Datasets:

Ayon128

/

github-issues

like 0

https://api.github.com/repos/huggingface/datasets/issues/3750

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3750/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3750/comments

https://api.github.com/repos/huggingface/datasets/issues/3750/events

https://github.com/huggingface/datasets/issues/3750

I_kwDODunzps5EF8SL

`NonMatchingSplitsSizesError` for cats_vs_dogs dataset

closed

CONTRIBUTOR

## Describe the bug Cannot download cats_vs_dogs dataset due to `NonMatchingSplitsSizesError`. ## Steps to reproduce the bug ```python from datasets import load_dataset dataset = load_dataset("cats_vs_dogs") ``` ## Expected results Loading is successful. ## Actual results ``` NonMatchingSplitsSizesError: [{'expected': SplitInfo(name='train', num_bytes=7503250, num_examples=23422, dataset_name='cats_vs_dogs'), 'recorded': SplitInfo(name='train', num_bytes=7262410, num_examples=23410, dataset_name='cats_vs_dogs')}] ``` ## Environment info Reproduced on a fresh [Colab notebook](https://colab.research.google.com/drive/13GTvrSJbBGvL2ybDdXCBZwATd6FOkMub?usp=sharing). ## Additional Context Originally reported in https://github.com/huggingface/transformers/issues/15698. cc @mariosasko

https://api.github.com/repos/huggingface/datasets/issues/3750/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3749

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3749/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3749/comments

https://api.github.com/repos/huggingface/datasets/issues/3749/events

https://github.com/huggingface/datasets/pull/3749

PR_kwDODunzps4zCKqg

Add tqdm arguments

closed

NONE

In this PR, tqdm arguments can be passed to the map() function and such, in order to be more flexible.

https://api.github.com/repos/huggingface/datasets/issues/3749/timeline

https://api.github.com/repos/huggingface/datasets/issues/3748

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3748/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3748/comments

https://api.github.com/repos/huggingface/datasets/issues/3748/events

https://github.com/huggingface/datasets/pull/3748

PR_kwDODunzps4zCEyM

Add tqdm arguments

closed

NONE

In this PR, there are two changes. 1. It is able to show the progress bar by adding the length of the iterator. 2. Pass in tqdm_kwargs so that can enable more feasibility for the control of tqdm library.

https://api.github.com/repos/huggingface/datasets/issues/3748/timeline

https://api.github.com/repos/huggingface/datasets/issues/3747

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3747/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3747/comments

https://api.github.com/repos/huggingface/datasets/issues/3747/events

https://github.com/huggingface/datasets/issues/3747

I_kwDODunzps5EDMoW

Passing invalid subset should throw an error

open

CONTRIBUTOR

## Describe the bug Only some datasets have a subset (as in `load_dataset(name, subset)`). If you pass an invalid subset, an error should be thrown. ## Steps to reproduce the bug ```python import datasets datasets.load_dataset('rotten_tomatoes', 'asdfasdfa') ``` ## Expected results This should break, since `'asdfasdfa'` isn't a subset of the `rotten_tomatoes` dataset. ## Actual results This API call silently succeeds.

https://api.github.com/repos/huggingface/datasets/issues/3747/timeline

https://api.github.com/repos/huggingface/datasets/issues/3746

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3746/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3746/comments

https://api.github.com/repos/huggingface/datasets/issues/3746/events

https://github.com/huggingface/datasets/pull/3746

PR_kwDODunzps4zAS-C

Use the same seed to shuffle shards and metadata in streaming mode

closed

MEMBER

When shuffling in streaming mode, those two entangled lists are shuffled independently. In this PR I changed this to shuffle the lists of same length with the exact same seed, in order for the files and metadata to still be aligned. ```python gen_kwargs = { "files": [os.path.join(data_dir, filename) for filename in all_files], "metadata_files": [all_metadata[filename] for filename in all_files], } ``` IMO this is important to avoid big but silent issues. Fix https://github.com/huggingface/datasets/issues/3744

https://api.github.com/repos/huggingface/datasets/issues/3746/timeline

https://api.github.com/repos/huggingface/datasets/issues/3745

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3745/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3745/comments

https://api.github.com/repos/huggingface/datasets/issues/3745/events

https://github.com/huggingface/datasets/pull/3745

PR_kwDODunzps4y__m2

Add mIoU metric

closed

CONTRIBUTOR

This PR adds the mean Intersection-over-Union metric to the library, useful for tasks like semantic segmentation. It is entirely based on mmseg's [implementation](https://github.com/open-mmlab/mmsegmentation/blob/master/mmseg/core/evaluation/metrics.py). I've removed any PyTorch dependency, and rely on Numpy only.

https://api.github.com/repos/huggingface/datasets/issues/3745/timeline

https://api.github.com/repos/huggingface/datasets/issues/3744

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3744/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3744/comments

https://api.github.com/repos/huggingface/datasets/issues/3744/events

https://github.com/huggingface/datasets/issues/3744

I_kwDODunzps5ECVCt

Better shards shuffling in streaming mode

closed

MEMBER

Sometimes a dataset script has a `_split_generators` that returns several files as well as the corresponding metadata of each file. It often happens that they end up in two separate lists in the `gen_kwargs`: ```python gen_kwargs = { "files": [os.path.join(data_dir, filename) for filename in all_files], "metadata_files": [all_metadata[filename] for filename in all_files], } ``` It happened for Multilingual Spoken Words for example in #3666 However currently **the two lists are shuffled independently** when shuffling the shards in streaming mode. This leads to `_generate_examples` not having the right metadata for each file. To prevent this issue I suggest that we always shuffle lists of the same length the exact same way to avoid such a big but silent issue. cc @polinaeterna

https://api.github.com/repos/huggingface/datasets/issues/3744/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3743

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3743/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3743/comments

https://api.github.com/repos/huggingface/datasets/issues/3743/events

https://github.com/huggingface/datasets/pull/3743

PR_kwDODunzps4y-2Do

initial monash time series forecasting repository

closed

CONTRIBUTOR

https://api.github.com/repos/huggingface/datasets/issues/3743/timeline

https://api.github.com/repos/huggingface/datasets/issues/3742

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3742/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3742/comments

https://api.github.com/repos/huggingface/datasets/issues/3742/events

https://github.com/huggingface/datasets/pull/3742

PR_kwDODunzps4y-1v5

Fix ValueError message formatting in int2str

closed

CONTRIBUTOR

Hi! I bumped into this particular `ValueError` during my work (because an instance of `np.int64` was passed instead of regular Python `int`), and so I had to `print(type(values))` myself. Apparently, it's just the missing `f` to make message an f-string. It ain't much for a contribution, but it's honest work. Hope it spares someone else a few seconds in the future 😃

https://api.github.com/repos/huggingface/datasets/issues/3742/timeline

https://api.github.com/repos/huggingface/datasets/issues/3741

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3741/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3741/comments

https://api.github.com/repos/huggingface/datasets/issues/3741/events

https://github.com/huggingface/datasets/pull/3741

PR_kwDODunzps4y-syt

Rm sphinx doc

closed

CONTRIBUTOR

Checklist - [x] Update circle ci yaml - [x] Delete sphinx static & python files in docs dir - [x] Update readme in docs dir - [ ] Update docs config in setup.py

https://api.github.com/repos/huggingface/datasets/issues/3741/timeline

https://api.github.com/repos/huggingface/datasets/issues/3740

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3740/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3740/comments

https://api.github.com/repos/huggingface/datasets/issues/3740/events

https://github.com/huggingface/datasets/pull/3740

PR_kwDODunzps4y9XAP

Support streaming for pubmed

closed

CONTRIBUTOR

This PR makes some minor changes to the `pubmed` dataset to allow for `streaming=True`. Fixes #3739. Basically, I followed the C4 dataset which works in streaming mode as an example, and made the following changes: * Change URL prefix from `ftp://` to `https://` * Explicilty `open` the filename and pass the XML contents to `etree.fromstring(xml_str)` The Github diff tool makes it look like the changes are larger than they are, sorry about that. I tested locally and the `pubmed` dataset now works in both normal and streaming modes. There is some overhead at the start of each shard in streaming mode as building the XML tree online is quite slow (each pubmed .xml.gz file is ~20MB), but the overhead gets amortized over all the samples in the shard. On my laptop with a single CPU worker I am able to stream at about ~600 samples/s.

