title stringlengths 1 290 | body stringlengths 0 228k ⌀ | html_url stringlengths 46 51 | comments list | pull_request dict | number int64 1 5.59k | is_pull_request bool 2 classes |
|---|---|---|---|---|---|---|
HF_DATASETS_OFFLINE=1 didn't stop datasets.builder from downloading | ## Describe the bug
According to https://huggingface.co/docs/datasets/loading_datasets.html#loading-a-dataset-builder, setting HF_DATASETS_OFFLINE to 1 should make datasets to "run in full offline mode". It didn't work for me. At the very beginning, datasets still tried to download "custom data configuration" for JSON, despite I have run the program once and cached all data into the same --cache_dir.
"Downloading" is not an issue when running with local disk, but crashes often with cloud storage because (1) multiply GPU processes try to access the same file, AND (2) FileLocker fails to synchronize all processes, due to storage throttling. 99% of times, when the main process releases FileLocker, the file is not actually ready for access in cloud storage and thus triggers "FileNotFound" errors for all other processes. Well, another way to resolve the problem is to investigate super reliable cloud storage, but that's out of scope here.
## Steps to reproduce the bug
```
export HF_DATASETS_OFFLINE=1
python run_clm.py --model_name_or_path=models/gpt-j-6B --train_file=trainpy.v2.train.json --validation_file=trainpy.v2.eval.json --cache_dir=datacache/trainpy.v2
```
## Expected results
datasets should stop all "downloading" behavior but reuse the cached JSON configuration. I think the problem here is part of the cache directory path, "default-471372bed4b51b53", is randomly generated, and it could change if some parameters changed. And I didn't find a way to use a fixed path to ensure datasets to reuse cached data every time.
## Actual results
The logging shows datasets are still downloading into "datacache/trainpy.v2/json/default-471372bed4b51b53/0.0.0/c2d554c3377ea79c7664b93dc65d0803b45e3279000f993c7bfd18937fd7f426".
```
12/16/2021 10:25:59 - WARNING - datasets.builder - Using custom data configuration default-471372bed4b51b53
12/16/2021 10:25:59 - INFO - datasets.builder - Generating dataset json (datacache/trainpy.v2/json/default-471372bed4b51b53/0.0.0/c2d554c3377ea79c7664b93dc65d0803b45e3279000f993c7bfd18937fd7f426)
Downloading and preparing dataset json/default to datacache/trainpy.v2/json/default-471372bed4b51b53/0.0.0/c2d554c3377ea79c7664b93dc65d0803b45e3279000f993c7bfd18937fd7f426...
100%|██████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 17623.13it/s]
12/16/2021 10:25:59 - INFO - datasets.utils.download_manager - Downloading took 0.0 min
12/16/2021 10:26:00 - INFO - datasets.utils.download_manager - Checksum Computation took 0.0 min
100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 1206.99it/s]
12/16/2021 10:26:00 - INFO - datasets.utils.info_utils - Unable to verify checksums.
12/16/2021 10:26:00 - INFO - datasets.builder - Generating split train
12/16/2021 10:26:01 - INFO - datasets.builder - Generating split validation
12/16/2021 10:26:02 - INFO - datasets.utils.info_utils - Unable to verify splits sizes.
Dataset json downloaded and prepared to datacache/trainpy.v2/json/default-471372bed4b51b53/0.0.0/c2d554c3377ea79c7664b93dc65d0803b45e3279000f993c7bfd18937fd7f426. Subsequent calls will reuse this data.
100%|█████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 53.54it/s]
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.16.1
- Platform: Linux
- Python version: 3.8.10
- PyArrow version: 6.0.1
| https://github.com/huggingface/datasets/issues/3447 | [
"Hi ! Indeed it says \"downloading and preparing\" but in your case it didn't need to download anything since you used local files (it would have thrown an error otherwise). I think we can improve the logging to make it clearer in this case",
"@lhoestq Thank you for explaining. I am sorry but I was not clear abou... | null | 3,447 | false |
Remove redundant local path information in audio/image datasets | Remove the redundant path information in the audio/image dataset as discussed in https://github.com/huggingface/datasets/pull/3430#issuecomment-994734828
TODOs:
* [ ] merge https://github.com/huggingface/datasets/pull/3430
* [ ] merge https://github.com/huggingface/datasets/pull/3364
* [ ] re-generate the info files of the updated audio datasets
cc: @patrickvonplaten @anton-l @nateraw (I expect this to break the audio/vision examples in Transformers; after this change you'll be able to access underlying paths as follows `dset = dset.cast_column("audio", Audio(..., decode=False)); path = dset[0]["audio"]`)
| https://github.com/huggingface/datasets/pull/3446 | [
"Cool, I'm in favor of this PR. Our official examples in speech already make use of `\"audio\"` so no need to change anything there. It would be great if we could prominently feature how one can get the audio path without decoding in the docs.",
"@patrickvonplaten Yes, I agree.\r\n\r\ncc @stevhliu we should add a... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3446",
"html_url": "https://github.com/huggingface/datasets/pull/3446",
"diff_url": "https://github.com/huggingface/datasets/pull/3446.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3446.patch",
"merged_at": null
} | 3,446 | true |
question | ## Dataset viewer issue for '*name of the dataset*'
**Link:** *link to the dataset viewer page*
*short description of the issue*
Am I the one who added this dataset ? Yes-No
| https://github.com/huggingface/datasets/issues/3445 | [
"Hi ! What's your question ?"
] | null | 3,445 | false |
Align the Dataset and IterableDataset processing API | ## Intro
items marked like <s>this</s> are done already :)
Currently the two classes have two distinct API for processing:
### The `.map()` method
Both have those parameters in common: function, batched, batch_size
- IterableDataset is missing those parameters:
<s>with_indices</s>, with_rank, <s>input_columns</s>, <s>drop_last_batch</s>, <s>remove_columns</s>, features, disable_nullable, fn_kwargs, num_proc
- Dataset also has additional parameters that are exclusive, due to caching:
keep_in_memory, load_from_cache_file, cache_file_name, writer_batch_size, suffix_template, new_fingerprint
- <s>There is also an important difference in terms of behavior:
**Dataset.map adds new columns** (with dict.update)
BUT
**IterableDataset discards previous columns** (it overwrites the dict)
IMO the two methods should have the same behavior. This would be an important breaking change though.</s>
- Dataset.map is eager while IterableDataset.map is lazy
### The `.shuffle()` method
- <s>Both have an optional seed parameter, but IterableDataset requires a mandatory parameter buffer_size to control the size of the local buffer used for approximate shuffling.</s>
- <s>IterableDataset is missing the parameter generator</s>
- Also Dataset has exclusive parameters due to caching: keep_in_memory, load_from_cache_file, indices_cache_file_name, writer_batch_size, new_fingerprint
### The `.with_format()` method
- IterableDataset only supports "torch" (it misses tf, jax, pandas, arrow) and is missing the parameters: columns, output_all_columns and format_kwargs
- other methods like `set_format`, `reset_format` or `formatted_as` are also missing
### Other methods
- Both have the same `remove_columns` method
- IterableDataset is missing: <s>cast</s>, <s>cast_column</s>, <s>filter</s>, <s>rename_column</s>, <s>rename_columns</s>, class_encode_column, flatten, prepare_for_task, train_test_split, shard
- Some other methods are missing but we can discuss them: set_transform, formatted_as, with_transform
- And others don't really make sense for an iterable dataset: select, sort, add_column, add_item
- Dataset is missing skip and take, that IterableDataset implements.
## Questions
I think it would be nice to be able to switch between streaming and regular dataset easily, without changing the processing code significantly.
1. What should be aligned and what shouldn't between those two APIs ?
IMO the minimum is to align the main processing methods.
It would mean aligning breaking the current `Iterable.map` to have the same behavior as `Dataset.map` (add columns with dict.update), and add multiprocessing as well as the missing parameters. DONE ✅
It would also mean implementing the missing methods: cast, cast_column, filter, rename_column, rename_columns, class_encode_column, flatten, prepare_for_task, train_test_split, shard. WIP 🟠
2. What are the breaking changes for IterableDataset ?
The main breaking change would be the change of behavior of `IterableDataset.map`, because currently it discards all the previous columns instead of keeping them. DONE ✅
3. Shall we also do some changes for regular datasets ?
I agree the simplest would be to have the exact same methods for both Dataset and IterableDataset. However this is probably not a good idea because it would prevent users from using the best benefits of them. That's why we can keep some aspects of regular datasets as they are:
- keep the eager Dataset.map with caching
- keep the with_transform method for lazy processing
- keep Dataset.select (it could also be added to IterableDataset even though it's not recommended)
We could have a completely aligned `map` method if both methods were lazy by default, but this is a very big breaking change so I'm not sure we can consider doing that.
For information, TFDS does lazy map by default, and has an additional `.cache()` method.
## Opinions ?
I'd love to gather some opinions about this here. If the two APIs are more aligned it would be awesome for the examples in `transformers`, and it would create a satisfactory experience for users that want to switch from one mode to the other.
cc @mariosasko @albertvillanova @thomwolf @patrickvonplaten @sgugger | https://github.com/huggingface/datasets/issues/3444 | [
"Yes I agree, these should be as aligned as possible. Maybe we can also check the feedback in the survey at http://hf.co/oss-survey and see if people mentioned related things on the API (in particular if we go the breaking change way, it would be good to be sure we are taking the right direction for the community).... | null | 3,444 | false |
Extend iter_archive to support file object input | This PR adds support to passing a file object to `[Streaming]DownloadManager.iter_archive`.
With this feature, we can iterate over a tar file inside another tar file. | https://github.com/huggingface/datasets/pull/3443 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3443",
"html_url": "https://github.com/huggingface/datasets/pull/3443",
"diff_url": "https://github.com/huggingface/datasets/pull/3443.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3443.patch",
"merged_at": "2021-12-17T17:53:02"
} | 3,443 | true |
Extend text to support yielding lines, paragraphs or documents | Add `config.row` option to `text` module to allow yielding lines (default, current case), paragraphs or documents.
Feel free to comment on the name of the config parameter `row`:
- Currently, the docs state datasets are made of rows and columns
- Other names I considered: `example`, `item` | https://github.com/huggingface/datasets/pull/3442 | [
"The parameter can also be named `split_by` with values \"line\", \"paragraph\" or \"document\" (no 's' at the end)",
"> The parameter can also be named `split_by` with values \"line\", \"paragraph\" or \"document\" (no 's' at the end)\r\n\r\n@lhoestq @mariosasko I would avoid the term `split` in this context and... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3442",
"html_url": "https://github.com/huggingface/datasets/pull/3442",
"diff_url": "https://github.com/huggingface/datasets/pull/3442.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3442.patch",
"merged_at": "2021-12-20T16:39:18"
} | 3,442 | true |
Add QuALITY dataset | ## Adding a Dataset
- **Name:** QuALITY
- **Description:** A challenging question answering with very long contexts (Twitter [thread](https://twitter.com/sleepinyourhat/status/1471225421794529281?s=20))
- **Paper:** No ArXiv link yet, but draft is [here](https://github.com/nyu-mll/quality/blob/main/quality_preprint.pdf)
- **Data:** GitHub repo [here](https://github.com/nyu-mll/quality)
- **Motivation:** This dataset would serve as a nice way to benchmark long-range Transformer models like BigBird, Longformer and their descendants. In particular, it would be very interesting to see how the S4 model fares on this given it's impressive performance on the Long Range Arena
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| https://github.com/huggingface/datasets/issues/3441 | [
"I'll take this one if no one hasn't yet!"
