11], in <cell line: 1>()
----> 1 rotten_tomatoes_builder.download_and_prepare(output_dir=".", max_shard_size="200MB", file_format="parquet")
File ~/BIGSCIENCE/env/lib/python3.9/site-packages/datasets/builder.py:818, in download_and_prepare(self, output_dir, download_config, download_mode, ignore_verifications, try_from_hf_gcs, dl_manager, base_path, use_auth_token, file_format, max_shard_size, storage_options, **download_and_prepare_kwargs)
File /usr/lib/python3.9/contextlib.py:124, in _GeneratorContextManager.__exit__(self, type, value, traceback)
122 if type is None:
123 try:
--> 124 next(self.gen)
125 except StopIteration:
126 return False
File ~/BIGSCIENCE/env/lib/python3.9/site-packages/datasets/builder.py:760, in incomplete_dir(dirname)
File /usr/lib/python3.9/shutil.py:722, in rmtree(path, ignore_errors, onerror)
720 os.rmdir(path)
721 except OSError:
--> 722 onerror(os.rmdir, path, sys.exc_info())
723 else:
724 try:
725 # symlinks to directories are forbidden, see bug #1669
File /usr/lib/python3.9/shutil.py:720, in rmtree(path, ignore_errors, onerror)
718 _rmtree_safe_fd(fd, path, onerror)
719 try:
--> 720 os.rmdir(path)
721 except OSError:
722 onerror(os.rmdir, path, sys.exc_info())
OSError: [Errno 22] Invalid argument: '/home/christopher/BIGSCIENCE/.'
```
## Steps to reproduce the bug
```python
rotten_tomatoes_builder = load_dataset_builder("rotten_tomatoes")
rotten_tomatoes_builder.download_and_prepare(output_dir="./test_folder", max_shard_size="200MB", file_format="parquet")
```
If `test_folder` contains any files they will all be deleted
## Expected results
Either a warning that all files will be deleted, but preferably that they not be deleted at all.
## Actual results
N/A
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.3.2
- Platform: Linux-5.15.0-48-generic-x86_64-with-glibc2.29
- Python version: 3.8.10
- PyArrow version: 8.0.0
- Pandas version: 1.4.3
| 5,105 |
https://github.com/huggingface/datasets/issues/5102 | Error in create a dataset from a Python generator | [
"Hi, thanks for reporting! The last line should be `dataset = Dataset.from_generator(my_gen)`.",
"Can I work on this one?"
] | ## Describe the bug
In HOW-TO-GUIDES > Load > [Python generator](https://huggingface.co/docs/datasets/v2.5.2/en/loading#python-generator), the code example defines the `my_gen` function, but when creating the dataset, an undefined `my_dict` is passed in.
```Python
>>> from datasets import Dataset
>>> def my_gen():
... for i in range(1, 4):
... yield {"a": i}
>>> dataset = Dataset.from_generator(my_dict)
``` | 5,102 |
https://github.com/huggingface/datasets/issues/5100 | datasets[s3] sagemaker can't run a model - datasets issue with Value and ClassLabel and cast() method | [] | null | 5,100 |
https://github.com/huggingface/datasets/issues/5099 | datasets doesn't support # in data paths | [
"`datasets` doesn't seem to urlencode the directory names here\r\n\r\nhttps://github.com/huggingface/datasets/blob/7feeb5648a63b6135a8259dedc3b1e19185ee4c7/src/datasets/utils/file_utils.py#L109-L111\r\n\r\nfor example we should have\r\n```python\r\nfrom datasets.utils.file_utils import hf_hub_url\r\n\r\nurl = hf_hu... | ## Describe the bug
dataset files with `#` symbol their paths aren't read correctly.