https://api.github.com/repos/huggingface/datasets/issues/3740/timeline

https://api.github.com/repos/huggingface/datasets/issues/3739

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3739/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3739/comments

https://api.github.com/repos/huggingface/datasets/issues/3739/events

https://github.com/huggingface/datasets/issues/3739

I_kwDODunzps5D-Arl

Pubmed dataset does not work in streaming mode

closed

CONTRIBUTOR

## Describe the bug Trying to use the `pubmed` dataset with `streaming=True` fails. ## Steps to reproduce the bug ```python import datasets pubmed_train = datasets.load_dataset('pubmed', split='train', streaming=True) print (next(iter(pubmed_train))) ``` ## Expected results I would expect to see the first training sample from the pubmed dataset. ## Actual results ``` Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/Users/abhinav/Documents/mosaicml/mosaicml_venv/lib/python3.8/site-packages/datasets/iterable_dataset.py", line 367, in __iter__ for key, example in self._iter(): File "/Users/abhinav/Documents/mosaicml/mosaicml_venv/lib/python3.8/site-packages/datasets/iterable_dataset.py", line 364, in _iter yield from ex_iterable File "/Users/abhinav/Documents/mosaicml/mosaicml_venv/lib/python3.8/site-packages/datasets/iterable_dataset.py", line 79, in __iter__ for key, example in self.generate_examples_fn(**self.kwargs): File "/Users/abhinav/.cache/huggingface/modules/datasets_modules/datasets/pubmed/9715addf10c42a7877a2149ae0c5f2fddabefc775cd1bd9b03ac3f012b86ce46/pubmed.py", line 373, in _generate_examples tree = etree.parse(filename) File "/Library/Developer/CommandLineTools/Library/Frameworks/Python3.framework/Versions/3.8/lib/python3.8/xml/etree/ElementTree.py", line 1202, in parse tree.parse(source, parser) File "/Library/Developer/CommandLineTools/Library/Frameworks/Python3.framework/Versions/3.8/lib/python3.8/xml/etree/ElementTree.py", line 584, in parse source = open(source, "rb") FileNotFoundError: [Errno 2] No such file or directory: 'gzip://pubmed21n0001.xml::ftp://ftp.ncbi.nlm.nih.gov/pubmed/baseline/pubmed21n0001.xml.gz' ``` ## Environment info <!-- You can run the command `datasets-cli env` and copy-and-paste its output below. --> - `datasets` version: 1.18.2 - Platform: macOS-11.4-x86_64-i386-64bit - Python version: 3.8.2 - PyArrow version: 6.0.0 ## Comments The error looks like an issue with `open` vs. `xopen` inside the `xml` package. It looks like it's trying to open the remote source URL, which has been edited with prefix `gzip://...`. Maybe there can be an explicit `xopen` before passing the raw data to `etree`, something like: ```python # Before tree = etree.parse(filename) root = tree.getroot() # After with xopen(filename) as f: data_str = f.read() root = etree.fromstring(data_str) ```

https://api.github.com/repos/huggingface/datasets/issues/3739/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3738

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3738/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3738/comments

https://api.github.com/repos/huggingface/datasets/issues/3738/events

https://github.com/huggingface/datasets/issues/3738

I_kwDODunzps5D9Yad

For data-only datasets, streaming and non-streaming don't behave the same

open

CONTRIBUTOR

See https://huggingface.co/datasets/huggingface/transformers-metadata: it only contains two JSON files. In streaming mode, the files are concatenated, and thus the rows might be dictionaries with different keys: ```python import datasets as ds iterable_dataset = ds.load_dataset("huggingface/transformers-metadata", split="train", streaming=True); rows = list(iterable_dataset.take(100)) rows[0] # {'model_type': 'albert', 'pytorch': True, 'tensorflow': True, 'flax': True, 'processor': 'AutoTokenizer'} rows[99] # {'model_class': 'BartModel', 'pipeline_tag': 'feature-extraction', 'auto_class': 'AutoModel'} ``` In normal mode, an exception is thrown: ```python import datasets as ds dataset = ds.load_dataset("huggingface/transformers-metadata", split="train"); ``` ``` ValueError: Couldn't cast model_class: string pipeline_tag: string auto_class: string to {'model_type': Value(dtype='string', id=None), 'pytorch': Value(dtype='bool', id=None), 'tensorflow': Value(dtype='bool', id=None), 'flax': Value(dtype='bool', id=None), 'processor': Value(dtype='string', id=None)} because column names don't match ```

https://api.github.com/repos/huggingface/datasets/issues/3738/timeline

https://api.github.com/repos/huggingface/datasets/issues/3737

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3737/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3737/comments

https://api.github.com/repos/huggingface/datasets/issues/3737/events

https://github.com/huggingface/datasets/pull/3737

PR_kwDODunzps4y7uFf

Make RedCaps streamable

closed

CONTRIBUTOR

Make RedCaps streamable. @lhoestq Using `data/redcaps_v1.0_annotations.zip` as a download URL gives an error locally when running `datasets-cli test` (will investigate this another time)

https://api.github.com/repos/huggingface/datasets/issues/3737/timeline

https://api.github.com/repos/huggingface/datasets/issues/3736

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3736/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3736/comments

https://api.github.com/repos/huggingface/datasets/issues/3736/events

https://github.com/huggingface/datasets/pull/3736

PR_kwDODunzps4y7rMR

Local paths in common voice

closed

MEMBER

Continuation of https://github.com/huggingface/datasets/pull/3664: - pass the `streaming` parameter to _split_generator - update @anton-l's code to use this parameter for `common_voice` - add a comment to explain why we use `download_and_extract` in non-streaming and `iter_archive` in streaming Now the `common_voice` dataset has a local path back in `ds["path"]`, and this field is `None` in streaming mode. cc @patrickvonplaten @anton-l @albertvillanova Fix #3663.

https://api.github.com/repos/huggingface/datasets/issues/3736/timeline

https://api.github.com/repos/huggingface/datasets/issues/3735

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3735/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3735/comments

https://api.github.com/repos/huggingface/datasets/issues/3735/events

https://github.com/huggingface/datasets/issues/3735

I_kwDODunzps5D9FxT

Performance of `datasets` at scale

open

MEMBER

# Performance of `datasets` at 1TB scale ## What is this? During the processing of a large dataset I monitored the performance of the `datasets` library to see if there are any bottlenecks. The insights of this analysis could guide the decision making to improve the performance of the library. ## Dataset The dataset is a 1.1TB extract from GitHub with 120M code files and is stored as 5000 `.json.gz` files. The goal of the preprocessing is to remove duplicates and filter files based on their stats. While the calculating of the hashes for deduplication and stats for filtering can be parallelized the filtering itself is run with a single process. After processing the files are pushed to the hub. ## Machine The experiment was run on a `m1` machine on GCP with 96 CPU cores and 1.3TB RAM. ## Performance breakdown - Loading the data **3.5h** (_30sec_ from cache) - **1h57min** single core loading (not sure what is going on here, corresponds to second progress bar) - **1h10min** multi core json reading - **20min** remaining time before and after the two main processes mentioned above - Process the data **2h** (_20min_ from cache) - **20min** Getting reading for processing - **40min** Hashing and files stats (96 workers) - **58min** Deduplication filtering (single worker) - Save parquet files **5h** - Saving 1000 parquet files (16 workers) - Push to hub **37min** - **34min** git add - **3min** git push (several hours with `Repository.git_push()`) ## Conclusion It appears that loading and saving the data is the main bottleneck at that scale (**8.5h**) whereas processing (**2h**) and pushing the data to the hub (**0.5h**) is relatively fast. To optimize the performance at this scale it would make sense to consider such an end-to-end example and target the bottlenecks which seem to be loading from and saving to disk. The processing itself seems to run relatively fast. ## Notes - map operation on a 1TB dataset with 96 workers requires >1TB RAM - map operation does not maintain 100% CPU utilization with 96 workers - sometimes when the script crashes all the data files have a corresponding `*.lock` file in the data folder (or multiple e.g. `*.lock.lock` when it happened a several times). This causes the cache **not** to be triggered (which is significant at that scale) - i guess because there are new data files - parallelizing `to_parquet` decreased the saving time from 17h to 5h, however adding more workers at this point had almost no effect. not sure if this is: a) a bug in my parallelization logic, b) i/o limit to load data form disk to memory or c) i/o limit to write from memory to disk. - Using `Repository.git_push()` was much slower than using command line `git-lfs` - 10-20MB/s vs. 300MB/s! The `Dataset.push_to_hub()` function is even slower as it only uploads one file at a time with only a few MB/s, whereas `Repository.git_push()` pushes files in parallel (each at a similar speed). cc @lhoestq @julien-c @LysandreJik @SBrandeis

https://api.github.com/repos/huggingface/datasets/issues/3735/timeline

https://api.github.com/repos/huggingface/datasets/issues/3734

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3734/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3734/comments

https://api.github.com/repos/huggingface/datasets/issues/3734/events

https://github.com/huggingface/datasets/pull/3734

PR_kwDODunzps4y7ZU2

Fix bugs in NewsQA dataset

closed

MEMBER

Fix #3733.

https://api.github.com/repos/huggingface/datasets/issues/3734/timeline

https://api.github.com/repos/huggingface/datasets/issues/3733

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3733/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3733/comments

https://api.github.com/repos/huggingface/datasets/issues/3733/events

https://github.com/huggingface/datasets/issues/3733

I_kwDODunzps5D8zFy

Bugs in NewsQA dataset

closed

MEMBER

## Describe the bug NewsQA dataset has the following bugs: - the field `validated_answers` is an exact copy of the field `answers` but with the addition of `'count': [0]` to each dict - the field `badQuestion` does not appear in `answers` nor `validated_answers` ## Steps to reproduce the bug By inspecting the dataset script we can see that: - the parsing of `validated_answers` is a copy-paste of the one for `answers` - the `badQuestion` field is ignored in the parsing of both `answers` and `validated_answers`

https://api.github.com/repos/huggingface/datasets/issues/3733/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3732