] | null | 3,441 | false |
datasets keeps reading from cached files, although I disabled it | ## Describe the bug
Hi,
I am trying to avoid dataset library using cached files, I get the following bug when this tried to read the cached files. I tried to do the followings:
```
from datasets import set_caching_enabled
set_caching_enabled(False)
```
also force redownlaod:
```
download_mode='force_redownload'
```
but none worked so far, this is on a cluster and on some of the machines this reads from the cached files, I really appreciate any idea on how to fully remove caching @lhoestq
many thanks
```
File "run_clm.py", line 496, in <module>
main()
File "run_clm.py", line 419, in main
train_result = trainer.train(resume_from_checkpoint=checkpoint)
File "/users/dara/codes/fewshot/debug/fewshot/third_party/trainers/trainer.py", line 943, in train
self._maybe_log_save_evaluate(tr_loss, model, trial, epoch, ignore_keys_for_eval)
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/transformers/trainer.py", line 1445, in _maybe_log_save_evaluate
metrics = self.evaluate(ignore_keys=ignore_keys_for_eval)
File "/users/dara/codes/fewshot/debug/fewshot/third_party/trainers/trainer.py", line 172, in evaluate
output = self.eval_loop(
File "/users/dara/codes/fewshot/debug/fewshot/third_party/trainers/trainer.py", line 241, in eval_loop
metrics = self.compute_pet_metrics(eval_datasets, model, self.extra_info[metric_key_prefix], task=task)
File "/users/dara/codes/fewshot/debug/fewshot/third_party/trainers/trainer.py", line 268, in compute_pet_metrics
centroids = self._compute_per_token_train_centroids(model, task=task)
File "/users/dara/codes/fewshot/debug/fewshot/third_party/trainers/trainer.py", line 353, in _compute_per_token_train_centroids
data = get_label_samples(self.get_per_task_train_dataset(task), label)
File "/users/dara/codes/fewshot/debug/fewshot/third_party/trainers/trainer.py", line 350, in get_label_samples
return dataset.filter(lambda example: int(example['labels']) == label)
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 470, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/datasets/fingerprint.py", line 406, in wrapper
out = func(self, *args, **kwargs)
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 2519, in filter
indices = self.map(
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 2036, in map
return self._map_single(
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 503, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 470, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/datasets/fingerprint.py", line 406, in wrapper
out = func(self, *args, **kwargs)
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 2248, in _map_single
return Dataset.from_file(cache_file_name, info=info, split=self.split)
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 654, in from_file
return cls(
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 593, in __init__
self.info.features = self.info.features.reorder_fields_as(inferred_features)
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/datasets/features/features.py", line 1092, in reorder_fields_as
return Features(recursive_reorder(self, other))
File "/users/dara/conda/envs/multisuccess/lib/python3.8/site-packages/datasets/features/features.py", line 1081, in recursive_reorder
raise ValueError(f"Keys mismatch: between {source} and {target}" + stack_position)
ValueError: Keys mismatch: between {'indices': Value(dtype='uint64', id=None)} and {'candidates_ids': Sequence(feature=Value(dtype='null', id=None), length=-1, id=None), 'labels': Value(dtype='int64', id=None), 'attention_mask': Sequence(feature=Value(dtype='int8', id=None), length=-1, id=None), 'input_ids': Sequence(feature=Value(dtype='int32', id=None), length=-1, id=None), 'extra_fields': {}, 'task': Value(dtype='string', id=None)}
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version:
- Platform: linux
- Python version: 3.8.12
- PyArrow version: 6.0.1
| https://github.com/huggingface/datasets/issues/3440 | [
"Hi ! What version of `datasets` are you using ? Can you also provide the logs you get before it raises the error ?"
] | null | 3,440 | false |
Add `cast_column` to `IterableDataset` | Closes #3369.
cc: @patrickvonplaten | https://github.com/huggingface/datasets/pull/3439 | [
"Awesome thanks a lot @mariosasko "
] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3439",
"html_url": "https://github.com/huggingface/datasets/pull/3439",
"diff_url": "https://github.com/huggingface/datasets/pull/3439.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3439.patch",
"merged_at": "2021-12-16T15:55:19"
} | 3,439 | true |
Update supported versions of Python in setup.py | Update the list of supported versions of Python in `setup.py` to keep the PyPI project description updated. | https://github.com/huggingface/datasets/pull/3438 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3438",
"html_url": "https://github.com/huggingface/datasets/pull/3438",
"diff_url": "https://github.com/huggingface/datasets/pull/3438.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3438.patch",
"merged_at": "2021-12-20T14:22:12"
} | 3,438 | true |
Update BLEURT hyperlink | The description of BLEURT on the hf.co website has a strange use of URL hyperlinking. This PR attempts to fix this, although I am not 100% sure Markdown syntax is allowed on the frontend or not.
| https://github.com/huggingface/datasets/pull/3437 | [
"seems like a very very low-prio improvement :)",
"@albertvillanova thanks for the feedback! I removed the formatting altogether since I think this is a bit simpler tor read than non-rendered Markdown"
] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3437",
"html_url": "https://github.com/huggingface/datasets/pull/3437",
"diff_url": "https://github.com/huggingface/datasets/pull/3437.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3437.patch",
"merged_at": "2021-12-17T13:28:25"
} | 3,437 | true |
Add the OneStopQa dataset | Adding OneStopQA, a multiple choice reading comprehension dataset annotated according to the STARC (Structured Annotations for Reading Comprehension) scheme. | https://github.com/huggingface/datasets/pull/3436 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3436",
"html_url": "https://github.com/huggingface/datasets/pull/3436",
"diff_url": "https://github.com/huggingface/datasets/pull/3436.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3436.patch",
"merged_at": "2021-12-17T13:25:29"
} | 3,436 | true |
Improve Wikipedia Loading Script | * More structured approach to detecting redirects
* Remove redundant template filter code (covered by strip_code)
* Add language-specific lists of additional media namespace aliases for filtering
* Add language-specific lists of category namespace aliases for new link text cleaning step
* Remove magic words (parser directions like __TOC__ that occasionally occur in text)
Fix #3400
With support from @albertvillanova
CC @yjernite | https://github.com/huggingface/datasets/pull/3435 | [
"I wanted to flag a change from since we discussed this: I initially wrote a function for using the Wikimedia APIs to collect namespace aliases, but decided that adding in more http requests to the script wasn't a great idea so instead used that code to build a static list that I just added directly to the code.\r\... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3435",
"html_url": "https://github.com/huggingface/datasets/pull/3435",
"diff_url": "https://github.com/huggingface/datasets/pull/3435.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3435.patch",
"merged_at": "2022-03-04T08:16:00"
} | 3,435 | true |
Add The People's Speech | ## Adding a Dataset
- **Name:** The People's Speech
- **Description:** a massive English-language dataset of audio transcriptions of full sentences.
- **Paper:** https://openreview.net/pdf?id=R8CwidgJ0yT
- **Data:** https://mlcommons.org/en/peoples-speech/
- **Motivation:** With over 30,000 hours of speech, this dataset is the largest and most diverse freely available English speech recognition corpus today.
[The article](https://thegradient.pub/new-datasets-to-democratize-speech-recognition-technology-2/) which may be useful when working on the dataset.
cc: @anton-l
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| https://github.com/huggingface/datasets/issues/3434 | [
"This dataset is now available on the Hub here: https://huggingface.co/datasets/MLCommons/peoples_speech"
] | null | 3,434 | false |
Add Multilingual Spoken Words dataset | ## Adding a Dataset
- **Name:** Multilingual Spoken Words
- **Description:** Multilingual Spoken Words Corpus is a large and growing audio dataset of spoken words in 50 languages for academic research and commercial applications in keyword spotting and spoken term search, licensed under CC-BY 4.0. The dataset contains more than 340,000 keywords, totaling 23.4 million 1-second spoken examples (over 6,000 hours).
Read more: https://mlcommons.org/en/news/spoken-words-blog/
- **Paper:** https://datasets-benchmarks-proceedings.neurips.cc/paper/2021/file/fe131d7f5a6b38b23cc967316c13dae2-Paper-round2.pdf
- **Data:** https://mlcommons.org/en/multilingual-spoken-words/
- **Motivation:**
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| https://github.com/huggingface/datasets/issues/3433 | [] | null | 3,433 | false |
Correctly indent builder config in dataset script docs | null | https://github.com/huggingface/datasets/pull/3432 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3432",
"html_url": "https://github.com/huggingface/datasets/pull/3432",
"diff_url": "https://github.com/huggingface/datasets/pull/3432.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3432.patch",
"merged_at": "2021-12-14T17:35:17"
} | 3,432 | true |
Unable to resolve any data file after loading once | when I rerun my program, it occurs this error
" Unable to resolve any data file that matches '['**train*']' at /data2/whr/lzy/open_domain_data/retrieval/wiki_dpr with any supported extension ['csv', 'tsv', 'json', 'jsonl', 'parquet', 'txt', 'zip']", so how could i deal with this problem?
thx.
And below is my code .
| https://github.com/huggingface/datasets/issues/3431 | [
"Hi ! `load_dataset` accepts as input either a local dataset directory or a dataset name from the Hugging Face Hub.\r\n\r\nSo here you are getting this error the second time because it tries to load the local `wiki_dpr` directory, instead of `wiki_dpr` from the Hub. It doesn't work since it's a **cache** directory,... | null | 3,431 | false |
Make decoding of Audio and Image feature optional | Add the `decode` argument (`True` by default) to the `Audio` and the `Image` feature to make it possible to toggle on/off decoding of these features.
Even though we've discussed that on Slack, I'm not removing the `_storage_dtype` argument of the Audio feature in this PR to avoid breaking the Audio feature tests. | https://github.com/huggingface/datasets/pull/3430 | [
"Closing this PR for now due to https://github.com/huggingface/datasets/issues/3145#issuecomment-993664104.",
"Okay, after some more thinking, I'm re-opening this PR for three reasons:\r\n* This feature will allow us to remove the `image_file_path`/`audio_file_path` columns in our vision/audio datasets. Currently... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3430",
"html_url": "https://github.com/huggingface/datasets/pull/3430",
"diff_url": "https://github.com/huggingface/datasets/pull/3430.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3430.patch",
"merged_at": "2022-01-25T18:57:52"
} | 3,430 | true |
Make cast cacheable (again) on Windows | `cast` currently emits the following warning when called on Windows:
```
Parameter 'function'=<function Dataset.cast.<locals>.<lambda> at 0x000001C930571EA0> of the transform datasets.arrow_dataset.Dataset._map_single couldn't be hashed properly, a random hash was used instead. Make sure your transforms and parameters are serializable with pickle or dill for the dataset fingerprinting
and caching to work. If you reuse this transform, the caching mechanism will consider it to be different
from the previous calls and recompute everything. This warning is only showed once. Subsequent hashing failures won't be showed.
```
It seems like the issue stems from the `config.PYARROW_VERSION` object not being serializable on Windows (tested with `dumps(lambda: config.PYARROW_VERSION)`), so I'm fixing this by capturing `config.PYARROW_VERSION.major` before the lambda definition. | https://github.com/huggingface/datasets/pull/3429 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3429",
"html_url": "https://github.com/huggingface/datasets/pull/3429",
"diff_url": "https://github.com/huggingface/datasets/pull/3429.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3429.patch",
"merged_at": "2021-12-14T14:39:50"
} | 3,429 | true |
Clean squad dummy data | Some unused files were remaining, this PR removes them. We just need to keep the dummy_data.zip file | https://github.com/huggingface/datasets/pull/3428 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3428",
"html_url": "https://github.com/huggingface/datasets/pull/3428",
"diff_url": "https://github.com/huggingface/datasets/pull/3428.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3428.patch",
"merged_at": "2021-12-13T18:57:50"
} | 3,428 | true |
Add The Pile Enron Emails subset | Add:
- Enron Emails subset of The Pile: "enron_emails" config
Close bigscience-workshop/data_tooling#310.
CC: @StellaAthena | https://github.com/huggingface/datasets/pull/3427 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3427",
"html_url": "https://github.com/huggingface/datasets/pull/3427",
"diff_url": "https://github.com/huggingface/datasets/pull/3427.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3427.patch",
"merged_at": "2021-12-14T17:30:55"
} | 3,427 | true |
Update disaster_response_messages download urls (+ add validation split) | Fixes #3240, fixes #3416 | https://github.com/huggingface/datasets/pull/3426 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3426",
"html_url": "https://github.com/huggingface/datasets/pull/3426",
"diff_url": "https://github.com/huggingface/datasets/pull/3426.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3426.patch",
"merged_at": "2021-12-14T14:38:29"
} | 3,426 | true |
Getting configs names takes too long |
## Steps to reproduce the bug
```python
from datasets import get_dataset_config_names
get_dataset_config_names("allenai/c4")
```
## Expected results
I would expect to get the answer quickly, at least in less than 10s
## Actual results
It takes about 45s on my environment
## Environment info
- `datasets` version: 1.16.1
- Platform: Linux-5.11.0-1022-aws-x86_64-with-glibc2.31
- Python version: 3.9.6
- PyArrow version: 4.0.1 | https://github.com/huggingface/datasets/issues/3425 | [
"maybe related to https://github.com/huggingface/datasets/issues/2859\r\n",
"It looks like it's currently calling `HfFileSystem.ls()` ~8 times at the root and for each subdirectory:\r\n- \"\"\r\n- \"en.noblocklist\"\r\n- \"en.noclean\"\r\n- \"en\"\r\n- \"multilingual\"\r\n- \"realnewslike\"\r\n\r\nCurrently `ls` ... | null | 3,425 | false |
Add RedCaps dataset | Add the RedCaps dataset. I'm not adding the generated `dataset_infos.json` file for now due to its size (11 MB).