## Steps to reproduce the bug
The data in folder `c#`of this [dataset](https://huggingface.co/datasets/loubnabnl/bigcode_csharp) can't be loaded. While the folder `c_sharp` with the same data is loaded properly
```python
ds = load_dataset('loubnabnl/bigcode_csharp', split="train", data_files=["data/c#/*"])
```
```
FileNotFoundError: Couldn't find file at https://huggingface.co/datasets/loubnabnl/bigcode_csharp/resolve/27a3166cff4bb18e11919cafa6f169c0f57483de/data/c#/data_0003.jsonl
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.5.2
- Platform: macOS-12.2.1-arm64-arm-64bit
- Python version: 3.9.13
- PyArrow version: 9.0.0
- Pandas version: 1.4.3
cc @lhoestq | 5,099 |
https://github.com/huggingface/datasets/issues/5098 | Classes label error when loading symbolic links using imagefolder | [
"It can be solved temporarily by remove `resolve` in \r\nhttps://github.com/huggingface/datasets/blob/bef23be3d9543b1ca2da87ab2f05070201044ddc/src/datasets/data_files.py#L278",
"Hi, thanks for reporting and suggesting a fix! We still need to account for `.`/`..` in the file path, so a more robust fix would be `P... | **Is your feature request related to a problem? Please describe.**
Like this: #4015
When there are **symbolic links** to pictures in the data folder, the parent folder name of the **real file** will be used as the class name instead of the parent folder of the symbolic link itself. Can you give an option to decide whether to enable symbolic link tracking?
This is inconsistent with the `torchvision.datasets.ImageFolder` behavior.
For example:
It use `others` in green circle as class label but not `abnormal`, I wish `load_dataset` not use the real file parent as label.
**Describe the solution you'd like**
A clear and concise description of what you want to happen.
**Describe alternatives you've considered**
A clear and concise description of any alternative solutions or features you've considered.
**Additional context**
Add any other context about the feature request here.
| 5,098 |
https://github.com/huggingface/datasets/issues/5097 | Fatal error with pyarrow/libarrow.so | [
"Thanks for reporting, @catalys1.\r\n\r\nThis seems a duplicate of:\r\n- #3310 \r\n\r\nThe source of the problem is in PyArrow:\r\n- [ARROW-15141: [C++] Fatal error condition occurred in aws_thread_launch](https://issues.apache.org/jira/browse/ARROW-15141)\r\n- [ARROW-17501: [C++] Fatal error condition occurred in ... | ## Describe the bug
When using datasets, at the very end of my jobs the program crashes (see trace below).
It doesn't seem to affect anything, as it appears to happen as the program is closing down. Just importing `datasets` is enough to cause the error.
## Steps to reproduce the bug
This is sufficient to reproduce the problem:
```bash
python -c "import datasets"
```
## Expected results
Program should run to completion without an error.
## Actual results
```bash
Fatal error condition occurred in /opt/vcpkg/buildtrees/aws-c-io/src/9e6648842a-364b708815.clean/source/event_loop.c:72: aws_thread_launch(&cleanup_thread, s_event_loop_destroy_async_thread_fn, el_group, &thread_options) == AWS_OP_SUCCESS
Exiting Application
################################################################################
Stack trace:
################################################################################
/u/user/miniconda3/envs/env/lib/python3.10/site-packages/pyarrow/libarrow.so.900(+0x200af06) [0x150dff547f06]
/u/user/miniconda3/envs/env/lib/python3.10/site-packages/pyarrow/libarrow.so.900(+0x20028e5) [0x150dff53f8e5]
/u/user/miniconda3/envs/env/lib/python3.10/site-packages/pyarrow/libarrow.so.900(+0x1f27e09) [0x150dff464e09]