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3732/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3732/comments

https://api.github.com/repos/huggingface/datasets/issues/3732/events

https://github.com/huggingface/datasets/pull/3732

PR_kwDODunzps4y7PTU

Support streaming in size estimation function in `push_to_hub`

closed

CONTRIBUTOR

This PR adds the streamable version of `os.path.getsize` (`fsspec` can return `None`, so we fall back to `fs.open` to make it more robust) to account for possible streamable paths in the nested `extra_nbytes_visitor` function inside `push_to_hub`.

https://api.github.com/repos/huggingface/datasets/issues/3732/timeline

https://api.github.com/repos/huggingface/datasets/issues/3731

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3731/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3731/comments

https://api.github.com/repos/huggingface/datasets/issues/3731/events

https://github.com/huggingface/datasets/pull/3731

PR_kwDODunzps4y5-hi

Fix Multi-News dataset metadata and card

closed

MEMBER

Fix #3730.

https://api.github.com/repos/huggingface/datasets/issues/3731/timeline

https://api.github.com/repos/huggingface/datasets/issues/3730

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3730/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3730/comments

https://api.github.com/repos/huggingface/datasets/issues/3730/events

https://github.com/huggingface/datasets/issues/3730

I_kwDODunzps5D7BYN

Checksum Error when loading multi-news dataset

closed

NONE

## Describe the bug When using the load_dataset function from datasets module to load the Multi-News dataset, does not load the dataset but throws Checksum Error instead. ## Steps to reproduce the bug ```python from datasets import load_dataset dataset = load_dataset("multi_news") ``` ## Expected results Should download and load Multi-News dataset. ## Actual results Throws the following error and cannot load data successfully: ``` NonMatchingChecksumError: Checksums didn't match for dataset source files: ['https://drive.google.com/uc?export=download&id=1vRY2wM6rlOZrf9exGTm5pXj5ExlVwJ0C'] ``` Could this issue please be looked at? Thanks!

https://api.github.com/repos/huggingface/datasets/issues/3730/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3729

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3729/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3729/comments

https://api.github.com/repos/huggingface/datasets/issues/3729/events

https://github.com/huggingface/datasets/issues/3729

I_kwDODunzps5D6dcq

Wrong number of examples when loading a text dataset

closed

NONE

## Describe the bug when I use load_dataset to read a txt file I find that the number of the samples is incorrect ## Steps to reproduce the bug ``` fr = open('train.txt','r',encoding='utf-8').readlines() print(len(fr)) # 1199637 datasets = load_dataset('text', data_files={'train': ['train.txt']}, streaming=False) print(len(datasets['train'])) # 1199649 ``` I also use command line operation to verify it ``` $ wc -l train.txt 1199637 train.txt ``` ## Expected results please fix that issue ## Environment info <!-- You can run the command `datasets-cli env` and copy-and-paste its output below. --> - `datasets` version: 1.8.3 - Platform:windows&linux - Python version:3.7 - PyArrow version:6.0.1

https://api.github.com/repos/huggingface/datasets/issues/3729/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3728

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3728/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3728/comments

https://api.github.com/repos/huggingface/datasets/issues/3728/events

https://github.com/huggingface/datasets/issues/3728

I_kwDODunzps5D6GS-

VoxPopuli

closed

MEMBER

## Adding a Dataset - **Name:** VoxPopuli - **Description:** A Large-Scale Multilingual Speech Corpus - **Paper:** https://arxiv.org/pdf/2101.00390.pdf - **Data:** https://github.com/facebookresearch/voxpopuli - **Motivation:** one of the largest (if not the largest) multilingual speech corpus: 400K hours of multilingual unlabeled speech + 17k hours of labeled speech Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md). 👀 @kahne @Molugan

https://api.github.com/repos/huggingface/datasets/issues/3728/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3727

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3727/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3727/comments

https://api.github.com/repos/huggingface/datasets/issues/3727/events

https://github.com/huggingface/datasets/pull/3727

PR_kwDODunzps4y34JN

Patch all module attributes in its namespace

closed

MEMBER

When patching module attributes, only those defined in its `__all__` variable were considered by default (only falling back to `__dict__` if `__all__` was None). However those are only a subset of all the module attributes in its namespace (`__dict__` variable). This PR fixes the problem of modules that have non-None `__all__` variable, but try to access an attribute present in `__dict__` (and not in `__all__`). For example, `pandas` has attribute `__version__` only present in `__dict__`. - Before version 1.4, pandas `__all__` was None, thus all attributes in `__dict__` were patched - From version 1.4, pandas `__all__` is not None, thus attributes in `__dict__` not present in `__all__` are ignored Fix #3724. CC: @severo @lvwerra

https://api.github.com/repos/huggingface/datasets/issues/3727/timeline

https://api.github.com/repos/huggingface/datasets/issues/3726

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3726/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3726/comments

https://api.github.com/repos/huggingface/datasets/issues/3726/events

https://github.com/huggingface/datasets/pull/3726

PR_kwDODunzps4y3iSv

Use config pandas version in CSV dataset builder

closed

MEMBER

Fix #3724.

https://api.github.com/repos/huggingface/datasets/issues/3726/timeline

https://api.github.com/repos/huggingface/datasets/issues/3725

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3725/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3725/comments

https://api.github.com/repos/huggingface/datasets/issues/3725/events

https://github.com/huggingface/datasets/pull/3725

PR_kwDODunzps4y3bOG

Pin pandas to avoid bug in streaming mode

closed

MEMBER

Temporarily pin pandas version to avoid bug in streaming mode (patching no longer works). Related to #3724.

https://api.github.com/repos/huggingface/datasets/issues/3725/timeline

https://api.github.com/repos/huggingface/datasets/issues/3724

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3724/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3724/comments

https://api.github.com/repos/huggingface/datasets/issues/3724/events

https://github.com/huggingface/datasets/issues/3724

I_kwDODunzps5D4SGh

Bug while streaming CSV dataset with pandas 1.4

closed

MEMBER

## Describe the bug If we upgrade to pandas `1.4`, the patching of the pandas module is no longer working ``` AttributeError: '_PatchedModuleObj' object has no attribute '__version__' ``` ## Steps to reproduce the bug ``` pip install pandas==1.4 ``` ```python from datasets import load_dataset ds = load_dataset("lvwerra/red-wine", split="train", streaming=True) item = next(iter(ds)) item ```

https://api.github.com/repos/huggingface/datasets/issues/3724/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3723

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3723/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3723/comments

https://api.github.com/repos/huggingface/datasets/issues/3723/events

https://github.com/huggingface/datasets/pull/3723

PR_kwDODunzps4y3RuI

Fix flatten of complex feature types

closed

CONTRIBUTOR

Fix `flatten` for the following feature types: Image/Audio, Translation, and TranslationVariableLanguages. Inspired by `cast`/`table_cast`, I've introduced a `table_flatten` function to handle the Image/Audio types. CC: @SBrandeis Fix #3686.

https://api.github.com/repos/huggingface/datasets/issues/3723/timeline

https://api.github.com/repos/huggingface/datasets/issues/3722

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3722/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3722/comments

https://api.github.com/repos/huggingface/datasets/issues/3722/events

https://github.com/huggingface/datasets/pull/3722

PR_kwDODunzps4y3NrP

added electricity load diagram dataset

closed

CONTRIBUTOR

Initial Electricity Load Diagram time series dataset.

https://api.github.com/repos/huggingface/datasets/issues/3722/timeline

https://api.github.com/repos/huggingface/datasets/issues/3721

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3721/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3721/comments

https://api.github.com/repos/huggingface/datasets/issues/3721/events

https://github.com/huggingface/datasets/pull/3721

PR_kwDODunzps4yzXCd

Multi-GPU support for `FaissIndex`

closed

CONTRIBUTOR

Per #3716 , current implementation does not take into consideration that `faiss` can run on multiple GPUs. In this commit, I provided multi-GPU support for `FaissIndex` by modifying the device management in `IndexableMixin.add_faiss_index` and `FaissIndex.load`. Now users are able to pass in 1. a positive integer (as usual) to use 1 GPU 2. a negative integer `-1` to use all GPUs 3. a list of integers e.g. `[0, 1]` to run only on those GPUs 4. Of course, passing in nothing still runs on CPU. This closes: #3716

https://api.github.com/repos/huggingface/datasets/issues/3721/timeline

https://api.github.com/repos/huggingface/datasets/issues/3720

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3720/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3720/comments

https://api.github.com/repos/huggingface/datasets/issues/3720/events

https://github.com/huggingface/datasets/issues/3720

I_kwDODunzps5DzXA4

Builder Configuration Update Required on Common Voice Dataset

open

NONE

Missing language in Common Voice dataset **Link:** https://huggingface.co/datasets/common_voice I tried to call the Urdu dataset using `load_dataset("common_voice", "ur", split="train+validation")` but couldn't due to builder configuration not found. I checked the source file here for the languages support: https://github.com/huggingface/datasets/blob/master/datasets/common_voice/common_voice.py and Urdu isn't included there. I assume a quick update will fix the issue as Urdu speech is now available at the Common Voice dataset. Am I the one who added this dataset? No

https://api.github.com/repos/huggingface/datasets/issues/3720/timeline

https://api.github.com/repos/huggingface/datasets/issues/3719

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3719/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3719/comments

https://api.github.com/repos/huggingface/datasets/issues/3719/events

https://github.com/huggingface/datasets/pull/3719

PR_kwDODunzps4yyFv7

Check if indices values in `Dataset.select` are within bounds

closed

CONTRIBUTOR

Fix #3707 Instead of reusing `_check_valid_index_key` from `datasets.formatting`, I defined a new function to provide a more meaningful error message.