TODOs:
- [x] dummy data
- [x] dataset card
Close #3316 | https://github.com/huggingface/datasets/pull/3424 | [
"Cool ! If you want you can include `dataset_infos.json` but only for the main configurations. That's what we do for example for translation datasets when there are too many configs",
"@lhoestq I've added an example that uses `map` to download the images."
] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3424",
"html_url": "https://github.com/huggingface/datasets/pull/3424",
"diff_url": "https://github.com/huggingface/datasets/pull/3424.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3424.patch",
"merged_at": "2022-01-12T14:13:15"
} | 3,424 | true |
data duplicate when setting num_works > 1 with streaming data | ## Describe the bug
The data is repeated num_works times when we load_dataset with streaming and set num_works > 1 when construct dataloader
## Steps to reproduce the bug
```python
# Sample code to reproduce the bug
import pandas as pd
import numpy as np
import os
from datasets import load_dataset
from torch.utils.data import DataLoader
from tqdm import tqdm
import shutil
NUM_OF_USER = 1000000
NUM_OF_ACTION = 50000
NUM_OF_SEQUENCE = 10000
NUM_OF_FILES = 32
NUM_OF_WORKERS = 16
if __name__ == "__main__":
shutil.rmtree("./dataset")
for i in range(NUM_OF_FILES):
sequence_data = pd.DataFrame(
{
"imei": np.random.randint(1, NUM_OF_USER, size=NUM_OF_SEQUENCE),
"sequence": np.random.randint(1, NUM_OF_ACTION, size=NUM_OF_SEQUENCE)
}
)
if not os.path.exists("./dataset"):
os.makedirs("./dataset")
sequence_data.to_csv(f"./dataset/sequence_data_{i}.csv",
index=False)
dataset = load_dataset("csv",
data_files=[os.path.join("./dataset",file) for file in os.listdir("./dataset") if file.endswith(".csv")],
split="train",
streaming=True).with_format("torch")
data_loader = DataLoader(dataset,
batch_size=1024,
num_workers=NUM_OF_WORKERS)
result = pd.DataFrame()
for i, batch in tqdm(enumerate(data_loader)):
result = pd.concat([result,
pd.DataFrame(batch)],
axis=0)
result.to_csv(f"num_work_{NUM_OF_WORKERS}.csv", index=False)
```
## Expected results
data do not duplicate
## Actual results
data duplicate NUM_OF_WORKERS = 16
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version:datasets==1.14.0
- Platform:transformers==4.11.3
- Python version:3.8
- PyArrow version:
| https://github.com/huggingface/datasets/issues/3423 | [
"Hi ! Thanks for reporting :)\r\n\r\nWhen using a PyTorch's data loader with `num_workers>1` and an iterable dataset, each worker streams the exact same data by default, resulting in duplicate data when iterating using the data loader.\r\n\r\nWe can probably fix this in `datasets` by checking `torch.utils.data.get_... | null | 3,423 | false |
Error about load_metric | ## Describe the bug
File "/opt/conda/lib/python3.8/site-packages/datasets/load.py", line 1371, in load_metric
metric = metric_cls(
TypeError: 'NoneType' object is not callable
## Steps to reproduce the bug
```python
metric = load_metric("glue", "sst2")
```
## Environment info
- `datasets` version: 1.16.1
- Platform: Linux-4.15.0-161-generic-x86_64-with-glibc2.10
- Python version: 3.8.3
- PyArrow version: 6.0.1
| https://github.com/huggingface/datasets/issues/3422 | [
"Hi ! I wasn't able to reproduce your error.\r\n\r\nCan you try to clear your cache at `~/.cache/huggingface/modules` and try again ?"
] | null | 3,422 | false |
Adding mMARCO dataset | Adding mMARCO (v1.1) to HF datasets. | https://github.com/huggingface/datasets/pull/3421 | [
"Hi @albertvillanova we've made a major overhaul of the loading script including all configurations we're making available. Could you please review it again?",
"@albertvillanova :ping_pong: ",
"Thanks @lhbonifacio for adding this dataset.\r\nHi there, i got an error about mmarco:\r\nConnectionError: Couldn't re... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3421",
"html_url": "https://github.com/huggingface/datasets/pull/3421",
"diff_url": "https://github.com/huggingface/datasets/pull/3421.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3421.patch",
"merged_at": null
} | 3,421 | true |
Add eli5_category dataset | This pull request adds a categorized Long-form question answering dataset `ELI5_Category`. It's a new variant of the [ELI5](https://huggingface.co/datasets/eli5) dataset that uses the Reddit tags to alleviate the training/validation overlapping in the origin ELI5 dataset.
A [report](https://celeritasml.netlify.app/posts/2021-12-01-eli5c/)(Section 2) on this dataset. | https://github.com/huggingface/datasets/pull/3420 | [
"> Thanks a lot for adding this dataset ! Good job with the dataset card and the dataset scripts - they're really good :)\r\n> \r\n> I just added minor changes\r\n\r\nThanks for fixing typos!"
] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3420",
"html_url": "https://github.com/huggingface/datasets/pull/3420",
"diff_url": "https://github.com/huggingface/datasets/pull/3420.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3420.patch",
"merged_at": "2021-12-14T17:53:02"
} | 3,420 | true |
`.to_json` is extremely slow after `.select` | ## Describe the bug
Saving a dataset to JSON with `to_json` is extremely slow after using `.select` on the original dataset.
## Steps to reproduce the bug
```python
from datasets import load_dataset
original = load_dataset("squad", split="train")
original.to_json("from_original.json") # Takes 0 seconds
selected_subset1 = original.select([i for i in range(len(original))])
selected_subset1.to_json("from_select1.json") # Takes 212 seconds
selected_subset2 = original.select([i for i in range(int(len(original) / 2))])
selected_subset2.to_json("from_select2.json") # Takes 90 seconds
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: master (https://github.com/huggingface/datasets/commit/6090f3cfb5c819f441dd4a4bb635e037c875b044)
- Platform: Linux-4.4.0-19041-Microsoft-x86_64-with-glibc2.27
- Python version: 3.9.7
- PyArrow version: 6.0.0
| https://github.com/huggingface/datasets/issues/3419 | [
"Hi ! It's slower indeed because a datasets on which `select`/`shard`/`train_test_split`/`shuffle` has been called has to do additional steps to retrieve the data of the dataset table in the right order.\r\n\r\nIndeed, if you call `dataset.select([0, 5, 10])`, the underlying table of the dataset is not altered to k... | null | 3,419 | false |
Add Wikisource dataset | Add loading script for Wikisource dataset.
Fix #3399.
CC: @geohci, @yjernite | https://github.com/huggingface/datasets/pull/3418 | [
"As we are removing the dataset scripts from GitHub and moving them to the Hugging Face Hub, I am going to transfer this script to the repo: https://huggingface.co/datasets/wikimedia/wikisource"
] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3418",
"html_url": "https://github.com/huggingface/datasets/pull/3418",
"diff_url": "https://github.com/huggingface/datasets/pull/3418.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3418.patch",
"merged_at": null
} | 3,418 | true |
Fix type of bridge field in QED | Use `Value("string")` instead of `Value("bool")` for the feature type of the `"bridge"` field in the QED dataset. If the value is `False`, set to `None`.
The following paragraph in the QED repo explains the purpose of this field:
>Each annotation in referential_equalities is a pair of spans, the question_reference and the sentence_reference, corresponding to an entity mention in the question and the selected_sentence respectively. As described in the paper, sentence_references can be "bridged in", in which case they do not correspond with any actual span in the selected_sentence. Hence, sentence_reference spans contain an additional field, bridge, which is a prepositional phrase when a reference is bridged, and is False otherwise. Prepositional phrases serve to link bridged references to an anchoring phrase in the selected_sentence. In the case a sentence_reference is bridged, the start and end, as well as the span string, map to such an anchoring phrase in the selected_sentence.
Fix #3346
cc @VictorSanh | https://github.com/huggingface/datasets/pull/3417 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3417",
"html_url": "https://github.com/huggingface/datasets/pull/3417",
"diff_url": "https://github.com/huggingface/datasets/pull/3417.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3417.patch",
"merged_at": "2021-12-14T14:39:05"
} | 3,417 | true |
disaster_response_messages unavailable | ## Dataset viewer issue for '* disaster_response_messages*'
**Link:** https://huggingface.co/datasets/disaster_response_messages
Dataset unavailable. Link dead: https://datasets.appen.com/appen_datasets/disaster_response_data/disaster_response_messages_training.csv
Am I the one who added this dataset ?No
| https://github.com/huggingface/datasets/issues/3416 | [
"Hi, thanks for reporting! This is a duplicate of https://github.com/huggingface/datasets/issues/3240. We are working on a fix.\r\n\r\n"
] | null | 3,416 | false |
Non-deterministic tests: CI tests randomly fail | ## Describe the bug
Some CI tests fail randomly.
1. In https://github.com/huggingface/datasets/pull/3375/commits/c10275fe36085601cb7bdb9daee9a8f1fc734f48, there were 3 failing tests, only on Linux:
```
=========================== short test summary info ============================
FAILED tests/test_streaming_download_manager.py::test_streaming_dl_manager_get_extraction_protocol[https://drive.google.com/uc?export=download&id=1k92sUfpHxKq8PXWRr7Y5aNHXwOCNUmqh-zip]
FAILED tests/test_streaming_download_manager.py::test_streaming_gg_drive - Fi...
FAILED tests/test_streaming_download_manager.py::test_streaming_gg_drive_zipped
= 3 failed, 3553 passed, 2950 skipped, 2 xfailed, 1 xpassed, 125 warnings in 192.79s (0:03:12) =
```
2. After re-running the CI (without any change in the code) in https://github.com/huggingface/datasets/pull/3375/commits/57bfe1f342cd3c59d2510b992d5f06a0761eb147, there was only 1 failing test (one on Linux and a different one on Windows):
- On Linux:
```
=========================== short test summary info ============================
FAILED tests/test_streaming_download_manager.py::test_streaming_gg_drive_zipped
= 1 failed, 3555 passed, 2950 skipped, 2 xfailed, 1 xpassed, 125 warnings in 199.76s (0:03:19) =
```
- On Windows:
```
=========================== short test summary info ===========================
FAILED tests/test_load.py::test_load_dataset_builder_for_community_dataset_without_script
= 1 failed, 3551 passed, 2954 skipped, 2 xfailed, 1 xpassed, 121 warnings in 478.58s (0:07:58) =
```
The test `tests/test_streaming_download_manager.py::test_streaming_gg_drive_zipped` passes locally.
3. After re-running again the CI (without any change in the code) in https://github.com/huggingface/datasets/pull/3375/commits/39f32f2119cf91b86867216bb5c356c586503c6a, ALL the tests passed.
| https://github.com/huggingface/datasets/issues/3415 | [
"I think it might come from two different issues:\r\n1. Google Drive is an unreliable host, mainly because of quota limitations\r\n2. the staging environment can sometimes raise some errors\r\n\r\nFor Google Drive tests we could set up some retries with backup URLs if necessary I guess.\r\nFor staging on the other ... | null | 3,415 | false |
Skip None encoding (line deleted by accident in #3195) | Return the line deleted by accident in #3195 while [resolving merge conflicts](https://github.com/huggingface/datasets/pull/3195/commits/8b0ed15be08559056b817836a07d47acda0c4510).
Fix #3181 (finally :))
| https://github.com/huggingface/datasets/pull/3414 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3414",
"html_url": "https://github.com/huggingface/datasets/pull/3414",
"diff_url": "https://github.com/huggingface/datasets/pull/3414.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3414.patch",
"merged_at": "2021-12-10T11:00:02"
} | 3,414 | true |
Add WIDER FACE dataset | Adds the WIDER FACE face detection benchmark.