/u/user/miniconda3/envs/env/lib/python3.10/site-packages/pyarrow/libarrow.so.900(+0x200ba3d) [0x150dff548a3d]
/u/user/miniconda3/envs/env/lib/python3.10/site-packages/pyarrow/libarrow.so.900(+0x1f25948) [0x150dff462948]
/u/user/miniconda3/envs/env/lib/python3.10/site-packages/pyarrow/libarrow.so.900(+0x200ba3d) [0x150dff548a3d]
/u/user/miniconda3/envs/env/lib/python3.10/site-packages/pyarrow/libarrow.so.900(+0x1ee0b46) [0x150dff41db46]
/u/user/miniconda3/envs/env/lib/python3.10/site-packages/pyarrow/libarrow.so.900(+0x194546a) [0x150dfee8246a]
/lib64/libc.so.6(+0x39b0c) [0x150e15eadb0c]
/lib64/libc.so.6(on_exit+0) [0x150e15eadc40]
/u/user/miniconda3/envs/env/bin/python(+0x28db18) [0x560ae370eb18]
/u/user/miniconda3/envs/env/bin/python(+0x28db4b) [0x560ae370eb4b]
/u/user/miniconda3/envs/env/bin/python(+0x28db90) [0x560ae370eb90]
/u/user/miniconda3/envs/env/bin/python(_PyRun_SimpleFileObject+0x1e6) [0x560ae37123e6]
/u/user/miniconda3/envs/env/bin/python(_PyRun_AnyFileObject+0x44) [0x560ae37124c4]
/u/user/miniconda3/envs/env/bin/python(Py_RunMain+0x35d) [0x560ae37135bd]
/u/user/miniconda3/envs/env/bin/python(Py_BytesMain+0x39) [0x560ae37137d9]
/lib64/libc.so.6(__libc_start_main+0xf3) [0x150e15e97493]
/u/user/miniconda3/envs/env/bin/python(+0x2125d4) [0x560ae36935d4]
Aborted (core dumped)
```
## Environment info
- `datasets` version: 2.5.1
- Platform: Linux-4.18.0-348.23.1.el8_5.x86_64-x86_64-with-glibc2.28
- Python version: 3.10.4
- PyArrow version: 9.0.0
- Pandas version: 1.4.3
| 5,097 |
https://github.com/huggingface/datasets/issues/5096 | Transfer some canonical datasets under an organization namespace | [
"The transfer of the dummy dataset to the dummy org works as expected:\r\n```python\r\nIn [1]: from datasets import load_dataset; ds = load_dataset(\"dummy_canonical_dataset\", download_mode=\"force_redownload\"); ds\r\nDownloading builder script: 100%|███████████████████████████████████████████████████████████████... | As discussed during our @huggingface/datasets meeting, we are planning to move some "canonical" dataset scripts under their corresponding organization namespace (if this does not exist).
On the contrary, if the dataset already exists under the organization namespace, we are deprecating the canonical one (and eventually delete it).
First, we should test it using a dummy dataset/organization.
TODO:
- [x] Test with a dummy dataset
- [x] Create dummy canonical dataset: https://huggingface.co/datasets/dummy_canonical_dataset
- [x] Create dummy organization: https://huggingface.co/dummy-canonical-org
- [x] Transfer dummy canonical dataset to dummy organization
- [ ] Transfer datasets
- [x] babi_qa => facebook
- [x] blbooks => TheBritishLibrary/blbooks
- [x] blbooksgenre => TheBritishLibrary/blbooksgenre
- [x] common_gen => allenai
- [x] commonsense_qa => tau
- [x] competition_math => hendrycks/competition_math
- [x] cord19 => allenai
- [x] emotion => dair-ai
- [ ] gem => GEM
- [x] hellaswag => Rowan
- [x] hendrycks_test => cais/mmlu
- [x] indonlu => indonlp
- [ ] multilingual_librispeech => facebook
- It already exists "facebook/multilingual_librispeech"
- [ ] oscar => oscar-corpus
- [x] peer_read => allenai
- [x] qasper => allenai
- [x] reddit => webis/tldr-17
- [x] russian_super_glue => russiannlp
- [x] rvl_cdip => aharley
- [x] s2orc => allenai
- [x] scicite => allenai
- [x] scifact => allenai
- [x] scitldr => allenai
- [x] swiss_judgment_prediction => rcds
- [x] the_pile => EleutherAI
- [ ] wmt14, wmt15, wmt16, wmt17, wmt18, wmt19,... => wmt
- [ ] Deprecate (and eventually remove) datasets that cannot be transferred because they already exist
- [x] banking77 => PolyAI
- [x] common_voice => mozilla-foundation
- [x] german_legal_entity_recognition => elenanereiss
- ...