https://api.github.com/repos/huggingface/datasets/issues/3719/timeline

https://api.github.com/repos/huggingface/datasets/issues/3718

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3718/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3718/comments

https://api.github.com/repos/huggingface/datasets/issues/3718/events

https://github.com/huggingface/datasets/pull/3718

PR_kwDODunzps4yx8r2

Fix Evidence Infer Treatment dataset

closed

MEMBER

This PR: - fixes a bug in the script, by removing an unnamed column with the row index: fix KeyError - fix the metadata JSON, by adding both configurations (1.1 and 2.0): fix ExpectedMoreDownloadedFiles - updates the dataset card Fix #3515.

https://api.github.com/repos/huggingface/datasets/issues/3718/timeline

https://api.github.com/repos/huggingface/datasets/issues/3717

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3717/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3717/comments

https://api.github.com/repos/huggingface/datasets/issues/3717/events

https://github.com/huggingface/datasets/issues/3717

I_kwDODunzps5DyAkn

wrong condition in `Features ClassLabel encode_example`

closed

NONE

## Describe the bug The `encode_example` function in *features.py* seems to have a wrong condition. ```python if not -1 <= example_data < self.num_classes: raise ValueError(f"Class label {example_data:d} greater than configured num_classes {self.num_classes}") ``` ## Expected results The `not - 1` condition change the result of the condition. For instance, if `example_data` equals 4 and ` self.num_classes` equals 4 too, `example_data < self.num_classes` will give `False` as expected . But if i add the `not - 1` condition, `not -1 <= example_data < self.num_classes` will give `True` and raise an exception. ## Environment info - `datasets` version: 1.18.3 - Python version: 3.8.10 - PyArrow version: 7.00

https://api.github.com/repos/huggingface/datasets/issues/3717/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3716

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3716/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3716/comments

https://api.github.com/repos/huggingface/datasets/issues/3716/events

https://github.com/huggingface/datasets/issues/3716

I_kwDODunzps5Dwqp0

`FaissIndex` to support multiple GPU and `custom_index`

closed

CONTRIBUTOR

**Is your feature request related to a problem? Please describe.** Currently, because `device` is of the type `int | None`, to leverage `faiss-gpu`'s multi-gpu support, you need to create a `custom_index`. However, if using a `custom_index` created by e.g. `faiss.index_cpu_to_all_gpus`, then `FaissIndex.save` does not work properly because it checks the device id (which is an int, so no multiple GPUs). **Describe the solution you'd like** I would like `FaissIndex` to support multiple GPUs, by passing in a list to `add_faiss_index`. **Describe alternatives you've considered** Alternatively, I would like it to at least provide a warning cause it wasn't the behavior that I expected. **Additional context** Relavent source code here: https://github.com/huggingface/datasets/blob/6ed6ac9448311930557810383d2cfd4fe6aae269/src/datasets/search.py#L340-L349 Device management needs changing to support multiple GPUs, probably by `isinstance` calls. I can provide a PR if you like :) Thanks for reading!

https://api.github.com/repos/huggingface/datasets/issues/3716/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3715

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3715/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3715/comments

https://api.github.com/repos/huggingface/datasets/issues/3715/events

https://github.com/huggingface/datasets/pull/3715

PR_kwDODunzps4yuKJj

Fix bugs in msr_sqa dataset

closed

CONTRIBUTOR

The last version has many problems, 1) Errors in table load-in. Split by a single comma instead of using pandas is wrong. 2) id reduplicated in _generate_examples function. 3) Missing information of history questions which make it hard to use. I fix it refer to https://github.com/HKUNLP/UnifiedSKG. And we test it to perform normally.

https://api.github.com/repos/huggingface/datasets/issues/3715/timeline

https://api.github.com/repos/huggingface/datasets/issues/3714

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3714/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3714/comments

https://api.github.com/repos/huggingface/datasets/issues/3714/events

https://github.com/huggingface/datasets/issues/3714

I_kwDODunzps5Dt5g6

tatoeba_mt: File not found error and key error

closed

NONE

## Dataset viewer issue for 'tatoeba_mt' **Link:** https://huggingface.co/datasets/Helsinki-NLP/tatoeba_mt My data loader script does not seem to work. The files are part of the local repository but cannot be found. An example where it should work is the subset for "afr-eng". Another problem is that I do not have validation data for all subsets and I don't know how to properly check whether validation exists in the configuration before I try to download it. An example is the subset for "afr-deu". Am I the one who added this dataset ? Yes

https://api.github.com/repos/huggingface/datasets/issues/3714/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3713

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3713/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3713/comments

https://api.github.com/repos/huggingface/datasets/issues/3713/events

https://github.com/huggingface/datasets/pull/3713

PR_kwDODunzps4yso6D

Rm sphinx doc

closed

CONTRIBUTOR

Checklist - [x] Update circle ci yaml - [x] Delete sphinx static & python files in docs dir - [x] Update readme in docs dir - [ ] Update docs config in setup.py

https://api.github.com/repos/huggingface/datasets/issues/3713/timeline

https://api.github.com/repos/huggingface/datasets/issues/3712

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3712/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3712/comments

https://api.github.com/repos/huggingface/datasets/issues/3712/events

https://github.com/huggingface/datasets/pull/3712

PR_kwDODunzps4ynVYy

Fix the error of msr_sqa dataset

closed

CONTRIBUTOR

Fix the error of _load_table_data function in msr_sqa dataset, it is wrong to use comma to split each row.

https://api.github.com/repos/huggingface/datasets/issues/3712/timeline

https://api.github.com/repos/huggingface/datasets/issues/3711

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3711/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3711/comments

https://api.github.com/repos/huggingface/datasets/issues/3711/events

https://github.com/huggingface/datasets/pull/3711

PR_kwDODunzps4ymmlK

Fix the error of _load_table_data function in msr_sqa dataset

closed

CONTRIBUTOR

The _load_table_data function from the last version is wrong, it is wrong to use comma to split each row.

https://api.github.com/repos/huggingface/datasets/issues/3711/timeline

https://api.github.com/repos/huggingface/datasets/issues/3710

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3710/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3710/comments

https://api.github.com/repos/huggingface/datasets/issues/3710/events

https://github.com/huggingface/datasets/pull/3710

PR_kwDODunzps4ymQMQ

Fix CI code quality issue

closed

MEMBER

Fix CI code quality issue introduced by #3695.

https://api.github.com/repos/huggingface/datasets/issues/3710/timeline

https://api.github.com/repos/huggingface/datasets/issues/3709

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3709/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3709/comments

https://api.github.com/repos/huggingface/datasets/issues/3709/events

https://github.com/huggingface/datasets/pull/3709

PR_kwDODunzps4yi0J4

Set base path to hub url for canonical datasets

closed

MEMBER

This should allow canonical datasets to use relative paths to download data files from the Hub cc @polinaeterna this will be useful if we have audio datasets that are canonical and for which you'd like to host data files

https://api.github.com/repos/huggingface/datasets/issues/3709/timeline

https://api.github.com/repos/huggingface/datasets/issues/3708

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3708/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3708/comments

https://api.github.com/repos/huggingface/datasets/issues/3708/events

https://github.com/huggingface/datasets/issues/3708

I_kwDODunzps5Dh7nS

Loading JSON gets stuck with many workers/threads

open

MEMBER

## Describe the bug Loading a JSON dataset with `load_dataset` can get stuck when running on a machine with many CPUs. This is especially an issue when loading a large dataset on a large machine. ## Steps to reproduce the bug I originally created the following script to reproduce the issue: ```python from datasets import load_dataset from multiprocessing import Process from tqdm import tqdm import datasets from transformers import set_seed def run_tasks_in_parallel(tasks, ds_list): for _ in tqdm(range(1000)): print('new batch') running_tasks = [Process(target=task, args=(ds, i)) for i, (task, ds) in enumerate(zip(tasks, ds_list))] for running_task in running_tasks: running_task.start() for running_task in running_tasks: running_task.join() def get_dataset(): dataset_name = 'transformersbook/codeparrot' ds = load_dataset(dataset_name+'-train', split="train", streaming=True) ds = ds.shuffle(buffer_size=1000, seed=1) return iter(ds) def get_next_element(ds, process_id, N=10000): for _ in range(N): _ = next(ds)['content'] print(f'process {process_id} done') return set_seed(1) datasets.utils.logging.set_verbosity_debug() n_processes = 8 tasks = [get_next_element for _ in range(n_processes)] args = [get_dataset() for _ in range(n_processes)] run_tasks_in_parallel(tasks, args) ``` Today I noticed that it can happen when running it on a single process on a machine with many cores without streaming. So just `load_dataset("transformersbook/codeparrot-train")` alone might cause the issue after waiting long enough or trying many times. It's a slightly random process which makes it especially hard to track down. When I encountered it today it had already processed 17GB of data (the size of the cache folder when it got stuck) before getting stuck. Here's my current understanding of the error. As far as I can tell it happens in the following block: https://github.com/huggingface/datasets/blob/be701e9e89ab38022612c7263edc015bc7feaff9/src/datasets/packaged_modules/json/json.py#L119-L139 When the try on line 121 fails and the `block_size` is increased it can happen that it can't read the JSON again and gets stuck indefinitely. A hint that points in that direction is that increasing the `chunksize` argument decreases the chance of getting stuck and vice versa. Maybe it is an issue with a lock on the file that is not properly released. ## Expected results Read a JSON before the end of the universe. ## Actual results Read a JSON not before the end of the universe. ## Environment info <!-- You can run the command `datasets-cli env` and copy-and-paste its output below. --> - `datasets` version: 1.18.3 - Platform: Linux-4.19.0-18-cloud-amd64-x86_64-with-glibc2.28 - Python version: 3.9.10 - PyArrow version: 7.0.0 @lhoestq we dicsussed this a while ago. @albertvillanova we discussed this today :)