TODOs:
* [x] dataset card
* [x] dummy data | https://github.com/huggingface/datasets/pull/3413 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3413",
"html_url": "https://github.com/huggingface/datasets/pull/3413",
"diff_url": "https://github.com/huggingface/datasets/pull/3413.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3413.patch",
"merged_at": "2022-01-12T14:13:47"
} | 3,413 | true |
Fix flaky test again for s3 serialization | Following https://github.com/huggingface/datasets/pull/3388 that wasn't enough (see CI error [here](https://app.circleci.com/pipelines/github/huggingface/datasets/9080/workflows/b971fb27-ff20-4220-9416-c19acdfdf6f4/jobs/55985)) | https://github.com/huggingface/datasets/pull/3412 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3412",
"html_url": "https://github.com/huggingface/datasets/pull/3412",
"diff_url": "https://github.com/huggingface/datasets/pull/3412.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3412.patch",
"merged_at": "2021-12-09T18:00:52"
} | 3,412 | true |
[chinese wwm] load_datasets behavior not as expected when using run_mlm_wwm.py script | ## Describe the bug
Model I am using (Bert, XLNet ...): bert-base-chinese
The problem arises when using:
* [https://github.com/huggingface/transformers/blob/master/examples/research_projects/mlm_wwm/run_mlm_wwm.py] the official example scripts: `rum_mlm_wwm.py`
The tasks I am working on is: pretraining whole word masking with my own dataset and ref.json file
I tried follow the run_mlm_wwm.py procedure to do whole word masking on pretraining task. my file is in .txt form, where one line represents one sample, with `9,264,784` chinese lines in total. the ref.json file is also contains 9,264,784 lines of whole word masking reference data for my chinese corpus. but when I try to adapt the run_mlm_wwm.py script, it shows that somehow after
`datasets["train"] = load_dataset(...`
`len(datasets["train"])` returns `9,265,365`
then, after `tokenized_datasets = datasets.map(...`
`len(tokenized_datasets["train"])` returns `9,265,279`
I'm really confused and tried to trace code by myself but can't know what happened after a week trial.
I want to know what happened in the `load_dataset()` function and `datasets.map` here and how did I get more lines of data than I input. so I'm here to ask.
## To reproduce
Sorry that I can't provide my data here since it did not belong to me. but I'm sure I remove the blank lines.
## Expected behavior
I expect the code run as it should. but the AssertionError in line 167 keeps raise as the line of reference json and datasets['train'] differs.
Thanks for your patient reading!
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.8.0
- Platform: Linux-5.4.0-91-generic-x86_64-with-glibc2.29
- Python version: 3.8.10
- PyArrow version: 3.0.0
| https://github.com/huggingface/datasets/issues/3411 | [
"@LysandreJik not so sure who to @\r\nCould you help?",
"Hi @hyusterr, I believe it is @wlhgtc from https://github.com/huggingface/transformers/pull/9887"
] | null | 3,411 | false |
Fix dependencies conflicts in Windows CI after conda update to 4.11 | For some reason the CI wasn't using python 3.6 but python 3.7 after the update to conda 4.11 | https://github.com/huggingface/datasets/pull/3410 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3410",
"html_url": "https://github.com/huggingface/datasets/pull/3410",
"diff_url": "https://github.com/huggingface/datasets/pull/3410.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3410.patch",
"merged_at": "2021-12-09T17:36:19"
} | 3,410 | true |
Pass new_fingerprint in multiprocessing | Following https://github.com/huggingface/datasets/pull/3045
Currently one can pass `new_fingerprint` to `.map()` to use a custom fingerprint instead of the one computed by hashing the map transform. However it's ignored if `num_proc>1`.
In this PR I fixed that by passing `new_fingerprint` to `._map_single()` when `num_proc>1`.
More specifically, `new_fingerprint` with a suffix based on the process `rank` is passed, so that each process has a different `new_fingerprint`
cc @TevenLeScao @vlievin | https://github.com/huggingface/datasets/pull/3409 | [
"@lhoestq Hi~, does this support that `datasets.map(func, batched=True, batch_size, num_proc>1, new_fingerprint=\"func_v1\")` even if `func` can't pickle. I also notice that you said \"Unfortunately you need picklable mapping functions to make multiprocessing work :confused: Also feel free to open an issue or send ... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3409",
"html_url": "https://github.com/huggingface/datasets/pull/3409",
"diff_url": "https://github.com/huggingface/datasets/pull/3409.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3409.patch",
"merged_at": "2021-12-09T17:38:43"
} | 3,409 | true |
Typo in Dataset viewer error message | ## Dataset viewer issue for '*name of the dataset*'
**Link:** *link to the dataset viewer page*
*short description of the issue*
When creating an empty dataset repo, the Dataset Preview provides a helpful message that no files were found. There is a tiny typo in that message: "ressource" should be "resource"
Am I the one who added this dataset ?
N/A
| https://github.com/huggingface/datasets/issues/3408 | [
"Fixed, thanks\r\n<img width=\"661\" alt=\"Capture d’écran 2021-12-22 à 12 02 30\" src=\"https://user-images.githubusercontent.com/1676121/147082881-cf700e8d-0511-4431-b214-d6cf8137db10.png\">\r\n"
] | null | 3,408 | false |
Use max number of data files to infer module | When inferring the module for datasets without script, set a maximum number of iterations over data files.
This PR fixes the issue of taking too long when hundred of data files present.
Please, feel free to agree on both numbers:
```
# Datasets without script
DATA_FILES_MAX_NUMBER = 10
ARCHIVED_DATA_FILES_MAX_NUMBER = 5
```
Fix #3404. | https://github.com/huggingface/datasets/pull/3407 | [
"Cool thanks :) Feel free to merge if it's all good for you"
] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3407",
"html_url": "https://github.com/huggingface/datasets/pull/3407",
"diff_url": "https://github.com/huggingface/datasets/pull/3407.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3407.patch",
"merged_at": "2021-12-14T17:08:41"
} | 3,407 | true |
Fix module inference for archive with a directory | Fix module inference for an archive file that contains files within a directory.
Fix #3405. | https://github.com/huggingface/datasets/pull/3406 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3406",
"html_url": "https://github.com/huggingface/datasets/pull/3406",
"diff_url": "https://github.com/huggingface/datasets/pull/3406.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3406.patch",
"merged_at": "2021-12-08T13:03:28"
} | 3,406 | true |
ZIP format inference does not work when files located in a dir inside the archive | ## Describe the bug
When a zipped file contains archived files within a directory, the function `infer_module_for_data_files_in_archives` does not work.
It only works for files located in the root directory of the ZIP file.
## Steps to reproduce the bug
```python
infer_module_for_data_files_in_archives(["path/to/zip/file.zip"], False)
``` | https://github.com/huggingface/datasets/issues/3405 | [] | null | 3,405 | false |
Optimize ZIP format inference | **Is your feature request related to a problem? Please describe.**
When hundreds of ZIP files are present in a dataset, format inference takes too long.
See: https://github.com/bigscience-workshop/data_tooling/issues/232#issuecomment-986685497
**Describe the solution you'd like**
Iterate over a maximum number of files.
CC: @lhoestq
| https://github.com/huggingface/datasets/issues/3404 | [] | null | 3,404 | false |
Cannot import name 'maybe_sync' | ## Describe the bug
Cannot seem to import datasets when running run_summarizer.py script on a VM set up on ovhcloud
## Steps to reproduce the bug
```python
from datasets import load_dataset
```
## Expected results
No error
## Actual results
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/opt/conda/lib/python3.7/site-packages/datasets/__init__.py", line 34, in <module>
from .arrow_dataset import Dataset, concatenate_datasets
File "/opt/conda/lib/python3.7/site-packages/datasets/arrow_dataset.py", line 48, in <module>
from .arrow_writer import ArrowWriter, OptimizedTypedSequence
File "/opt/conda/lib/python3.7/site-packages/datasets/arrow_writer.py", line 27, in <module>
from .features import (
File "/opt/conda/lib/python3.7/site-packages/datasets/features/__init__.py", line 2, in <module>
from .audio import Audio
File "/opt/conda/lib/python3.7/site-packages/datasets/features/audio.py", line 8, in <module>
from ..utils.streaming_download_manager import xopen
File "/opt/conda/lib/python3.7/site-packages/datasets/utils/streaming_download_manager.py", line 16, in <module>
from ..filesystems import COMPRESSION_FILESYSTEMS
File "/opt/conda/lib/python3.7/site-packages/datasets/filesystems/__init__.py", line 13, in <module>
from .s3filesystem import S3FileSystem # noqa: F401
File "/opt/conda/lib/python3.7/site-packages/datasets/filesystems/s3filesystem.py", line 1, in <module>
import s3fs
File "/opt/conda/lib/python3.7/site-packages/s3fs/__init__.py", line 1, in <module>
from .core import S3FileSystem, S3File
File "/opt/conda/lib/python3.7/site-packages/s3fs/core.py", line 11, in <module>
from fsspec.asyn import AsyncFileSystem, sync, sync_wrapper, maybe_sync
ImportError: cannot import name 'maybe_sync' from 'fsspec.asyn' (/opt/conda/lib/python3.7/site-packages/fsspec/asyn.py)
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.16.0
- Platform: OVH Cloud Tesla V100 Machine
- Python version: 3.7.9
- PyArrow version: 6.0.1
| https://github.com/huggingface/datasets/issues/3403 | [
"Hi ! Can you try updating `fsspec` ? The minimum version is `2021.05.0`",
"hey @lhoestq. I'm using `fsspec-2021.11.1` but still getting that error.",
"Maybe this discussion can help:\r\n\r\nhttps://github.com/fsspec/filesystem_spec/issues/597#issuecomment-958646964",
"Thanks @lhoestq. Downgrading `fsspec and... | null | 3,403 | false |
More robust first elem check in encode/cast example | Fix #3306 | https://github.com/huggingface/datasets/pull/3402 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3402",
"html_url": "https://github.com/huggingface/datasets/pull/3402",
"diff_url": "https://github.com/huggingface/datasets/pull/3402.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3402.patch",
"merged_at": "2021-12-08T13:02:15"
} | 3,402 | true |
Add Wikimedia pre-processed datasets | ## Adding a Dataset
- **Name:** Add pre-processed data to:
- *wikimedia/wikipedia*: https://huggingface.co/datasets/wikimedia/wikipedia
- *wikimedia/wikisource*: https://huggingface.co/datasets/wikimedia/wikisource
- **Description:** Add pre-processed data to the Hub for all languages
- **Paper:** *link to the dataset paper if available*
- **Data:** *link to the Github repository or current dataset location*
- **Motivation:** This will be very useful for the NLP community, as the pre-processing has a high cost for lot of researchers (both in computation and in knowledge)
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
CC: @geohci, @yjernite | https://github.com/huggingface/datasets/issues/3401 | [] | null | 3,401 | false |
Improve Wikipedia loading script | As reported by @geohci, the "wikipedia" processing/loading script could be improved by some additional small suggested processing functions:
- _extract_content(filepath):
- Replace .startswith("#redirect") with more structured approach: if elem.find(f"./{namespace}redirect") is None: continue
- _parse_and_clean_wikicode(raw_content, parser):
- Remove rm_template from cleaning -- this is redundant with .strip_code() from mwparserformhell
- Build a language-specific list of namespace prefixes to filter out per below get_namespace_prefixes
- Optional: strip prefixes like categories -- e.g., Category:Towns in Tianjin becomes Towns in Tianjin
- Optional: strip magic words
| https://github.com/huggingface/datasets/issues/3400 | [
"Thanks! See https://public.paws.wmcloud.org/User:Isaac_(WMF)/HuggingFace%20Wikipedia%20Processing.ipynb for more implementation details / some data around the overhead induced by adding the extra preprocessing steps (stripping link prefixes and magic words)",
"Closed by:\r\n- #3435"
] | null | 3,400 | false |
Add Wikisource dataset | ## Adding a Dataset
- **Name:** *wikisource*
- **Description:** *short description of the dataset (or link to social media or blog post)*
- **Paper:** *link to the dataset paper if available*
- **Data:** *link to the Github repository or current dataset location*
- **Motivation:** Additional high quality textual data, besides Wikipedia.
Add loading script as "canonical" dataset (as it is the case for ""wikipedia").