EDIT: the list above is continuously being updated | 5,096 |
https://github.com/huggingface/datasets/issues/5094 | Multiprocessing with `Dataset.map` and `PyTorch` results in deadlock | [
"Hi ! Could it be an Out of Memory issue that could have killed one of the processes ? can you check your memory ?",
"Hi! I don't think it is a memory issue. I'm monitoring the main and spawn python processes and threads with `htop` and the memory does not peak. Besides, the example I've posted above should not b... | ## Describe the bug
There seems to be an issue with using multiprocessing with `datasets.Dataset.map` (i.e. setting `num_proc` to a value greater than one) combined with a function that uses `torch` under the hood. The subprocesses that `datasets.Dataset.map` spawns [a this step](https://github.com/huggingface/datasets/blob/1b935dab9d2f171a8c6294269421fe967eb55e34/src/datasets/arrow_dataset.py#L2663) go into wait mode forever.
## Steps to reproduce the bug
The below code goes into deadlock when `NUMBER_OF_PROCESSES` is greater than one.
```python
NUMBER_OF_PROCESSES = 2
from transformers import AutoTokenizer, AutoModel
from datasets import load_dataset
dataset = load_dataset("glue", "mrpc", split="train")
tokenizer = AutoTokenizer.from_pretrained("sentence-transformers/all-MiniLM-L6-v2")
model = AutoModel.from_pretrained("sentence-transformers/all-MiniLM-L6-v2")
model.to("cpu")
def cls_pooling(model_output):
return model_output.last_hidden_state[:, 0]
def generate_embeddings_batched(examples):
sentences_batch = list(examples['sentence1'])
encoded_input = tokenizer(
sentences_batch, padding=True, truncation=True, return_tensors="pt"
)
encoded_input = {k: v.to("cpu") for k, v in encoded_input.items()}
model_output = model(**encoded_input)
embeddings = cls_pooling(model_output)
examples['embeddings'] = embeddings.detach().cpu().numpy() # 64, 384
return examples
embeddings_dataset = dataset.map(
generate_embeddings_batched,
batched=True,
batch_size=10,
num_proc=NUMBER_OF_PROCESSES
)
```
While debugging it I've seen that it gets "stuck" when calling `torch.nn.Embedding.forward` but some testing shows that the same happens with other functions from `torch.nn`.
## Environment info
- Platform: Linux-5.14.0-1052-oem-x86_64-with-glibc2.31
- Python version: 3.9.14
- PyArrow version: 9.0.0
- Pandas version: 1.5.0
Not sure if this is a HF problem, a PyTorch problem or something I'm doing wrong..
Thanks!
| 5,094 |
https://github.com/huggingface/datasets/issues/5093 | Mismatch between tutoriel and doc | [
"Hi, thanks for reporting! This line should be replaced with \r\n```python\r\ndataset = dataset.map(lambda examples: tokenizer(examples[\"text\"], return_tensors=\"np\"), batched=True)\r\n```\r\nfor it to work (the `return_tensors` part inside the `tokenizer` call).",
"Can I work on this?",
"Fixed in https://gi... | ## Describe the bug
In the "Process text data" tutorial, [`map` has `return_tensors` as kwarg](https://huggingface.co/docs/datasets/main/en/nlp_process#map). It does not seem to appear in the [function documentation](https://huggingface.co/docs/datasets/main/en/package_reference/main_classes#datasets.Dataset.map), nor to work.
## Steps to reproduce the bug
MWE:
```python
from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("bert-base-cased")
from datasets import load_dataset
dataset = load_dataset("lhoestq/demo1", split="train")
dataset = dataset.map(lambda examples: tokenizer(examples["review"]), batched=True, return_tensors="pt")
```
## Expected results
return_tensors to be a valid kwarg :smiley:
## Actual results
```python
>> TypeError: map() got an unexpected keyword argument 'return_tensors'
```
## Environment info
- `datasets` version: 2.3.2
- Platform: Linux-5.14.0-1052-oem-x86_64-with-glibc2.29
- Python version: 3.8.10
- PyArrow version: 8.0.0
- Pandas version: 1.4.3
| 5,093 |
https://github.com/huggingface/datasets/issues/5090 | Review sync issues from GitHub to Hub | [
"Nice!!"
] | ## Describe the bug
We have discovered that sometimes there were sync issues between GitHub and Hub datasets, after a merge commit to main branch.