https://api.github.com/repos/huggingface/datasets/issues/3708/timeline

https://api.github.com/repos/huggingface/datasets/issues/3707

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3707/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3707/comments

https://api.github.com/repos/huggingface/datasets/issues/3707/events

https://github.com/huggingface/datasets/issues/3707

I_kwDODunzps5DhEUP

`.select`: unexpected behavior with `indices`

closed

NONE

## Describe the bug The `.select` method will not throw when sending `indices` bigger than the dataset length; `indices` will be wrapped instead. This behavior is not documented anywhere, and is not intuitive. ## Steps to reproduce the bug ```python from datasets import Dataset ds = Dataset.from_dict({"text": ["d", "e", "f"], "label": [4, 5, 6]}) res1 = ds.select([1, 2, 3])['text'] res2 = ds.select([1000])['text'] ``` ## Expected results Both results should throw an `Error`. ## Actual results `res1` will give `['e', 'f', 'd']` `res2` will give `['e']` ## Environment info Bug found from this environment: - `datasets` version: 1.16.1 - Platform: macOS-10.16-x86_64-i386-64bit - Python version: 3.8.7 - PyArrow version: 6.0.1 It was also replicated on `master`.

https://api.github.com/repos/huggingface/datasets/issues/3707/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3706

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3706/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3706/comments

https://api.github.com/repos/huggingface/datasets/issues/3706/events

https://github.com/huggingface/datasets/issues/3706

I_kwDODunzps5DfEn6

Unable to load dataset 'big_patent'

closed

NONE

## Describe the bug Unable to load the "big_patent" dataset ## Steps to reproduce the bug ```python load_dataset('big_patent', 'd', 'validation') ``` ## Expected results Download big_patents' validation split from the 'd' subset ## Getting an error saying: {FileNotFoundError}Local file ..\huggingface\datasets\downloads\6159313604f4f2c01e7d1cac52139343b6c07f73f6de348d09be6213478455c5\bigPatentData\train.tar.gz doesn't exist ## Environment info <!-- You can run the command `datasets-cli env` and copy-and-paste its output below. --> - `datasets` version:1.18.3 - Platform: Windows - Python version:3.8 - PyArrow version:7.0.0

https://api.github.com/repos/huggingface/datasets/issues/3706/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3705

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3705/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3705/comments

https://api.github.com/repos/huggingface/datasets/issues/3705/events

https://github.com/huggingface/datasets/pull/3705

PR_kwDODunzps4yfhyj

Raise informative error when loading a save_to_disk dataset

closed

MEMBER

People recurrently report error when trying to load a dataset (using `load_dataset`) that was previously saved using `save_to_disk`. This PR raises an informative error message telling them they should use `load_from_disk` instead. Close #3700.

https://api.github.com/repos/huggingface/datasets/issues/3705/timeline

https://api.github.com/repos/huggingface/datasets/issues/3704

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3704/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3704/comments

https://api.github.com/repos/huggingface/datasets/issues/3704/events

https://github.com/huggingface/datasets/issues/3704

I_kwDODunzps5DeZmH

OSCAR-2109 datasets are misaligned and truncated

closed

NONE

## Describe the bug The `oscar-corpus/OSCAR-2109` data appears to be misaligned and truncated by the dataset builder for subsets that contain more than one part and for cases where the texts contain non-unix newlines. ## Steps to reproduce the bug A few examples, although I'm not sure how deterministic the particular (mis)alignment is in various configurations: ```python from datasets import load_dataset dataset = load_dataset("oscar-corpus/OSCAR-2109", "deduplicated_fi", split="train", use_auth_token=True) entry = dataset[0] # entry["text"] is from fi_part_3.txt.gz # entry["meta"] is from fi_meta_part_2.jsonl.gz dataset = load_dataset("oscar-corpus/OSCAR-2109", "deduplicated_no", split="train", use_auth_token=True) entry = dataset[900000] # entry["text"] is from no_part_3.txt.gz and contains a blank line # entry["meta"] is from no_meta_part_1.jsonl.gz dataset = load_dataset("oscar-corpus/OSCAR-2109", "deduplicated_mk", split="train", streaming=True, use_auth_token=True) # 9088 texts in the dataset are empty ``` For `deduplicated_fi`, all exported raw texts from the dataset are 17GB rather than 20GB as reported in the data splits overview table. The token count with `wc -w` for the raw texts is 2,067,556,874 rather than the expected 2,357,264,196 from the data splits table. For `deduplicated_no` all exported raw texts contain 624,040,887 rather than the expected 776,354,517 tokens. For `deduplicated_mk` it is 122,236,936 rather than 134,544,934 tokens. I'm not expecting the `wc -w` counts to line up exactly with the data splits table, but for comparison the `wc -w` count for `deduplicated_mk` on the raw texts is 134,545,424. ## Issues * The meta / text files are not paired correctly when loading, so the extracted texts do not have the right offsets, the metadata is not associated with the correct text, and the text files may not be processed to the end or may be processed beyond the end (empty texts). * The line count offset is not reset per file so the texts aren't aligned to the right offsets in any parts beyond the first part, leading to truncation when in effect blank lines are not skipped. * Non-unix newline characters are treated as newlines when reading the text files while the metadata only counts unix newlines for its line offsets, leading to further misalignments between the metadata and the extracted texts, and which also results in truncation. ## Expected results All texts from the OSCAR release are extracted according to the metadata and aligned with the correct metadata. ## Fixes Not necessarily the exact fixes/checks you may want to use (I didn't test all languages or do any cross-platform testing, I'm not sure all the details are compatible with streaming), however to highlight the issues: ```diff diff --git a/OSCAR-2109.py b/OSCAR-2109.py index bbac1076..5eee8de7 100644 --- a/OSCAR-2109.py +++ b/OSCAR-2109.py @@ -20,6 +20,7 @@ import collections import gzip import json +import os import datasets @@ -387,9 +388,20 @@ class Oscar2109(datasets.GeneratorBasedBuilder): with open(checksum_file, encoding="utf-8") as f: data_filenames = [line.split()[1] for line in f if line] data_urls = [self.config.base_data_path + data_filename for data_filename in data_filenames] - text_files = dl_manager.download([url for url in data_urls if url.endswith(".txt.gz")]) - metadata_files = dl_manager.download([url for url in data_urls if url.endswith(".jsonl.gz")]) + # sort filenames so corresponding parts are aligned + text_files = sorted(dl_manager.download([url for url in data_urls if url.endswith(".txt.gz")])) + metadata_files = sorted(dl_manager.download([url for url in data_urls if url.endswith(".jsonl.gz")])) + assert len(text_files) == len(metadata_files) metadata_and_text_files = list(zip(metadata_files, text_files)) + for meta_path, text_path in metadata_and_text_files: + # check that meta/text part numbers are the same + if "part" in os.path.basename(text_path): + assert ( + os.path.basename(text_path).replace(".txt.gz", "").split("_")[-1] + == os.path.basename(meta_path).replace(".jsonl.gz", "").split("_")[-1] + ) + else: + assert len(metadata_and_text_files) == 1 return [ datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"metadata_and_text_files": metadata_and_text_files}), ] @@ -397,10 +409,14 @@ class Oscar2109(datasets.GeneratorBasedBuilder): def _generate_examples(self, metadata_and_text_files): """This function returns the examples in the raw (text) form by iterating on all the files.""" id_ = 0 - offset = 0 for meta_path, text_path in metadata_and_text_files: + # line offsets are per text file + offset = 0 logger.info("generating examples from = %s", text_path) - with gzip.open(open(text_path, "rb"), "rt", encoding="utf-8") as text_f: + # some texts contain non-Unix newlines that should not be + # interpreted as line breaks for the line counts in the metadata + # with readline() + with gzip.open(open(text_path, "rb"), "rt", encoding="utf-8", newline="\n") as text_f: with gzip.open(open(meta_path, "rb"), "rt", encoding="utf-8") as meta_f: for line in meta_f: # read meta @@ -411,7 +427,12 @@ class Oscar2109(datasets.GeneratorBasedBuilder): offset += 1 text_f.readline() # read text - text = "".join([text_f.readline() for _ in range(meta["nb_sentences"])]).rstrip() + text_lines = [text_f.readline() for _ in range(meta["nb_sentences"])] + # all lines contain text (no blank lines or EOF) + assert all(text_lines) + assert "\n" not in text_lines offset += meta["nb_sentences"] + # only strip the trailing newline + text = "".join(text_lines).rstrip("\n") yield id_, {"id": id_, "text": text, "meta": meta} id_ += 1 ``` I've tested this with a number of smaller deduplicated languages with 1-20 parts and the resulting datasets looked correct in terms of word count and size when compared to the data splits table and raw texts, and the text/metadata alignments were correct in all my spot checks. However, there are many many languages I didn't test and I'm not sure that there aren't any texts containing blank lines in the corpus, for instance. For the cases I tested, the assertions related to blank lines and EOF made it easier to verify that the text and metadata were aligned as intended, since there would be little chance of spurious alignments of variable-length texts across so much data.