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
CC: @geohci, @yjernite | https://github.com/huggingface/datasets/issues/3399 | [
"See notebook by @geohci: https://public.paws.wmcloud.org/User:Isaac_(WMF)/HuggingFace%20Wikisource%20Processing.ipynb"
] | null | 3,399 | false |
Add URL field to Wikimedia dataset instances: wikipedia,... | As reported by @geohci, in order to host pre-processed data in the Hub, we should add the full URL to data instances (new field "url"), so that we conform to proper attribution from license requirement. See, e.g.: https://fair-trec.github.io/docs/Fair_Ranking_2021_Participant_Instructions.pdf#subsection.3.2
This should be done for all pre-processed datasets under "wikimedia" org in the Hub: https://huggingface.co/wikimedia
| https://github.com/huggingface/datasets/issues/3398 | [
"@geohci, I think the field \"url\" does not appear in the Wikimedia dumps. Therefore I guess we should generate it, using the \"title\" field and making some transformation of it (replacing spaces with underscores) and prepending the domain (created using the language)?",
"Indeed:\r\n\r\n> To re-distribute text ... | null | 3,398 | false |
add BNL newspapers | This pull request adds the BNL's [processed newspaper collections](https://data.bnl.lu/data/historical-newspapers/) as a dataset. This is partly done to support BigScience see: https://github.com/bigscience-workshop/data_tooling/issues/192.
The Datacard is more sparse than I would like but I plan to make a separate pull request to try and make this more complete at a later date.
I had to manually add the `dummy_data` but I believe I've done this correctly (the tests pass locally).
| https://github.com/huggingface/datasets/pull/3397 | [
"\r\n> Also, maybe calling the dataset as \"bnl_historical_newspapers\" and setting \"processed\" as one configuration name?\r\n\r\nThis sounds like a good idea but my only question around this is how easy it would be to use the same approach for processing the other newspaper collections [https://data.bnl.lu/data/... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3397",
"html_url": "https://github.com/huggingface/datasets/pull/3397",
"diff_url": "https://github.com/huggingface/datasets/pull/3397.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3397.patch",
"merged_at": "2022-01-17T18:35:34"
} | 3,397 | true |
Install Audio dependencies to support audio decoding | ## Dataset viewer issue for '*openslr*', '*projecte-aina/parlament_parla*'
**Link:** *https://huggingface.co/datasets/openslr*
**Link:** *https://huggingface.co/datasets/projecte-aina/parlament_parla*
Error:
```
Status code: 400
Exception: ImportError
Message: To support decoding audio files, please install 'librosa'.
```
Am I the one who added this dataset ? Yes-No
- openslr: No
- projecte-aina/parlament_parla: Yes
| https://github.com/huggingface/datasets/issues/3396 | [
"https://huggingface.co/datasets/projecte-aina/parlament_parla -> works (but we still have to show an audio player)\r\n\r\nhttps://huggingface.co/datasets/openslr -> another issue: `Message: [Errno 2] No such file or directory: '/home/hf/datasets-preview-backend/zip:/asr_javanese/data/00/00004fe6aa.flac'`",
... | null | 3,396 | false |
Fix formatting in IterableDataset.map docs | Fix formatting in the recently added `Map` section of the streaming docs. | https://github.com/huggingface/datasets/pull/3395 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3395",
"html_url": "https://github.com/huggingface/datasets/pull/3395",
"diff_url": "https://github.com/huggingface/datasets/pull/3395.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3395.patch",
"merged_at": "2021-12-08T10:11:32"
} | 3,395 | true |
Preserve all feature types when saving a dataset on the Hub with `push_to_hub` | Currently, if one of the dataset features is of type `ClassLabel`, saving the dataset with `push_to_hub` and reloading the dataset with `load_dataset` will return the feature of type `Value`. To fix this, we should do something similar to `save_to_disk` (which correctly preserves the types) and not only push the parquet files in `push_to_hub`, but also the dataset `info` (stored in a JSON file). | https://github.com/huggingface/datasets/issues/3394 | [
"According to this [comment in the forum](https://discuss.huggingface.co/t/save-datasetdict-to-huggingface-hub/12075/8?u=lhoestq), using `push_to_hub` on a dataset with `ClassLabel` can also make the feature simply disappear when it's reloaded !",
"Maybe we can also fix https://github.com/huggingface/datasets/iss... | null | 3,394 | false |
Common Voice Belarusian Dataset | ## Adding a Dataset
- **Name:** *Common Voice Belarusian Dataset*
- **Description:** *[commonvoice.mozilla.org/be](https://commonvoice.mozilla.org/be)*
- **Data:** *[commonvoice.mozilla.org/be/datasets](https://commonvoice.mozilla.org/be/datasets)*
- **Motivation:** *It has more than 7GB of data, so it will be great to have it in this package so anyone can try to train something for Belarusian language.*
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| https://github.com/huggingface/datasets/issues/3393 | [] | null | 3,393 | false |
Dataset viewer issue for `dansbecker/hackernews_hiring_posts` | ## Dataset viewer issue for `dansbecker/hackernews_hiring_posts`
**Link:** https://huggingface.co/datasets/dansbecker/hackernews_hiring_posts
*short description of the issue*
Dataset preview not showing for uploaded DatasetDict. See https://discuss.huggingface.co/t/dataset-preview-not-showing-for-uploaded-datasetdict/12603
Am I the one who added this dataset ?
No -> @dansbecker | https://github.com/huggingface/datasets/issues/3392 | [
"This issue was fixed by me calling `all_datasets.push_to_hub(\"hackernews_hiring_posts\")`.\r\n\r\nThe previous problems were from calling `all_datasets.save_to_disk` and then pushing with `my_repo.git_add` and `my_repo.push_to_hub`.\r\n"
] | null | 3,392 | false |
method to select columns | **Is your feature request related to a problem? Please describe.**
* There is currently no way to select some columns of a dataset. In pandas, one can use `df[['col1', 'col2']]` to select columns, but in `datasets`, it results in error.
**Describe the solution you'd like**
* A new method that can be used to create a new dataset with only a list of specified columns.
**Describe alternatives you've considered**
`.remove_columns(self, columns: Union[str, List[str]], inverse: bool = False)`
Or
`.select(self, indices: Iterable = None, columns: List[str] = None)`
| https://github.com/huggingface/datasets/issues/3391 | [
"duplicate of #2655"
] | null | 3,391 | false |
Loading dataset throws "KeyError: 'Field "builder_name" does not exist in table schema'" | ## Describe the bug
I have prepared dataset to datasets and now I am trying to load it back Finnish-NLP/voxpopuli_fi
I get "KeyError: 'Field "builder_name" does not exist in table schema'"
My dataset folder and files should be like @patrickvonplaten has here https://huggingface.co/datasets/flax-community/german-common-voice-processed
How my voxpopuli dataset looks like:
Part of the processing (path column is the absolute path to audio files)
```
def add_audio_column(example):
example['audio'] = example['path']
return example
voxpopuli = voxpopuli.map(add_audio_column)
voxpopuli.cast_column("audio", Audio())
voxpopuli["audio"] <-- to my knowledge this does load the local files and prepares those arrays
voxpopuli = voxpopuli.cast_column("audio", Audio(sampling_rate=16_000)) resampling 16kHz
```
I have then saved it to disk_
`voxpopuli.save_to_disk('/asr_disk/datasets_processed_new/voxpopuli')`
and made folder structure same as @patrickvonplaten
I also get same error while trying to load_dataset from his repo:
## Steps to reproduce the bug
```python
dataset = load_dataset("Finnish-NLP/voxpopuli_fi")
```
## Expected results
Dataset is loaded correctly and looks like in the first picture
## Actual results
Loading throws keyError:
KeyError: 'Field "builder_name" does not exist in table schema'
Resources I have been trying to follow:
https://huggingface.co/docs/datasets/audio_process.html
https://huggingface.co/docs/datasets/share_dataset.html
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.16.2.dev0
- Platform: Ubuntu 20.04.2 LTS
- Python version: 3.8.12
- PyArrow version: 6.0.1
| https://github.com/huggingface/datasets/issues/3390 | [
"Got solved it with push_to_hub, closing"
] | null | 3,390 | false |
Add EDGAR | ## Adding a Dataset
- **Name:** EDGAR Database
- **Description:** https://www.sec.gov/edgar/about EDGAR, the Electronic Data Gathering, Analysis, and Retrieval system, is the primary system for companies and others submitting documents under the Securities Act of 1933, the Securities Exchange Act of 1934, the Trust Indenture Act of 1939, and the Investment Company Act of 1940. Containing millions of company and individual filings, EDGAR benefits investors, corporations, and the U.S. economy overall by increasing the efficiency, transparency, and fairness of the securities markets. The system processes about 3,000 filings per day, serves up 3,000 terabytes of data to the public annually, and accommodates 40,000 new filers per year on average. EDGAR® and EDGARLink® are registered trademarks of the SEC.
- **Data:** https://www.sec.gov/os/accessing-edgar-data
- **Motivation:** Enabling and improving FSI (Financial Services Industry) datasets to increase ease of use
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
| https://github.com/huggingface/datasets/issues/3389 | [
"cc @juliensimon ",
"Datasets are not tracked in this repository anymore. But you can make your own dataset in the huggingface hub"
] | null | 3,389 | false |
Fix flaky test of the temporary directory used by load_from_disk | The test is flaky, here is an example of random CI failure:
https://github.com/huggingface/datasets/commit/73ed6615b4b3eb74d5311684f7b9e05cdb76c989
I fixed that by not checking the content of the random part of the temporary directory name | https://github.com/huggingface/datasets/pull/3388 | [
"CI failed because of a server error - merging"
] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3388",
"html_url": "https://github.com/huggingface/datasets/pull/3388",
"diff_url": "https://github.com/huggingface/datasets/pull/3388.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3388.patch",
"merged_at": "2021-12-06T11:24:49"
} | 3,388 | true |
Create Language Modeling task | Create Language Modeling task to be able to specify the input "text" column in a dataset.
This can be useful for datasets which are not exclusively used for language modeling and have more than one column:
- for text classification datasets (with columns "review" and "rating", for example), the Language Modeling task can be used to specify the "text" column ("review" in this case).
TODO:
- [ ] Add the LanguageModeling task to all dataset scripts which can be used for language modeling | https://github.com/huggingface/datasets/pull/3387 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3387",
"html_url": "https://github.com/huggingface/datasets/pull/3387",
"diff_url": "https://github.com/huggingface/datasets/pull/3387.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3387.patch",
"merged_at": "2021-12-17T17:18:27"
} | 3,387 | true |
Fix typos in dataset cards | This PR:
- Fix typos in dataset cards
- Fix Papers With Code ID for:
- Bilingual Corpus of Arabic-English Parallel Tweets
- Tweets Hate Speech Detection
- Add pretty name tags | https://github.com/huggingface/datasets/pull/3386 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3386",
"html_url": "https://github.com/huggingface/datasets/pull/3386",
"diff_url": "https://github.com/huggingface/datasets/pull/3386.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3386.patch",
"merged_at": "2021-12-06T09:30:54"
} | 3,386 | true |
None batched `with_transform`, `set_transform` | **Is your feature request related to a problem? Please describe.**
A `torch.utils.data.Dataset.__getitem__` operates on a single example.
But 🤗 `Datasets.with_transform` doesn't seem to allow non-batched transform.
**Describe the solution you'd like**
Have a `batched=True` argument in `Datasets.with_transform`
**Describe alternatives you've considered**
* Convert a non-batched transform function to batched one myself.
* Wrap a 🤗 Dataset with torch Dataset, and add a `__getitem__`. 🙄
* Have `lazy=False` in `Dataset.map`, and returns a `LazyDataset` if `lazy=True`. This way the same `map` interface can be used, and existing code can be updated with one argument change. | https://github.com/huggingface/datasets/issues/3385 | [
"Hi ! Thanks for the suggestion :)\r\nIt makes sense to me, and it can surely be implemented by wrapping the user's function to make it a batched function. However I'm not a big fan of the inconsistency it would create with `map`: `with_transform` is batched by default while `map` isn't.\r\n\r\nIs there something y... | null | 3,385 | false |
Adding mMARCO dataset | We are adding mMARCO dataset to HuggingFace datasets repo.
This way, all the languages covered in the translation are available in a easy way. | https://github.com/huggingface/datasets/pull/3384 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3384",
"html_url": "https://github.com/huggingface/datasets/pull/3384",
"diff_url": "https://github.com/huggingface/datasets/pull/3384.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3384.patch",
"merged_at": null
} | 3,384 | true |
add Georgian data in cc100. | update cc100 dataset to support loading Georgian (ka) data which is originally available in CC100 dataset source.