For example:
- this merge commit: https://github.com/huggingface/datasets/commit/d74a9e8e4bfff1fed03a4cab99180a841d7caf4b
- was not properly synced with the Hub: https://github.com/huggingface/datasets/actions/runs/3002495269/jobs/4819769684
```
[main 9e641de] Add Papers with Code ID to scifact dataset (#4941)
Author: Albert Villanova del Moral <albertvillanova@users.noreply.huggingface.co>
1 file changed, 42 insertions(+), 14 deletions(-)
push failed !
GitCommandError(['git', 'push'], 1, b'remote: ---------------------------------------------------------- \nremote: Sorry, your push was rejected during YAML metadata verification: \nremote: - Error: "license" does not match any of the allowed types \nremote: ---------------------------------------------------------- \nremote: Please find the documentation at: \nremote: https://huggingface.co/docs/hub/models-cards#model-card-metadata \nremote: ---------------------------------------------------------- \nTo [https://huggingface.co/datasets/scifact.git\n](https://huggingface.co/datasets/scifact.git/n) ! [remote rejected] main -> main (pre-receive hook declined)\nerror: failed to push some refs to \'[https://huggingface.co/datasets/scifact.git\](https://huggingface.co/datasets/scifact.git/)'', b'')
```
We are reviewing sync issues in previous commits to recover them and repushing to the Hub.
TODO: Review
- [x] #4941
- scifact
- [x] #4931
- scifact
- [x] #4753
- wikipedia
- [x] #4554
- wmt17, wmt19, wmt_t2t
- Fixed with "Release 2.4.0" commit: https://github.com/huggingface/datasets/commit/401d4c4f9b9594cb6527c599c0e7a72ce1a0ea49
- https://huggingface.co/datasets/wmt17/commit/5c0afa83fbbd3508ff7627c07f1b27756d1379ea
- https://huggingface.co/datasets/wmt19/commit/b8ad5bf1960208a376a0ab20bc8eac9638f7b400
- https://huggingface.co/datasets/wmt_t2t/commit/b6d67191804dd0933476fede36754a436b48d1fc
- [x] #4607
- [x] #4416
- lccc
- Fixed with "Release 2.3.0" commit: https://huggingface.co/datasets/lccc/commit/8b1f8cf425b5653a0a4357a53205aac82ce038d1
- [x] #4367
| 5,090 |
https://github.com/huggingface/datasets/issues/5089 | Resume failed process | [] | **Is your feature request related to a problem? Please describe.**
When a process (`map`, `filter`, etc.) crashes part-way through, you lose all progress.
**Describe the solution you'd like**
It would be good if the cache reflected the partial progress, so that after we restart the script, the process can restart where it left off.
**Describe alternatives you've considered**
Doing processing outside of `datasets`, by writing the dataset to json files and building a restart mechanism myself.
**Additional context**
N/A
| 5,089 |
https://github.com/huggingface/datasets/issues/5088 | load_datasets("json", ...) don't read local .json.gz properly | [
"Hi @junwang-wish, thanks for reporting.\r\n\r\nUnfortunately, I'm not able to reproduce the bug. Which version of `datasets` are you using? Does the problem persist if you update `datasets`?\r\n```shell\r\npip install -U datasets\r\n``` ",
"Thanks @albertvillanova I updated `datasets` from `2.5.1` to `2.5.2` and... | ## Describe the bug
I have a local file `*.json.gz` and it can be read by `pandas.read_json(lines=True)`, but cannot be read by `load_datasets("json")` (resulting in 0 lines)
## Steps to reproduce the bug
```python
fpath = '/data/junwang/.cache/general/57b6f2314cbe0bc45dda5b78f0871df2/test.json.gz'
ds_panda = DatasetDict(
test=Dataset.from_pandas(
pd.read_json(fpath, lines=True)
)
)
ds_direct = load_dataset(
'json', data_files={
'test': fpath
}, features=Features(
text_input=Value(dtype="string", id=None),
text_output=Value(dtype="string", id=None)
)
)
len(ds_panda['test']), len(ds_direct['test'])
```
## Expected results
Lines of `ds_panda['test']` and `ds_direct['test']` should match.
## Actual results
```
Using custom data configuration default-c0ef2598760968aa
Downloading and preparing dataset json/default to /data/junwang/.cache/huggingface/datasets/json/default-c0ef2598760968aa/0.0.0/e6070c77f18f01a5ad4551a8b7edfba20b8438b7cad4d94e6ad9378022ce4aab...