https://api.github.com/repos/huggingface/datasets/issues/3704/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3703

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3703/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3703/comments

https://api.github.com/repos/huggingface/datasets/issues/3703/events

https://github.com/huggingface/datasets/issues/3703

I_kwDODunzps5DdykU

ImportError: To be able to use this metric, you need to install the following dependencies['seqeval'] using 'pip install seqeval' for instance'

open

NONE

hi : I want to use the seqeval indicator because of direct load_ When metric ('seqeval '), it will prompt that the network connection fails. So I downloaded the seqeval Py to load locally. Loading code: metric = load_ metric(path='mymetric/seqeval/seqeval.py') But tips: Traceback (most recent call last): File "/home/ubuntu/Python3.6_project/zyf_project/transformers/examples/pytorch/token-classification/run_ner.py", line 604, in <module> main() File "/home/ubuntu/Python3.6_project/zyf_project/transformers/examples/pytorch/token-classification/run_ner.py", line 481, in main metric = load_metric(path='mymetric/seqeval/seqeval.py') File "/home/ubuntu/Python3.6_project/zyf_project/transformers_venv_0209/lib/python3.7/site-packages/datasets/load.py", line 610, in load_metric dataset=False, File "/home/ubuntu/Python3.6_project/zyf_project/transformers_venv_0209/lib/python3.7/site-packages/datasets/load.py", line 450, in prepare_module f"To be able to use this {module_type}, you need to install the following dependencies" ImportError: To be able to use this metric, you need to install the following dependencies['seqeval'] using 'pip install seqeval' for instance' **What should I do? Please help me, thank you**

https://api.github.com/repos/huggingface/datasets/issues/3703/timeline

https://api.github.com/repos/huggingface/datasets/issues/3702

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3702/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3702/comments

https://api.github.com/repos/huggingface/datasets/issues/3702/events

https://github.com/huggingface/datasets/pull/3702

PR_kwDODunzps4yahKc

Update data URL of lm1b dataset

closed

NONE

The http address doesn't work anymore

https://api.github.com/repos/huggingface/datasets/issues/3702/timeline

https://api.github.com/repos/huggingface/datasets/issues/3701

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3701/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3701/comments

https://api.github.com/repos/huggingface/datasets/issues/3701/events

https://github.com/huggingface/datasets/pull/3701

PR_kwDODunzps4yZ8Dw

Pin ElasticSearch

closed

MEMBER

Until we manage to support ES 8.0, I'm setting the version to `<8.0.0` Currently we're getting this error on 8.0: ```python ValueError: Either 'hosts' or 'cloud_id' must be specified ``` When instantiating a `Elasticsearch()` object

https://api.github.com/repos/huggingface/datasets/issues/3701/timeline

https://api.github.com/repos/huggingface/datasets/issues/3700

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3700/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3700/comments

https://api.github.com/repos/huggingface/datasets/issues/3700/events

https://github.com/huggingface/datasets/issues/3700

I_kwDODunzps5DXkjQ

Unable to load a dataset

closed

NONE

## Describe the bug Unable to load a dataset from Huggingface that I have just saved. ## Steps to reproduce the bug On Google colab `! pip install datasets ` `from datasets import load_dataset` `my_path = "wiki_dataset"` `dataset = load_dataset('wikipedia', "20200501.fr")` `dataset.save_to_disk(my_path)` `dataset = load_dataset(my_path)` ## Expected results Loading the dataset ## Actual results ValueError: Couldn't cast _data_files: list<item: struct<filename: string>> child 0, item: struct<filename: string> child 0, filename: string _fingerprint: string _format_columns: null _format_kwargs: struct<> _format_type: null _indexes: struct<> _output_all_columns: bool _split: string to {'builder_name': Value(dtype='string', id=None), 'citation': Value(dtype='string', id=None), 'config_name': Value(dtype='string', id=None), 'dataset_size': Value(dtype='int64', id=None), 'description': Value(dtype='string', id=None), 'download_checksums': {}, 'download_size': Value(dtype='int64', id=None), 'features': {'title': {'dtype': Value(dtype='string', id=None), 'id': Value(dtype='null', id=None), '_type': Value(dtype='string', id=None)}, 'text': {'dtype': Value(dtype='string', id=None), 'id': Value(dtype='null', id=None), '_type': Value(dtype='string', id=None)}}, 'homepage': Value(dtype='string', id=None), 'license': Value(dtype='string', id=None), 'post_processed': Value(dtype='null', id=None), 'post_processing_size': Value(dtype='null', id=None), 'size_in_bytes': Value(dtype='int64', id=None), 'splits': {'train': {'name': Value(dtype='string', id=None), 'num_bytes': Value(dtype='int64', id=None), 'num_examples': Value(dtype='int64', id=None), 'dataset_name': Value(dtype='string', id=None)}}, 'supervised_keys': Value(dtype='null', id=None), 'task_templates': Value(dtype='null', id=None), 'version': {'version_str': Value(dtype='string', id=None), 'description': Value(dtype='string', id=None), 'major': Value(dtype='int64', id=None), 'minor': Value(dtype='int64', id=None), 'patch': Value(dtype='int64', id=None)}} because column names don't match ## Environment info - `datasets` version: 1.18.3 - Platform: Linux-5.4.144+-x86_64-with-Ubuntu-18.04-bionic - Python version: 3.7.12 - PyArrow version: 6.0.1

https://api.github.com/repos/huggingface/datasets/issues/3700/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3699

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3699/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3699/comments

https://api.github.com/repos/huggingface/datasets/issues/3699/events

https://github.com/huggingface/datasets/pull/3699

PR_kwDODunzps4yY49I

Add dev-only config to Natural Questions dataset

closed

MEMBER

As suggested by @lhoestq and @thomwolf, a new config has been added to Natural Questions dataset, so that only dev split can be downloaded. Fix #413.

https://api.github.com/repos/huggingface/datasets/issues/3699/timeline

https://api.github.com/repos/huggingface/datasets/issues/3698

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3698/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3698/comments

https://api.github.com/repos/huggingface/datasets/issues/3698/events

https://github.com/huggingface/datasets/pull/3698

PR_kwDODunzps4yXtyQ

Add finetune-data CodeFill

closed

NONE

https://api.github.com/repos/huggingface/datasets/issues/3698/timeline

https://api.github.com/repos/huggingface/datasets/issues/3697

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3697/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3697/comments

https://api.github.com/repos/huggingface/datasets/issues/3697/events

https://github.com/huggingface/datasets/pull/3697

PR_kwDODunzps4yXeXo

Add code-fill datasets for pretraining/finetuning/evaluating

closed

NONE

https://api.github.com/repos/huggingface/datasets/issues/3697/timeline

https://api.github.com/repos/huggingface/datasets/issues/3696

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3696/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3696/comments

https://api.github.com/repos/huggingface/datasets/issues/3696/events

https://github.com/huggingface/datasets/pull/3696

PR_kwDODunzps4yXXgH

Force unique keys in newsqa dataset

closed

MEMBER

Currently, it may raise `DuplicatedKeysError`. Fix #3630.

https://api.github.com/repos/huggingface/datasets/issues/3696/timeline

https://api.github.com/repos/huggingface/datasets/issues/3695

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3695/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3695/comments

https://api.github.com/repos/huggingface/datasets/issues/3695/events

https://github.com/huggingface/datasets/pull/3695

PR_kwDODunzps4yXP44

Fix ClassLabel to/from dict when passed names_file

closed

MEMBER

Currently, `names_file` is a field of the data class `ClassLabel`, thus appearing when transforming it to dict (when saving infos). Afterwards, when trying to read it from infos, it conflicts with the other field `names`. This PR, removes `names_file` as a field of the data class `ClassLabel`. - it is only used at instantiation to generate the `labels` field Fix #3631.