All tests are passed.
Dummy data generated.
metadata generated. | https://github.com/huggingface/datasets/pull/3383 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3383",
"html_url": "https://github.com/huggingface/datasets/pull/3383",
"diff_url": "https://github.com/huggingface/datasets/pull/3383.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3383.patch",
"merged_at": "2021-12-14T14:37:22"
} | 3,383 | true |
#3337 Add typing overloads to Dataset.__getitem__ for mypy | Add typing overloads to Dataset.__getitem__ for mypy
Fixes #3337
**Iterable**
Iterable from `collections` cannot have a type, so you can't do `Iterable[int]` for example. `typing` has a Generic version that builds upon the one from `collections`.
**Flake8**
I had to add `# noqa: F811`, this is a bug from Flake8.
datasets uses flake8==3.7.9 which released in October 2019 if I update flake8 (4.0.1), I no longer get these errors, but I did not want to make the update without your approval. (It also triggers other errors like no args in f-strings.) | https://github.com/huggingface/datasets/pull/3382 | [
"Locally the `make quality` passes with the same dependencies. I would suggest upgrading flake8. (I can take care of it in another PR)\r\ncc @lhoestq ",
"Thank you for fixing flake8! I think we are ready to merge then. "
] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3382",
"html_url": "https://github.com/huggingface/datasets/pull/3382",
"diff_url": "https://github.com/huggingface/datasets/pull/3382.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3382.patch",
"merged_at": "2021-12-14T10:28:54"
} | 3,382 | true |
Unable to load audio_features from common_voice dataset | ## Describe the bug
I am not able to load audio features from common_voice dataset
## Steps to reproduce the bug
```
from datasets import load_dataset
import torchaudio
test_dataset = load_dataset("common_voice", "hi", split="test[:2%]")
resampler = torchaudio.transforms.Resample(48_000, 16_000)
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
```
## Expected results
This piece of code should return test_dataset after loading audio features.
## Actual results
Reusing dataset common_voice (/home/jovyan/.cache/huggingface/datasets/common_voice/hi/6.1.0/b879a355caa529b11f2249400b61cadd0d9433f334d5c60f8c7216ccedfecfe1)
/opt/conda/lib/python3.7/site-packages/transformers/configuration_utils.py:341: UserWarning: Passing `gradient_checkpointing` to a config initialization is deprecated and will be removed in v5 Transformers. Using `model.gradient_checkpointing_enable()` instead, or if you are using the `Trainer` API, pass `gradient_checkpointing=True` in your `TrainingArguments`.
"Passing `gradient_checkpointing` to a config initialization is deprecated and will be removed in v5 "
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
0%| | 0/3 [00:00<?, ?ex/s]formats: can't open input file `common_voice_hi_23795358.mp3': No such file or directory
0%| | 0/3 [00:00<?, ?ex/s]
Traceback (most recent call last):
File "demo_file.py", line 23, in <module>
test_dataset = test_dataset.map(speech_file_to_array_fn)
File "/opt/conda/lib/python3.7/site-packages/datasets/arrow_dataset.py", line 2036, in map
desc=desc,
File "/opt/conda/lib/python3.7/site-packages/datasets/arrow_dataset.py", line 518, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
File "/opt/conda/lib/python3.7/site-packages/datasets/arrow_dataset.py", line 485, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
File "/opt/conda/lib/python3.7/site-packages/datasets/fingerprint.py", line 411, in wrapper
out = func(self, *args, **kwargs)
File "/opt/conda/lib/python3.7/site-packages/datasets/arrow_dataset.py", line 2368, in _map_single
example = apply_function_on_filtered_inputs(example, i, offset=offset)
File "/opt/conda/lib/python3.7/site-packages/datasets/arrow_dataset.py", line 2277, in apply_function_on_filtered_inputs
processed_inputs = function(*fn_args, *additional_args, **fn_kwargs)
File "/opt/conda/lib/python3.7/site-packages/datasets/arrow_dataset.py", line 1978, in decorated
result = f(decorated_item, *args, **kwargs)
File "demo_file.py", line 19, in speech_file_to_array_fn
speech_array, sampling_rate = torchaudio.load(batch["path"])
File "/opt/conda/lib/python3.7/site-packages/torchaudio/backend/sox_io_backend.py", line 154, in load
filepath, frame_offset, num_frames, normalize, channels_first, format)
RuntimeError: Error loading audio file: failed to open file common_voice_hi_23795358.mp3
## Environment info
- `datasets` version: 1.16.1
- Platform: Linux-4.14.243 with-debian-bullseye-sid
- Python version: 3.7.9
- PyArrow version: 6.0.1
| https://github.com/huggingface/datasets/issues/3381 | [
"Hi ! Feel free to access `batch[\"audio\"][\"array\"]` and `batch[\"audio\"][\"sampling_rate\"]` instead\r\n\r\n`datasets` 1.16 introduced some changes in `common_voice` and now the `path` field is no longer a path to a local file (but rather the path to the file in the archive it's extracted from)",
"Thanks for... | null | 3,381 | false |
[Quick poll] Give your opinion on the future of the Hugging Face Open Source ecosystem! | Thanks to all of you, `datasets` will pass 11.5k stars :star2: this week!
If you have a couple of minutes and want to participate in shaping the future of the ecosystem, please share your thoughts:
[**hf.co/oss-survey**](https://hf.co/oss-survey)
(please reply in the above feedback form rather than to this thread)
Thank you all on behalf of the HuggingFace team! 🤗 | https://github.com/huggingface/datasets/issues/3380 | [] | null | 3,380 | false |
iter_archive on zipfiles with better compression type check | Hello @lhoestq , thank you for your detailed answer on previous PR !
I made this new PR because I misused git on the previous one #3347.
Related issue #3272.
# Comments :
* For extension check I used the `_get_extraction_protocol` function in **download_manager.py** with a slight change and called it `_get_extraction_protocol_local`:
**I removed this part :**
```python
elif path.endswith(".tar.gz") or path.endswith(".tgz"):
raise NotImplementedError(
f"Extraction protocol for TAR archives like '{urlpath}' is not implemented in streaming mode. Please use `dl_manager.iter_archive` instead."
)
```
**And also changed :**
```diff
- extension = path.split(".")[-1]
+ extension = "tar" if path.endswith(".tar.gz") else path.split(".")[-1]
```
The reason for this is a compression like **.tar.gz** will be considered a **.gz** which is handled with **zipfile**, though **tar.gz** can only be opened using **tarfile**.
Please tell me if there's anything to change.
# Tasks :
- [x] download_manager.py
- [x] streaming_download_manager.py | https://github.com/huggingface/datasets/pull/3379 | [
"Hello @lhoestq, thank you for your answer.\r\n\r\nI don't use pytest a lot so I think I might need some help on it :) but I tried some tests for `streaming_download_manager.py` only. I don't know how to test `download_manager.py` since we need to use local files.\r\n\r\n# Comments : \r\n* In **download_manager.py*... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3379",
"html_url": "https://github.com/huggingface/datasets/pull/3379",
"diff_url": "https://github.com/huggingface/datasets/pull/3379.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3379.patch",
"merged_at": "2023-01-24T12:53:08"
} | 3,379 | true |
Add The Pile subsets | Add The Pile subsets:
- pubmed
- ubuntu_irc
- europarl
- hacker_news
- nih_exporter
Close bigscience-workshop/data_tooling#301.
CC: @StellaAthena | https://github.com/huggingface/datasets/pull/3378 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3378",
"html_url": "https://github.com/huggingface/datasets/pull/3378",
"diff_url": "https://github.com/huggingface/datasets/pull/3378.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3378.patch",
"merged_at": "2021-12-09T18:11:23"
} | 3,378 | true |
COCO 🥥 on the 🤗 Hub? | This is a draft PR since I ran into few small problems.
I referred to this TFDS code: https://github.com/tensorflow/datasets/blob/2538a08c184d53b37bfcf52cc21dd382572a88f4/tensorflow_datasets/object_detection/coco.py
cc: @mariosasko | https://github.com/huggingface/datasets/pull/3377 | [
"@mariosasko I fixed couple of bugs",
"TO-DO: \r\n- [x] Add unlabeled 2017 splits, train and validation splits of 2015\r\n- [x] Add Class Labels as list instead",
"@mariosasko added fine & coarse grained labels, will fix the bugs (currently getting set up with VM, my internet is too slow to run the tests and do... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3377",
"html_url": "https://github.com/huggingface/datasets/pull/3377",
"diff_url": "https://github.com/huggingface/datasets/pull/3377.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3377.patch",
"merged_at": null
} | 3,377 | true |
Update clue benchmark | Fix #3374 | https://github.com/huggingface/datasets/pull/3376 | [
"The CI error is due to missing tags in the CLUE dataset card - merging !"
] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3376",
"html_url": "https://github.com/huggingface/datasets/pull/3376",
"diff_url": "https://github.com/huggingface/datasets/pull/3376.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3376.patch",
"merged_at": "2021-12-08T14:14:41"
} | 3,376 | true |
Support streaming zipped dataset repo by passing only repo name | Proposed solution:
- I have added the method `iter_files` to DownloadManager and StreamingDownloadManager
- I use this in modules: "csv", "json", "text"
- I test for CSV/JSONL/TXT zipped (and non-zipped) files, both in streaming and non-streaming modes
Fix #3373. | https://github.com/huggingface/datasets/pull/3375 | [
"I just tested and I think this only opens one file ? If there are several files in the ZIP, only the first one is opened. To open several files from a ZIP, one has to call `open` several times.\r\n\r\nWhat about updating the CSV loader to make it `download_and_extract` zip files, and open each extracted file ?",
... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3375",
"html_url": "https://github.com/huggingface/datasets/pull/3375",
"diff_url": "https://github.com/huggingface/datasets/pull/3375.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3375.patch",
"merged_at": "2021-12-16T18:03:31"
} | 3,375 | true |
NonMatchingChecksumError for the CLUE:cluewsc2020, chid, c3 and tnews | Hi, it seems like there are updates in cluewsc2020, chid, c3 and tnews, since i could not load them due to the checksum error. | https://github.com/huggingface/datasets/issues/3374 | [
"Seems like the issue still exists,:\r\n`Downloading and preparing dataset clue/chid (download: 127.15 MiB, generated: 259.71 MiB, post-processed: Unknown size, total: 386.86 MiB) to /mnt/cache/tanhaochen/.cache/huggingface/datasets/clue/chid/1.0.0/e55b490cb7809dcd8db31b9a87119f2e2ec87cdc060da8a9ac070b070ca3e379...... | null | 3,374 | false |
Support streaming zipped CSV dataset repo by passing only repo name | Given a community 🤗 dataset repository containing only a zipped CSV file (only raw data, no loading script), I would like to load it in streaming mode without passing `data_files`:
```
ds_name = "bigscience-catalogue-data/vietnamese_poetry_from_fsoft_ai_lab"
ds = load_dataset(ds_name, split="train", streaming=True, use_auth_token=True)
item = next(iter(ds))
```
Currently, it gives a `FileNotFoundError` because there is no glob (no "\*" after "zip://": "zip://*") in the passed URL:
```
'zip://::https://huggingface.co/datasets/bigscience-catalogue-data/vietnamese_poetry_from_fsoft_ai_lab/resolve/e5d45f1bd9a8a798cc14f0a45ebc1ce91907c792/poems_dataset.zip'
```
| https://github.com/huggingface/datasets/issues/3373 | [] | null | 3,373 | false |
[SEO improvement] Add Dataset Metadata to make datasets indexable | Some people who host datasets on github seem to include a table of metadata at the end of their README.md to make the dataset indexable by [Google Dataset Search](https://datasetsearch.research.google.com/) (See [here](https://github.com/google-research/google-research/tree/master/goemotions#dataset-metadata) and [here](https://github.com/cvdfoundation/google-landmark#dataset-metadata)). This could be a useful addition to canonical datasets; perhaps even community datasets.
I'll include a screenshot (as opposed to markdown) as an example so as not to have a github issue indexed as a dataset:
> 
**_PS: It might very well be the case that this is already covered by some other markdown magic I'm not aware of._**
| https://github.com/huggingface/datasets/issues/3372 | [] | null | 3,372 | false |
New: Americas NLI dataset | This PR adds the [Americas NLI](https://arxiv.org/abs/2104.08726) dataset, extension of XNLI to 10 low-resource indigenous languages spoken in the Americas: Ashaninka, Aymara, Bribri, Guarani, Nahuatl, Otomi, Quechua, Raramuri, Shipibo-Konibo, and Wixarika.