Dataset json downloaded and prepared to /data/junwang/.cache/huggingface/datasets/json/default-c0ef2598760968aa/0.0.0/e6070c77f18f01a5ad4551a8b7edfba20b8438b7cad4d94e6ad9378022ce4aab. Subsequent calls will reuse this data.
(62087, 0)
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version:
- Platform: Ubuntu 18.04.4 LTS
- Python version: 3.8.13
- PyArrow version: 9.0.0
| 5,088 |
https://github.com/huggingface/datasets/issues/5086 | HTTPError: 404 Client Error: Not Found for url | [
"FYI @lewtun ",
"Hi @km5ar, thanks for reporting.\r\n\r\nThis should be fixed in the notebook:\r\n- the filename `datasets-issues-with-hf-doc-builder.jsonl` no longer exists on the repo; instead, current filename is `datasets-issues-with-comments.jsonl`\r\n- see: https://huggingface.co/datasets/lewtun/github-issu... | ## Describe the bug
I was following chap 5 from huggingface course: https://huggingface.co/course/chapter5/6?fw=tf
However, I'm not able to download the datasets, with a 404 erros
<img width="1160" alt="iShot2022-10-06_15 54 50" src="https://user-images.githubusercontent.com/54015474/194406327-ae62c2f3-1da5-4686-8631-13d879a0edee.png">
## Steps to reproduce the bug
```python
from huggingface_hub import hf_hub_url
data_files = hf_hub_url(
repo_id="lewtun/github-issues",
filename="datasets-issues-with-hf-doc-builder.jsonl",
repo_type="dataset",
)
from datasets import load_dataset
issues_dataset = load_dataset("json", data_files=data_files, split="train")
issues_dataset
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.5.2
- Platform: macOS-10.16-x86_64-i386-64bit
- Python version: 3.9.12
- PyArrow version: 9.0.0
- Pandas version: 1.4.4
| 5,086 |
https://github.com/huggingface/datasets/issues/5085 | Filtering on an empty dataset returns a corrupted dataset. | [
"~~It seems like #5043 fix (merged recently) is the root cause of such behaviour. When we empty indices mapping (because the dataset length equals to zero), we can no longer get column item like: `ds_filter_2['sentence']` which uses\r\n`ds_filter_1._indices.column(0)`~~\r\n\r\n**UPDATE:**\r\nEmpty datasets are retu... | ## Describe the bug
When filtering a dataset twice, where the first result is an empty dataset, the second dataset seems corrupted.
## Steps to reproduce the bug
```python
datasets = load_dataset("glue", "sst2")
dataset_split = datasets['validation']
ds_filter_1 = dataset_split.filter(lambda x: False) # Some filtering condition that leads to an empty dataset
assert ds_filter_1.num_rows == 0
sentences = ds_filter_1['sentence']
assert len(sentences) == 0
ds_filter_2 = ds_filter_1.filter(lambda x: False) # Some other filtering condition
assert ds_filter_2.num_rows == 0
assert 'sentence' in ds_filter_2.column_names
sentences = ds_filter_2['sentence']
```
## Expected results
The last line should be returning an empty list, same as 4 lines above.
## Actual results
The last line currently raises `IndexError: index out of bounds`.
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 2.5.2
- Platform: macOS-11.6.6-x86_64-i386-64bit
- Python version: 3.9.11
- PyArrow version: 7.0.0
- Pandas version: 1.4.1
| 5,085 |
https://github.com/huggingface/datasets/issues/5083 | Support numpy/torch/tf/jax formatting for IterableDataset | [
"hii @lhoestq, can you assign this issue to me? Though i am new to open source still I would love to put my best foot forward. I can see there isn't anyone right now assigned to this issue.",
"Hi @zutarich ! This issue was fixed by #5852 - sorry I forgot to close it\r\n\r\nFeel free to look for other issues and p... | Right now `IterableDataset` doesn't do any formatting.
In particular this code should return a numpy array:
```python
from datasets import load_dataset
ds = load_dataset("imagenet-1k", split="train", streaming=True).with_format("np")
print(next(iter(ds))["image"])
```
Right now it returns a PIL.Image.