https://api.github.com/repos/huggingface/datasets/issues/3695/timeline

https://api.github.com/repos/huggingface/datasets/issues/3693

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3693/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3693/comments

https://api.github.com/repos/huggingface/datasets/issues/3693/events

https://github.com/huggingface/datasets/pull/3693

PR_kwDODunzps4yTTcQ

Standardize to `Example::`

closed

CONTRIBUTOR

https://api.github.com/repos/huggingface/datasets/issues/3693/timeline

https://api.github.com/repos/huggingface/datasets/issues/3692

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3692/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3692/comments

https://api.github.com/repos/huggingface/datasets/issues/3692/events

https://github.com/huggingface/datasets/pull/3692

PR_kwDODunzps4yShiu

Update data URL in pubmed dataset

closed

MEMBER

Fix #3655.

https://api.github.com/repos/huggingface/datasets/issues/3692/timeline

https://api.github.com/repos/huggingface/datasets/issues/3691

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3691/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3691/comments

https://api.github.com/repos/huggingface/datasets/issues/3691/events

https://github.com/huggingface/datasets/pull/3691

PR_kwDODunzps4yQThV

Upgrade black to version ~=22.0

closed

MEMBER

Upgrades the `datasets` library quality tool `black` to use the first stable release of `black`, version 22.0.

https://api.github.com/repos/huggingface/datasets/issues/3691/timeline

https://api.github.com/repos/huggingface/datasets/issues/3690

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3690/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3690/comments

https://api.github.com/repos/huggingface/datasets/issues/3690/events

https://github.com/huggingface/datasets/pull/3690

PR_kwDODunzps4yP2p5

Update docs to new frontend/UI

closed

CONTRIBUTOR

### TLDR: Update `datasets` `docs` to the new syntax (markdown and mdx files) & frontend (as how it looks on [hf.co/transformers](https://huggingface.co/docs/transformers/index)) | Light mode | Dark mode | |-----------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------| | <img width="400" alt="Screenshot 2022-02-17 at 14 15 34" src="https://user-images.githubusercontent.com/11827707/154489358-e2fb3708-8d72-4fb6-93f0-51d4880321c0.png"> | <img width="400" alt="Screenshot 2022-02-17 at 14 16 27" src="https://user-images.githubusercontent.com/11827707/154489596-c5a1311b-181c-4341-adb3-d60a7d3abe85.png"> | ## Checklist - [x] update datasets docs to new syntax (should call `doc-builder convert`) (this PR) - [x] discuss `@property` methods frontend https://github.com/huggingface/doc-builder/pull/87 - [x] discuss `inject_arrow_table_documentation` (this PR) https://github.com/huggingface/datasets/pull/3690#discussion_r801847860 - [x] update datasets docs path on moon-landing https://github.com/huggingface/moon-landing/pull/2089 - [x] convert pyarrow docstring from Numpydoc style to groups style https://github.com/huggingface/doc-builder/pull/89(https://stackoverflow.com/a/24385103/6558628) - [x] handle `Raises` section on frontend and doc-builder https://github.com/huggingface/doc-builder/pull/86 - [x] check imgs path (this PR) (nothing to update here) - [x] doc exaples block has to follow format `Examples::` https://github.com/huggingface/datasets/pull/3693 - [x] fix [this docstring](https://github.com/huggingface/datasets/blob/6ed6ac9448311930557810383d2cfd4fe6aae269/src/datasets/arrow_dataset.py#L3339) (causing svelte compilation error) - [x] Delete sphinx related files - [x] Delete sphinx CI - [x] Update docs config in setup.py - [x] add `versions.yml` in doc-build https://github.com/huggingface/doc-build/pull/1 - [x] add `versions.yml` in doc-build-dev https://github.com/huggingface/doc-build-dev/pull/1 - [x] https://github.com/huggingface/moon-landing/pull/2089 - [x] format docstrings for example `datasets.DatasetBuilder.download_and_prepare` args format look wrong - [x] create new github actions. (can probably be in a separate PR) (see the transformers equivalents below) 1. [build_dev_documentation.yml](https://github.com/huggingface/transformers/blob/master/.github/workflows/build_dev_documentation.yml) 2. [build_documentation.yml](https://github.com/huggingface/transformers/blob/master/.github/workflows/build_documentation.yml) 3. [delete_dev_documentation.yml](https://github.com/huggingface/transformers/blob/master/.github/workflows/delete_dev_documentation.yml) ## Note to reviewers The number of changed files is a lot (100+) because I've converted all `.rst` files to `.mdx` files & they are compiling fine on the svelte side (also, moved all the imgs to to [doc-imgs repo](https://huggingface.co/datasets/huggingface/documentation-images/tree/main/datasets)). Moreover, you should just review them on preprod and see if the rendering look fine. _Therefore, I'd suggest to focus on the changed_ **`.py`** and **CI files** (github workflows, etc. you can use [this filter here](https://github.com/huggingface/datasets/pull/3690/files?file-filters%5B%5D=.py&file-filters%5B%5D=.yml&show-deleted-files=true&show-viewed-files=true)) during the review & ignore `.mdx` files. (if there's a bug in `.mdx` files, we can always handle it in a separate PR afterwards).

https://api.github.com/repos/huggingface/datasets/issues/3690/timeline

https://api.github.com/repos/huggingface/datasets/issues/3689

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3689/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3689/comments

https://api.github.com/repos/huggingface/datasets/issues/3689/events

https://github.com/huggingface/datasets/pull/3689

PR_kwDODunzps4yPnp7

Fix streaming for servers not supporting HTTP range requests

closed

MEMBER

Some servers do not support HTTP range requests, whereas this is required to stream some file formats (like ZIP). ~~This PR implements a workaround for those cases, by download the files locally in a temporary directory (cleaned up by the OS once the process is finished).~~ This PR raises custom error explaining that streaming is not possible because data host server does not support HTTP range requests. Fix #3677.

https://api.github.com/repos/huggingface/datasets/issues/3689/timeline

https://api.github.com/repos/huggingface/datasets/issues/3688

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3688/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3688/comments

https://api.github.com/repos/huggingface/datasets/issues/3688/events

https://github.com/huggingface/datasets/issues/3688

I_kwDODunzps5DL_yR

Pyarrow version error

closed

NONE

## Describe the bug I installed datasets(version 1.17.0, 1.18.0, 1.18.3) but i'm right now nor able to import it because of pyarrow. when i try to import it, i get the following error: `To use datasets, the module pyarrow>=3.0.0 is required, and the current version of pyarrow doesn't match this condition`. i tryed with all version of pyarrow execpt `4.0.0` but still get the same error. ## Steps to reproduce the bug ```python import datasets ``` ## Expected results A clear and concise description of the expected results. ## Actual results AttributeError Traceback (most recent call last) <ipython-input-19-652e886d387f> in <module> ----> 1 import datasets ~\AppData\Local\Continuum\anaconda3\lib\site-packages\datasets\__init__.py in <module> 26 27 ---> 28 if _version.parse(pyarrow.__version__).major < 3: 29 raise ImportWarning( 30 "To use `datasets`, the module `pyarrow>=3.0.0` is required, and the current version of `pyarrow` doesn't match this condition.\n" AttributeError: 'Version' object has no attribute 'major' ## Environment info Traceback (most recent call last): File "c:\users\alex\appdata\local\continuum\anaconda3\lib\runpy.py", line 193, in _run_module_as_main "__main__", mod_spec) File "c:\users\alex\appdata\local\continuum\anaconda3\lib\runpy.py", line 85, in _run_code exec(code, run_globals) File "C:\Users\Alex\AppData\Local\Continuum\anaconda3\Scripts\datasets-cli.exe\__main__.py", line 5, in <module> File "c:\users\alex\appdata\local\continuum\anaconda3\lib\site-packages\datasets\__init__.py", line 28, in <module> if _version.parse(pyarrow.__version__).major < 3: AttributeError: 'Version' object has no attribute 'major' - `datasets` version: - Platform: Linux(Ubuntu) and Windows: conda on the both - Python version: 3.7 - PyArrow version: 7.0.0

https://api.github.com/repos/huggingface/datasets/issues/3688/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3687

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3687/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3687/comments

https://api.github.com/repos/huggingface/datasets/issues/3687/events

https://github.com/huggingface/datasets/issues/3687

I_kwDODunzps5DLwRO

Can't get the text data when calling to_tf_dataset

closed

NONE

I am working with the SST2 dataset, and am using TensorFlow 2.5 I'd like to convert it to a `tf.data.Dataset` by calling the `to_tf_dataset` method. The following snippet is what I am using to achieve this: ``` from datasets import load_dataset from transformers import DefaultDataCollator data_collator = DefaultDataCollator(return_tensors="tf") dataset = load_dataset("sst") train_dataset = dataset["train"].to_tf_dataset(columns=['sentence'], label_cols="label", shuffle=True, batch_size=8,collate_fn=data_collator) ``` However, this only gets me the labels; the text--the most important part--is missing: ``` for s in train_dataset.take(1): print(s) #prints something like: ({}, <tf.Tensor: shape=(8,), ...>) ``` As you can see, it only returns the label part, not the data, as indicated by the empty dictionary, `{}`. So far, I've played with various settings of the method arguments, but to no avail; I do not want to perform any text processing at this time. On my quest to achieve what I want ( a `tf.data.Dataset`), I've consulted these resources: [https://www.philschmid.de/huggingface-transformers-keras-tf](https://www.philschmid.de/huggingface-transformers-keras-tf) [https://huggingface.co/docs/datasets/use_dataset.html?highlight=tensorflow](https://huggingface.co/docs/datasets/use_dataset.html?highlight=tensorflow) I was surprised to not find more extensive examples on how to transform a Hugginface dataset to one compatible with TensorFlow. If you could point me to where I am going wrong, please do so. Thanks in advance for your support. --- Edit: In the [docs](https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.to_tf_dataset), I found the following description: _In general, only columns that the model can use as input should be included here (numeric data only)._ Does this imply that no textual, i.e., `string` data can be loaded?