One odd thing (not sure) is that I had to set
`datasets-cli dummy_data ./datasets/americas_nli/ --auto_generate --n_lines 7500`
`n_lines` very large to successfully generate the dummy files for all the subsets. Happy to get some guidance here.
Otherwise, I hope everything is in order :)
e: missed a step, onto fixing the tests
e2: there you go -- hope it's ok to have added more languages with their ISO codes to `languages.json`, need those tests to pass :laughing: | https://github.com/huggingface/datasets/pull/3371 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3371",
"html_url": "https://github.com/huggingface/datasets/pull/3371",
"diff_url": "https://github.com/huggingface/datasets/pull/3371.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3371.patch",
"merged_at": "2021-12-08T13:58:11"
} | 3,371 | true |
Document a training loop for streaming dataset | I added some docs about streaming dataset. In particular I added two subsections:
- one on how to use `map` for preprocessing
- one on how to use a streaming dataset in a pytorch training loop
cc @patrickvonplaten @stevhliu if you have some comments
cc @Rocketknight1 later we can add the one for TF and I might need your help ^^' | https://github.com/huggingface/datasets/pull/3370 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3370",
"html_url": "https://github.com/huggingface/datasets/pull/3370",
"diff_url": "https://github.com/huggingface/datasets/pull/3370.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3370.patch",
"merged_at": "2021-12-03T13:34:34"
} | 3,370 | true |
[Audio] Allow resampling for audio datasets in streaming mode | Many audio datasets like Common Voice always need to be resampled. This can very easily be done in non-streaming mode as follows:
```python
from datasets import load_dataset
ds = load_dataset("common_voice", "ab", split="test")
ds = ds.cast_column("audio", Audio(sampling_rate=16_000))
```
However in streaming mode it fails currently:
```python
from datasets import load_dataset
ds = load_dataset("common_voice", "ab", split="test", streaming=True)
ds = ds.cast_column("audio", Audio(sampling_rate=16_000))
```
with the following error:
```
AttributeError: 'IterableDataset' object has no attribute 'cast_column'
```
It would be great if we could add such a feature (I'm not 100% sure though how complex this would be) | https://github.com/huggingface/datasets/issues/3369 | [
"This requires implementing `cast_column` for iterable datasets, it could be a very nice addition !\r\n\r\n<s>It can also be useful to be able to disable the audio/image decoding for the dataset viewer (see PR https://github.com/huggingface/datasets/pull/3430) cc @severo </s>\r\nEDIT: actually following https://git... | null | 3,369 | false |
Fix dict source_datasets tagset validator | Currently, the `source_datasets` tag validation does not support passing a dict with configuration keys.
This PR:
- Extends `tagset_validator` to support regex tags
- Uses `tagset_validator` to validate dict `source_datasets` | https://github.com/huggingface/datasets/pull/3368 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3368",
"html_url": "https://github.com/huggingface/datasets/pull/3368",
"diff_url": "https://github.com/huggingface/datasets/pull/3368.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3368.patch",
"merged_at": "2021-12-02T15:48:37"
} | 3,368 | true |
Fix typo in other-structured-to-text task tag | Fix typo in task tag:
- `other-stuctured-to-text` (before)
- `other-structured-to-text` (now) | https://github.com/huggingface/datasets/pull/3367 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3367",
"html_url": "https://github.com/huggingface/datasets/pull/3367",
"diff_url": "https://github.com/huggingface/datasets/pull/3367.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3367.patch",
"merged_at": "2021-12-02T16:07:13"
} | 3,367 | true |
Add multimodal datasets | Epic issue to track the addition of multimodal datasets:
- [ ] #2526
- [x] #1842
- [ ] #1810
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
@VictorSanh feel free to add and sort by priority any interesting dataset. I have added the multimodal dataset requests which were already present as issues. | https://github.com/huggingface/datasets/issues/3366 | [] | null | 3,366 | false |
Add task tags for multimodal datasets | ## **Is your feature request related to a problem? Please describe.**
Currently, task tags are either exclusively related to text or speech processing:
- https://github.com/huggingface/datasets/blob/master/src/datasets/utils/resources/tasks.json
## **Describe the solution you'd like**
We should also add tasks related to:
- multimodality
- image
- video
CC: @VictorSanh @lewtun @lhoestq @merveenoyan @SBrandeis | https://github.com/huggingface/datasets/issues/3365 | [] | null | 3,365 | false |
Use the Audio feature in the AutomaticSpeechRecognition template | This updates the ASR template and all supported datasets to use the `Audio` feature | https://github.com/huggingface/datasets/pull/3364 | [
"Cool !\r\n\r\nI noticed that you removed the `audio_file_path_column` field of the template, note that you also have to update all the dataset_infos.json file that still contain this outdated field. For example in the common_voice you can find this:\r\n```\r\n\"task_templates\": [{\"task\": \"automatic-speech-reco... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3364",
"html_url": "https://github.com/huggingface/datasets/pull/3364",
"diff_url": "https://github.com/huggingface/datasets/pull/3364.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3364.patch",
"merged_at": null
} | 3,364 | true |
Update URL of Jeopardy! dataset | Updates the URL of the Jeopardy! dataset.
Fix #3361 | https://github.com/huggingface/datasets/pull/3363 | [
"Closing this PR in favor of #3266.",
"I think you should also close this branch"
] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3363",
"html_url": "https://github.com/huggingface/datasets/pull/3363",
"diff_url": "https://github.com/huggingface/datasets/pull/3363.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3363.patch",
"merged_at": null
} | 3,363 | true |
Adapt image datasets | This PR:
* adapts the ImageClassification template to use the new Image feature
* adapts the following datasets to use the new Image feature:
* beans (+ fixes streaming)
* cast_vs_dogs (+ fixes streaming)
* cifar10
* cifar100
* fashion_mnist
* mnist
* head_qa
cc @nateraw | https://github.com/huggingface/datasets/pull/3362 | [
"This PR can be merged after #3163 is merged (this PR is pretty big because I was working on the forked branch).\r\n\r\n@lhoestq @albertvillanova Could you please take a look at the changes in `src/datasets/utils/streaming_download_manager.py`? These changes were required to support streaming of the `cats_vs_dogs` ... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3362",
"html_url": "https://github.com/huggingface/datasets/pull/3362",
"diff_url": "https://github.com/huggingface/datasets/pull/3362.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3362.patch",
"merged_at": "2021-12-09T18:37:41"
} | 3,362 | true |
Jeopardy _URL access denied | ## Describe the bug
http://skeeto.s3.amazonaws.com/share/JEOPARDY_QUESTIONS1.json.gz returns Access Denied now.
However, https://drive.google.com/file/d/0BwT5wj_P7BKXb2hfM3d2RHU1ckE/view?usp=sharing from the original Reddit post https://www.reddit.com/r/datasets/comments/1uyd0t/200000_jeopardy_questions_in_a_json_file/ may work.
## Steps to reproduce the bug
```shell
> python
Python 3.7.12 (default, Sep 5 2021, 08:34:29)
[Clang 11.0.3 (clang-1103.0.32.62)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
```
```python
>>> from datasets import load_dataset
>>> load_dataset("jeopardy")
```
## Expected results
The download completes.
## Actual results
```shell
Downloading: 4.18kB [00:00, 1.60MB/s]
Downloading: 2.03kB [00:00, 1.04MB/s]
Using custom data configuration default
Downloading and preparing dataset jeopardy/default (download: 12.13 MiB, generated: 34.46 MiB, post-processed: Unknown size, total: 46.59 MiB) to /Users/mike/.cache/huggingface/datasets/jeopardy/default/0.1.0/25ee3e4a73755e637b8810f6493fd36e4523dea3ca8a540529d0a6e24c7f9810...
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/Users/mike/Library/Caches/pypoetry/virtualenvs/promptsource-hsdAcWsQ-py3.7/lib/python3.7/site-packages/datasets/load.py", line 1632, in load_dataset
use_auth_token=use_auth_token,
File "/Users/mike/Library/Caches/pypoetry/virtualenvs/promptsource-hsdAcWsQ-py3.7/lib/python3.7/site-packages/datasets/builder.py", line 608, in download_and_prepare
dl_manager=dl_manager, verify_infos=verify_infos, **download_and_prepare_kwargs
File "/Users/mike/Library/Caches/pypoetry/virtualenvs/promptsource-hsdAcWsQ-py3.7/lib/python3.7/site-packages/datasets/builder.py", line 675, in _download_and_prepare
split_generators = self._split_generators(dl_manager, **split_generators_kwargs)
File "/Users/mike/.cache/huggingface/modules/datasets_modules/datasets/jeopardy/25ee3e4a73755e637b8810f6493fd36e4523dea3ca8a540529d0a6e24c7f9810/jeopardy.py", line 72, in _split_generators
filepath = dl_manager.download_and_extract(_DATA_URL)
File "/Users/mike/Library/Caches/pypoetry/virtualenvs/promptsource-hsdAcWsQ-py3.7/lib/python3.7/site-packages/datasets/utils/download_manager.py", line 284, in download_and_extract
return self.extract(self.download(url_or_urls))
File "/Users/mike/Library/Caches/pypoetry/virtualenvs/promptsource-hsdAcWsQ-py3.7/lib/python3.7/site-packages/datasets/utils/download_manager.py", line 197, in download
download_func, url_or_urls, map_tuple=True, num_proc=download_config.num_proc, disable_tqdm=False
File "/Users/mike/Library/Caches/pypoetry/virtualenvs/promptsource-hsdAcWsQ-py3.7/lib/python3.7/site-packages/datasets/utils/py_utils.py", line 197, in map_nested
return function(data_struct)
File "/Users/mike/Library/Caches/pypoetry/virtualenvs/promptsource-hsdAcWsQ-py3.7/lib/python3.7/site-packages/datasets/utils/download_manager.py", line 217, in _download
return cached_path(url_or_filename, download_config=download_config)
File "/Users/mike/Library/Caches/pypoetry/virtualenvs/promptsource-hsdAcWsQ-py3.7/lib/python3.7/site-packages/datasets/utils/file_utils.py", line 305, in cached_path
use_auth_token=download_config.use_auth_token,
File "/Users/mike/Library/Caches/pypoetry/virtualenvs/promptsource-hsdAcWsQ-py3.7/lib/python3.7/site-packages/datasets/utils/file_utils.py", line 594, in get_from_cache
raise ConnectionError("Couldn't reach {}".format(url))
ConnectionError: Couldn't reach http://skeeto.s3.amazonaws.com/share/JEOPARDY_QUESTIONS1.json.gz
```
---
```shell
> curl http://skeeto.s3.amazonaws.com/share/JEOPARDY_QUESTIONS1.json.gz
```
```xml
<?xml version="1.0" encoding="UTF-8"?>
<Error><Code>AccessDenied</Code><Message>Access Denied</Message><RequestId>70Y9R36XNPEQXMGV</RequestId><HostId>G6F5AK4qo7JdaEdKGMtS0P6gdLPeFOdEfSEfvTOZEfk9km0/jAfp08QLfKSTFFj1oWIKoAoBehM=</HostId></Error>
```
## Environment info
<!-- You can run the command `datasets-cli env` and copy-and-paste its output below. -->
- `datasets` version: 1.14.0
- Platform: macOS Catalina 10.15.7
- Python version: 3.7.12
- PyArrow version: 6.0.1
| https://github.com/huggingface/datasets/issues/3361 | [
"Just a side note: duplicate #3264"
] | null | 3,361 | false |
Add The Pile USPTO subset | Add:
- USPTO subset of The Pile: "uspto" config
Close bigscience-workshop/data_tooling#297.
CC: @StellaAthena | https://github.com/huggingface/datasets/pull/3360 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3360",
"html_url": "https://github.com/huggingface/datasets/pull/3360",
"diff_url": "https://github.com/huggingface/datasets/pull/3360.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3360.patch",
"merged_at": "2021-12-03T11:45:27"
} | 3,360 | true |
Add The Pile Free Law subset | Add:
- Free Law subset of The Pile: "free_law" config
Close bigscience-workshop/data_tooling#75.