Setting `streaming=False` does return a numpy array after #5072 | 5,083 |
https://github.com/huggingface/datasets/issues/5081 | Bug loading `sentence-transformers/parallel-sentences` | [
"tagging @nreimers ",
"The dataset is sadly not really compatible to be loaded with `load_dataset`. So far it is better to git clone it and to use the files directly.\r\n\r\nA data loading script would be needed to be added to this dataset. But this was too much overhead / not really intuitive how to create it.",... | ## Steps to reproduce the bug
```python
from datasets import load_dataset
dataset = load_dataset("sentence-transformers/parallel-sentences")
```
raises this:
```
/home/phmay/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/download/streaming_download_manager.py:697: FutureWarning: the 'mangle_dupe_cols' keyword is deprecated and will be removed in a future version. Please take steps to stop the use of 'mangle_dupe_cols'
return pd.read_csv(xopen(filepath_or_buffer, "rb", use_auth_token=use_auth_token), **kwargs)
/home/phmay/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/download/streaming_download_manager.py:697: FutureWarning: the 'mangle_dupe_cols' keyword is deprecated and will be removed in a future version. Please take steps to stop the use of 'mangle_dupe_cols'
return pd.read_csv(xopen(filepath_or_buffer, "rb", use_auth_token=use_auth_token), **kwargs)
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Cell In [4], line 1
----> 1 dataset = load_dataset("sentence-transformers/parallel-sentences", split="train")
File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/load.py:1693, in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, ignore_verifications, keep_in_memory, save_infos, revision, use_auth_token, task, streaming, **config_kwargs)
1690 try_from_hf_gcs = path not in _PACKAGED_DATASETS_MODULES
1692 # Download and prepare data
-> 1693 builder_instance.download_and_prepare(
1694 download_config=download_config,
1695 download_mode=download_mode,
1696 ignore_verifications=ignore_verifications,
1697 try_from_hf_gcs=try_from_hf_gcs,
1698 use_auth_token=use_auth_token,
1699 )
1701 # Build dataset for splits
1702 keep_in_memory = (
1703 keep_in_memory if keep_in_memory is not None else is_small_dataset(builder_instance.info.dataset_size)
1704 )
File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/builder.py:807, in DatasetBuilder.download_and_prepare(self, output_dir, download_config, download_mode, ignore_verifications, try_from_hf_gcs, dl_manager, base_path, use_auth_token, file_format, max_shard_size, storage_options, **download_and_prepare_kwargs)
801 if not downloaded_from_gcs:
802 prepare_split_kwargs = {
803 "file_format": file_format,
804 "max_shard_size": max_shard_size,
805 **download_and_prepare_kwargs,
806 }
--> 807 self._download_and_prepare(
808 dl_manager=dl_manager,
809 verify_infos=verify_infos,
810 **prepare_split_kwargs,
811 **download_and_prepare_kwargs,
812 )
813 # Sync info
814 self.info.dataset_size = sum(split.num_bytes for split in self.info.splits.values())
File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/builder.py:898, in DatasetBuilder._download_and_prepare(self, dl_manager, verify_infos, **prepare_split_kwargs)
894 split_dict.add(split_generator.split_info)
896 try:
897 # Prepare split will record examples associated to the split
--> 898 self._prepare_split(split_generator, **prepare_split_kwargs)
899 except OSError as e:
900 raise OSError(
901 "Cannot find data file. "
902 + (self.manual_download_instructions or "")
903 + "\nOriginal error:\n"
904 + str(e)
905 ) from None
File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/builder.py:1513, in ArrowBasedBuilder._prepare_split(self, split_generator, file_format, max_shard_size)
1506 shard_id += 1
1507 writer = writer_class(
1508 features=writer._features,
1509 path=fpath.replace("SSSSS", f"{shard_id:05d}"),
1510 storage_options=self._fs.storage_options,
1511 embed_local_files=embed_local_files,
1512 )
-> 1513 writer.write_table(table)
1514 finally:
1515 num_shards = shard_id + 1
File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/arrow_writer.py:540, in ArrowWriter.write_table(self, pa_table, writer_batch_size)
538 if self.pa_writer is None:
539 self._build_writer(inferred_schema=pa_table.schema)
--> 540 pa_table = table_cast(pa_table, self._schema)
541 if self.embed_local_files:
542 pa_table = embed_table_storage(pa_table)
File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/table.py:2044, in table_cast(table, schema)
2032 """Improved version of pa.Table.cast.