https://api.github.com/repos/huggingface/datasets/issues/3687/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3686

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3686/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3686/comments

https://api.github.com/repos/huggingface/datasets/issues/3686/events

https://github.com/huggingface/datasets/issues/3686

I_kwDODunzps5DLsAK

`Translation` features cannot be `flatten`ed

closed

CONTRIBUTOR

## Describe the bug (`Dataset.flatten`)[https://github.com/huggingface/datasets/blob/master/src/datasets/arrow_dataset.py#L1265] fails for columns with feature (`Translation`)[https://github.com/huggingface/datasets/blob/3edbeb0ec6519b79f1119adc251a1a6b379a2c12/src/datasets/features/translation.py#L8] ## Steps to reproduce the bug ```python from datasets import load_dataset dataset = load_dataset("europa_ecdc_tm", "en2fr", split="train[:10]") print(dataset.features) # {'translation': Translation(languages=['en', 'fr'], id=None)} print(dataset[0]) # {'translation': {'en': 'Vaccination against hepatitis C is not yet available.', 'fr': 'Aucune vaccination contre l’hépatite C n’est encore disponible.'}} dataset.flatten() ``` ## Expected results `dataset.flatten` should flatten the `Translation` column as if it were a dict of `Value("string")` ```python dataset[0] # {'translation.en': 'Vaccination against hepatitis C is not yet available.', 'translation.fr': 'Aucune vaccination contre l’hépatite C n’est encore disponible.' } dataset.features # {'translation.en': Value("string"), 'translation.fr': Value("string")} ``` ## Actual results ```python In [31]: dset.flatten() --------------------------------------------------------------------------- KeyError Traceback (most recent call last) <ipython-input-31-bb88eb5276ee> in <module> ----> 1 dset.flatten() [...]\site-packages\datasets\fingerprint.py in wrapper(*args, **kwargs) 411 # Call actual function 412 --> 413 out = func(self, *args, **kwargs) 414 415 # Update fingerprint of in-place transforms + update in-place history of transforms [...]\site-packages\datasets\arrow_dataset.py in flatten(self, new_fingerprint, max_depth) 1294 break 1295 dataset.info.features = self.features.flatten(max_depth=max_depth) -> 1296 dataset._data = update_metadata_with_features(dataset._data, dataset.features) 1297 logger.info(f'Flattened dataset from depth {depth} to depth {1 if depth + 1 < max_depth else "unknown"}.') 1298 dataset._fingerprint = new_fingerprint [...]\site-packages\datasets\arrow_dataset.py in update_metadata_with_features(table, features) 534 def update_metadata_with_features(table: Table, features: Features): 535 """To be used in dataset transforms that modify the features of the dataset, in order to update the features stored in the metadata of its schema.""" --> 536 features = Features({col_name: features[col_name] for col_name in table.column_names}) 537 if table.schema.metadata is None or b"huggingface" not in table.schema.metadata: 538 pa_metadata = ArrowWriter._build_metadata(DatasetInfo(features=features)) [...]\site-packages\datasets\arrow_dataset.py in <dictcomp>(.0) 534 def update_metadata_with_features(table: Table, features: Features): 535 """To be used in dataset transforms that modify the features of the dataset, in order to update the features stored in the metadata of its schema.""" --> 536 features = Features({col_name: features[col_name] for col_name in table.column_names}) 537 if table.schema.metadata is None or b"huggingface" not in table.schema.metadata: 538 pa_metadata = ArrowWriter._build_metadata(DatasetInfo(features=features)) KeyError: 'translation.en' ``` ## Environment info - `datasets` version: 1.18.3 - Platform: Windows-10-10.0.19041-SP0 - Python version: 3.7.10 - PyArrow version: 3.0.0

https://api.github.com/repos/huggingface/datasets/issues/3686/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3685

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3685/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3685/comments

https://api.github.com/repos/huggingface/datasets/issues/3685/events

https://github.com/huggingface/datasets/pull/3685

PR_kwDODunzps4yLw3m

Add support for `Audio` and `Image` feature in `push_to_hub`

closed

CONTRIBUTOR

Add support for the `Audio` and the `Image` feature in `push_to_hub`. The idea is to remove local path information and store file content under "bytes" in the Arrow table before the push. My initial approach (https://github.com/huggingface/datasets/commit/34c652afeff9686b6b8bf4e703c84d2205d670aa) was to use a map transform similar to [`decode_nested_example`](https://github.com/huggingface/datasets/blob/5e0f6068741464f833ff1802e24ecc2064aaea9f/src/datasets/features/features.py#L1023-L1056) while having decoding turned off, but I wasn't satisfied with the code quality, so I ended up using the `temporary_assignment` decorator to override `cast_storage`, which allows me to directly modify the underlying storage (the final op is similar to `Dataset.cast`) and results in a much simpler code. Additionally, I added the `allow_cast` flag that can disable this behavior in the situations where it's not needed (e.g. the dataset is already in the correct format for the Hub, etc.) EDIT: `allow_cast` renamed to `embed_external_files`

https://api.github.com/repos/huggingface/datasets/issues/3685/timeline

https://api.github.com/repos/huggingface/datasets/issues/3684

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3684/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3684/comments

https://api.github.com/repos/huggingface/datasets/issues/3684/events

https://github.com/huggingface/datasets/pull/3684

PR_kwDODunzps4yIOer

[fix]: iwslt2017 download urls

closed

NONE

Fixes #2076.

https://api.github.com/repos/huggingface/datasets/issues/3684/timeline

https://api.github.com/repos/huggingface/datasets/issues/3683

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3683/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3683/comments

https://api.github.com/repos/huggingface/datasets/issues/3683/events

https://github.com/huggingface/datasets/pull/3683

PR_kwDODunzps4yGKoj

added told-br (brazilian hate speech) dataset

closed

CONTRIBUTOR

Hey, Adding ToLD-Br. Feel free to ask for modifications. Thanks!!

https://api.github.com/repos/huggingface/datasets/issues/3683/timeline

https://api.github.com/repos/huggingface/datasets/issues/3682

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3682/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3682/comments

https://api.github.com/repos/huggingface/datasets/issues/3682/events

https://github.com/huggingface/datasets/pull/3682

PR_kwDODunzps4yGFml

adding told-br for toxic/abusive hatespeech detection

closed

CONTRIBUTOR

Hey, I'm adding our dataset from our paper published at AACL 2020. Feel free to ask for modifications. Thanks!

https://api.github.com/repos/huggingface/datasets/issues/3682/timeline

https://api.github.com/repos/huggingface/datasets/issues/3681

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3681/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3681/comments

https://api.github.com/repos/huggingface/datasets/issues/3681/events

https://github.com/huggingface/datasets/pull/3681

PR_kwDODunzps4yFcpM

Fix TestCommand to move dataset_infos instead of copying

open

MEMBER

Why do we copy instead of moving the file? CC: @lhoestq @lvwerra

https://api.github.com/repos/huggingface/datasets/issues/3681/timeline

https://api.github.com/repos/huggingface/datasets/issues/3680

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3680/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3680/comments

https://api.github.com/repos/huggingface/datasets/issues/3680/events

https://github.com/huggingface/datasets/pull/3680

PR_kwDODunzps4yFXm8

Fix TestCommand to copy dataset_infos to local dir with only data files

closed

MEMBER

Currently this case is missed. CC: @lvwerra

https://api.github.com/repos/huggingface/datasets/issues/3680/timeline

https://api.github.com/repos/huggingface/datasets/issues/3679

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3679/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3679/comments

https://api.github.com/repos/huggingface/datasets/issues/3679/events

https://github.com/huggingface/datasets/issues/3679

I_kwDODunzps5C_9O1

Download datasets from a private hub

closed

NONE

In the context of a private hub deployment, customers would like to use load_dataset() to load datasets from their hub, not from the public hub. This doesn't seem to be configurable at the moment and it would be nice to add this feature. The obvious workaround is to clone the repo first and then load it from local storage, but this adds an extra step. It'd be great to have the same experience regardless of where the hub is hosted. The same issue exists with the transformers library and the CLI. I'm going to create issues there as well, and I'll reference them below.

https://api.github.com/repos/huggingface/datasets/issues/3679/timeline

completed

https://api.github.com/repos/huggingface/datasets/issues/3678

https://api.github.com/repos/huggingface/datasets

https://api.github.com/repos/huggingface/datasets/issues/3678/labels{/name}

https://api.github.com/repos/huggingface/datasets/issues/3678/comments

https://api.github.com/repos/huggingface/datasets/issues/3678/events

https://github.com/huggingface/datasets/pull/3678

PR_kwDODunzps4yCt91

Add code example in wikipedia card

closed