CC: @StellaAthena | https://github.com/huggingface/datasets/pull/3359 | [
"@albertvillanova Is there a specific reason you’re adding the Pile under “the” instead of under “pile”? That does not appear to be consistent with other datasets.",
"Hi @StellaAthena,\r\n\r\nI asked myself the same question, but at the end I decided to be consistent with previously added Pile subsets:\r\n- #2817... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3359",
"html_url": "https://github.com/huggingface/datasets/pull/3359",
"diff_url": "https://github.com/huggingface/datasets/pull/3359.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3359.patch",
"merged_at": "2021-12-01T17:30:43"
} | 3,359 | true |
add new field, and get errors | after adding new field **tokenized_examples["example_id"]**, and get errors below,
I think it is due to changing data to tensor, and **tokenized_examples["example_id"]** is string list
**all fields**
```
***************** train_dataset 1: Dataset({
features: ['attention_mask', 'end_positions', 'example_id', 'input_ids', 'start_positions', 'token_type_ids'],
num_rows: 87714
})
```
**Errors**
```
Traceback (most recent call last):
File "/usr/local/lib/python3.7/site-packages/transformers/tokenization_utils_base.py", line 705, in convert_to_tensors
tensor = as_tensor(value)
ValueError: too many dimensions 'str'
``` | https://github.com/huggingface/datasets/issues/3358 | [
"Hi, \r\n\r\ncould you please post this question on our [Forum](https://discuss.huggingface.co/) as we keep issues for bugs and feature requests? ",
"> Hi,\r\n> \r\n> could you please post this question on our [Forum](https://discuss.huggingface.co/) as we keep issues for bugs and feature requests?\r\n\r\nok."
] | null | 3,358 | false |
Update languages in aeslc dataset card | After having worked a bit with the dataset.
As far as I know, it is solely in English (en-US). There are only a few mails in Spanish, French or German (less than a dozen I would estimate). | https://github.com/huggingface/datasets/pull/3357 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3357",
"html_url": "https://github.com/huggingface/datasets/pull/3357",
"diff_url": "https://github.com/huggingface/datasets/pull/3357.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3357.patch",
"merged_at": "2022-09-23T13:16:48"
} | 3,357 | true |
to_tf_dataset() refactor | This is the promised cleanup to `to_tf_dataset()` now that the course is out of the way! The main changes are:
- A collator is always required (there was way too much hackiness making things like labels work without it)
- Lots of cleanup and a lot of code moved to `_get_output_signature`
- Should now handle it gracefully when the data collator adds unexpected columns | https://github.com/huggingface/datasets/pull/3356 | [
"Also, please don't merge yet - I need to make sure all the code samples and notebooks have a collate_fn specified, since we're removing the ability for this method to work without one!",
"Hi @lhoestq @mariosasko, the other PRs this was depending on in Transformers and huggingface/notebooks are now merged, so thi... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3356",
"html_url": "https://github.com/huggingface/datasets/pull/3356",
"diff_url": "https://github.com/huggingface/datasets/pull/3356.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3356.patch",
"merged_at": "2021-12-09T10:26:53"
} | 3,356 | true |
Extend support for streaming datasets that use pd.read_excel | This PR fixes error:
```
ValueError: Cannot seek streaming HTTP file
```
CC: @severo | https://github.com/huggingface/datasets/pull/3355 | [
"TODO in the future: https://github.com/huggingface/datasets/pull/3355#discussion_r761138011\r\n- If we finally find a use case where the `pd.read_excel()` can work in streaming mode (using fsspec), that is, without using the `.read()`, I propose to try this first, catch the ValueError and then try with `.read`, bu... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3355",
"html_url": "https://github.com/huggingface/datasets/pull/3355",
"diff_url": "https://github.com/huggingface/datasets/pull/3355.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3355.patch",
"merged_at": "2021-12-17T07:24:18"
} | 3,355 | true |
Remove duplicate name from dataset cards | Remove duplicate name from dataset card for:
- ajgt_twitter_ar
- emotone_ar | https://github.com/huggingface/datasets/pull/3354 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3354",
"html_url": "https://github.com/huggingface/datasets/pull/3354",
"diff_url": "https://github.com/huggingface/datasets/pull/3354.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3354.patch",
"merged_at": "2021-12-01T13:14:29"
} | 3,354 | true |
add one field "example_id", but I can't see it in the "comput_loss" function | Hi, I add one field **example_id**, but I can't see it in the **comput_loss** function, how can I do this? below is the information of inputs
```
*********************** inputs: {'attention_mask': tensor([[1, 1, 1, ..., 0, 0, 0],
[1, 1, 1, ..., 0, 0, 0],
[1, 1, 1, ..., 0, 0, 0],
...,
[1, 1, 1, ..., 0, 0, 0],
[1, 1, 1, ..., 0, 0, 0],
[1, 1, 1, ..., 0, 0, 0]], device='cuda:0'), 'end_positions': tensor([ 25, 97, 93, 44, 25, 112, 109, 134], device='cuda:0'), 'input_ids': tensor([[ 101, 2054, 2390, ..., 0, 0, 0],
[ 101, 2054, 2515, ..., 0, 0, 0],
[ 101, 2054, 2106, ..., 0, 0, 0],
...,
[ 101, 2339, 2001, ..., 0, 0, 0],
[ 101, 2054, 2515, ..., 0, 0, 0],
[ 101, 2054, 2003, ..., 0, 0, 0]], device='cuda:0'), 'start_positions': tensor([ 20, 90, 89, 41, 25, 96, 106, 132], device='cuda:0'), 'token_type_ids': tensor([[0, 0, 0, ..., 0, 0, 0],
[0, 0, 0, ..., 0, 0, 0],
[0, 0, 0, ..., 0, 0, 0],
...,
[0, 0, 0, ..., 0, 0, 0],
[0, 0, 0, ..., 0, 0, 0],
[0, 0, 0, ..., 0, 0, 0]], device='cuda:0')}
```
```
# This function preprocesses a question answering dataset, tokenizing the question and context text
# and finding the right offsets for the answer spans in the tokenized context (to use as labels).
# Adapted from https://github.com/huggingface/transformers/blob/master/examples/pytorch/question-answering/run_qa.py
def prepare_train_dataset_qa(examples, tokenizer, max_seq_length=None):
questions = [q.lstrip() for q in examples["question"]]
max_seq_length = tokenizer.model_max_length
# tokenize both questions and the corresponding context
# if the context length is longer than max_length, we split it to several
# chunks of max_length
tokenized_examples = tokenizer(
questions,
examples["context"],
truncation="only_second",
max_length=max_seq_length,
stride=min(max_seq_length // 2, 128),
return_overflowing_tokens=True,
return_offsets_mapping=True,
padding="max_length"
)
# Since one example might give us several features if it has a long context,
# we need a map from a feature to its corresponding example.
sample_mapping = tokenized_examples.pop("overflow_to_sample_mapping")
# The offset mappings will give us a map from token to character position
# in the original context. This will help us compute the start_positions
# and end_positions to get the final answer string.
offset_mapping = tokenized_examples.pop("offset_mapping")
tokenized_examples["start_positions"] = []
tokenized_examples["end_positions"] = []
tokenized_examples["example_id"] = []
for i, offsets in enumerate(offset_mapping):
input_ids = tokenized_examples["input_ids"][i]
# We will label features not containing the answer the index of the CLS token.
cls_index = input_ids.index(tokenizer.cls_token_id)
sequence_ids = tokenized_examples.sequence_ids(i)
# from the feature idx to sample idx
sample_index = sample_mapping[i]
# get the answer for a feature
answers = examples["answers"][sample_index]
tokenized_examples["example_id"].append(examples["id"][sample_index])
if len(answers["answer_start"]) == 0:
tokenized_examples["start_positions"].append(cls_index)
tokenized_examples["end_positions"].append(cls_index)
else:
# Start/end character index of the answer in the text.
start_char = answers["answer_start"][0]
end_char = start_char + len(answers["text"][0])
# Start token index of the current span in the text.
token_start_index = 0
while sequence_ids[token_start_index] != 1:
token_start_index += 1
# End token index of the current span in the text.
token_end_index = len(input_ids) - 1
while sequence_ids[token_end_index] != 1:
token_end_index -= 1
# Detect if the answer is out of the span (in which case this feature is labeled with the CLS index).
if not (offsets[token_start_index][0] <= start_char and
offsets[token_end_index][1] >= end_char):
tokenized_examples["start_positions"].append(cls_index)
tokenized_examples["end_positions"].append(cls_index)
else:
# Otherwise move the token_start_index and token_end_index to the two ends of the answer.
# Note: we could go after the last offset if the answer is the last word (edge case).
while token_start_index < len(offsets) and \
offsets[token_start_index][0] <= start_char:
token_start_index += 1
tokenized_examples["start_positions"].append(
token_start_index - 1)
while offsets[token_end_index][1] >= end_char:
token_end_index -= 1
tokenized_examples["end_positions"].append(token_end_index + 1)
return tokenized_examples
```
_Originally posted by @yanllearnn in https://github.com/huggingface/datasets/issues/3333#issuecomment-983457161_ | https://github.com/huggingface/datasets/issues/3353 | [
"Hi ! Your function looks fine, I used to map `squad` locally and it indeed added the `example_id` field correctly.\r\n\r\nHowever I think that in the `compute_loss` method only a subset of the fields are available: the model inputs. Since `example_id` is not a model input (it's not passed as a parameter to the mod... | null | 3,353 | false |
Make LABR dataset streamable | Fix LABR dataset to make it streamable.
Related to: #3350. | https://github.com/huggingface/datasets/pull/3352 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3352",
"html_url": "https://github.com/huggingface/datasets/pull/3352",
"diff_url": "https://github.com/huggingface/datasets/pull/3352.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3352.patch",
"merged_at": "2021-12-01T10:49:01"
} | 3,352 | true |
Add VCTK dataset | Fixes #1837. | https://github.com/huggingface/datasets/pull/3351 | [
"Hello @patrickvonplaten, I hope it's okay to ping you with a (dumb) question!\r\n\r\nI've been trying to get `dl_manager.download_and_extract(_DL_URL)` to work with no avail. I verified that this is a problem on two different machines (lab server, GCP), so I doubt it's an issue with network connectivity. Here is t... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3351",
"html_url": "https://github.com/huggingface/datasets/pull/3351",
"diff_url": "https://github.com/huggingface/datasets/pull/3351.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3351.patch",
"merged_at": "2021-12-28T15:05:07"
} | 3,351 | true |
Avoid content-encoding issue while streaming datasets | This PR will fix streaming of datasets served with gzip content-encoding:
```
ClientPayloadError: 400, message='Can not decode content-encoding: gzip'
```
Fix #2918.
CC: @severo | https://github.com/huggingface/datasets/pull/3350 | [] | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3350",
"html_url": "https://github.com/huggingface/datasets/pull/3350",
"diff_url": "https://github.com/huggingface/datasets/pull/3350.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3350.patch",
"merged_at": "2021-12-01T08:15:00"
} | 3,350 | true |
raise exception instead of using assertions. | fix for the remaining files https://github.com/huggingface/datasets/issues/3171 | https://github.com/huggingface/datasets/pull/3349 | [
"@mariosasko - Thanks for the review & suggestions. Updated as per the suggestions. ",
"@mariosasko - Hello, Are there any additional changes required from my end??. Wondering if this PR can be merged or still pending on additional steps.",
"@mariosasko - The approved changes in the PR now has conflicts with th... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3349",
"html_url": "https://github.com/huggingface/datasets/pull/3349",
"diff_url": "https://github.com/huggingface/datasets/pull/3349.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3349.patch",
"merged_at": "2021-12-20T16:07:27"
} | 3,349 | true |
BLEURT: Match key names to correspond with filename | In order to properly locate downloaded ckpt files key name needs to match filename. Correcting change introduced in #3235 | https://github.com/huggingface/datasets/pull/3348 | [
"Thanks for the suggestion! I think the current checked-in `CHECKPOINT_URLS` is already not working. I believe anyone who tried using the new ckpts (`BLEURT-20-X`) can't unless this fix is in. The zip file from bleurt side unzips to directory name matching the filename (capitalized for new ones). For example withou... | {
"url": "https://api.github.com/repos/huggingface/datasets/pulls/3348",
"html_url": "https://github.com/huggingface/datasets/pull/3348",
"diff_url": "https://github.com/huggingface/datasets/pull/3348.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/3348.patch",
"merged_at": "2021-12-07T16:06:57"
} | 3,348 | true |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.