2033
2034 It supports casting to feature types stored in the schema metadata.
(...)
2041 table (:obj:`pyarrow.Table`): the casted table
2042 """
2043 if table.schema != schema:
-> 2044 return cast_table_to_schema(table, schema)
2045 elif table.schema.metadata != schema.metadata:
2046 return table.replace_schema_metadata(schema.metadata)
File ~/miniconda3/envs/paraphrase-mining/lib/python3.9/site-packages/datasets/table.py:2005, in cast_table_to_schema(table, schema)
2003 features = Features.from_arrow_schema(schema)
2004 if sorted(table.column_names) != sorted(features):
-> 2005 raise ValueError(f"Couldn't cast\n{table.schema}\nto\n{features}\nbecause column names don't match")
2006 arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]
2007 return pa.Table.from_arrays(arrays, schema=schema)
ValueError: Couldn't cast
Action taken on Parliament's resolutions: see Minutes: string
Následný postup na základě usnesení Parlamentu: viz zápis: string
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 742
to
{'Membership of Parliament: see Minutes': Value(dtype='string', id=None), 'Състав на Парламента: вж. протоколи': Value(dtype='string', id=None)}
because column names don't match
```
## Expected results
no error
## Actual results
error
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version:
- Platform: Linux
- Python version: Python 3.9.13
- PyArrow version: pyarrow 9.0.0
- transformers 4.22.2
- datasets 2.5.2 | 5,081 |
https://github.com/huggingface/datasets/issues/5080 | Use hfh for caching | [
"There is some discussion in https://github.com/huggingface/huggingface_hub/pull/1088 if it can help :)"
] | ## Is your feature request related to a problem?
As previously discussed in our meeting with @Wauplin and agreed on our last datasets team sync meeting, I'm investigating how `datasets` can use `hfh` for caching.
## Describe the solution you'd like
Due to the peculiarities of the `datasets` cache, I would propose adopting `hfh` caching system in stages.
First, we could easily start using `hfh` caching for:
- dataset Python scripts
- dataset READMEs
- dataset infos JSON files (now deprecated)
Second, we could also use `hfh` caching for data files downloaded from the Hub.
Further investigation is needed for:
- files downloaded from non-Hub hosts
- extracted files from downloaded archive/compressed files
- generated Arrow files
## Additional context
Docs about the `hfh` caching system:
- [Manage huggingface_hub cache-system](https://huggingface.co/docs/huggingface_hub/main/en/how-to-cache)
- [Cache-system reference](https://huggingface.co/docs/huggingface_hub/main/en/package_reference/cache)
The `transformers` library has already adopted `hfh` for caching. See:
- huggingface/transformers#18438
- huggingface/transformers#18857
- huggingface/transformers#18966
| 5,080 |
https://github.com/huggingface/datasets/issues/5075 | Throw EnvironmentError when token is not present | [
"@mariosasko I've raised a PR #5076 against this issue. Please help to review. Thanks."
] | Throw EnvironmentError instead of OSError ([link](https://github.com/huggingface/datasets/blob/6ad430ba0cdeeb601170f732d4bd977f5c04594d/src/datasets/arrow_dataset.py#L4306) to the line) in `push_to_hub` when the Hub token is not present. | 5,075 |
https://github.com/huggingface/datasets/issues/5074 | Replace AssertionErrors with more meaningful errors | [
"Hi, can I pick up this issue?",
"#self-assign",
"Looks like the top-level `datasource` directory was removed when https://github.com/huggingface/datasets/pull/4974 was merged, so there are 3 source files to fix."
] | Replace the AssertionErrors with more meaningful errors such as ValueError, TypeError, etc.
The files with AssertionErrors that need to be replaced:
```
src/datasets/arrow_reader.py
src/datasets/builder.py
src/datasets/utils/version.py
``` | 5,074 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.