id
int64 953M
3.35B
| number
int64 2.72k
7.75k
| title
stringlengths 1
290
| state
stringclasses 2
values | created_at
timestamp[s]date 2021-07-26 12:21:17
2025-08-23 00:18:43
| updated_at
timestamp[s]date 2021-07-26 13:27:59
2025-08-23 12:34:39
| closed_at
timestamp[s]date 2021-07-26 13:27:59
2025-08-20 16:35:55
⌀ | html_url
stringlengths 49
51
| pull_request
dict | user_login
stringlengths 3
26
| is_pull_request
bool 2
classes | comments
listlengths 0
30
|
|---|---|---|---|---|---|---|---|---|---|---|---|
1,454,647,861
| 5,261
|
Add PubTables-1M
|
open
| 2022-11-18T07:56:36
| 2022-11-18T08:02:18
| null |
https://github.com/huggingface/datasets/issues/5261
| null |
NielsRogge
| false
|
[
"cc @albertvillanova the author would like to add this dataset to the hub: https://github.com/microsoft/table-transformer/issues/68#issuecomment-1319114621. Could you help him out?"
] |
1,453,921,697
| 5,260
|
consumer-finance-complaints dataset not loading
|
open
| 2022-11-17T20:10:26
| 2022-11-18T10:16:53
| null |
https://github.com/huggingface/datasets/issues/5260
| null |
adiprasad
| false
|
[
"Thanks for reporting, @adiprasad.\r\n\r\nWe are having a look at it.",
"I have opened an issue in that dataset Community tab on the Hub: https://huggingface.co/datasets/consumer-finance-complaints/discussions/1\r\n\r\nPlease note that in the meantime, you can load the dataset by passing `ignore_verifications=True`:\r\n```python\r\n>>> ds = load_dataset(\"consumer-finance-complaints\", ignore_verifications=True)\r\n>>> ds\r\nDatasetDict({\r\n train: Dataset({\r\n features: ['Date Received', 'Product', 'Sub Product', 'Issue', 'Sub Issue', 'Complaint Text', 'Company Public Response', 'Company', 'State', 'Zip Code', 'Tags', 'Consumer Consent Provided', 'Submitted via', 'Date Sent To Company', 'Company Response To Consumer', 'Timely Response', 'Consumer Disputed', 'Complaint ID'],\r\n num_rows: 3079747\r\n })\r\n})\r\n```",
"PR fixing this issue: https://huggingface.co/datasets/consumer-finance-complaints/discussions/2"
] |
1,453,555,923
| 5,259
|
datasets 2.7 introduces sharding error
|
closed
| 2022-11-17T15:36:52
| 2022-12-24T01:44:02
| 2022-11-18T12:52:05
|
https://github.com/huggingface/datasets/issues/5259
| null |
DCNemesis
| false
|
[
"I notice a comment in the code says:\r\n`Having lists of different sizes makes sharding ambigious, raise an error in this case until we decide how to define sharding without ambiguity for users` \r\n \r\n ... which suggests this update was pushed knowing that it might break some things. But, it didn't seem to have a useful error message of an argument that could be passed to avoid the error.",
"Sorry for the inconvenience, I opened a PR in your repo to fix this: https://huggingface.co/datasets/sil-ai/bloom-speech/discussions/2\r\n\r\nBasically we've always considered lists in `gen_kwargs` to be a shard list that we can split and pass into different workers to generate the dataset (e.g. if you pass `num_proc=` in `load_dataset()` to generate the dataset in parallel), but it was documented only recently",
"@lhoestq Thanks for the help. It looks like that took care of it."
] |
1,453,516,636
| 5,258
|
Restore order of split names in dataset_info for canonical datasets
|
closed
| 2022-11-17T15:13:15
| 2023-02-16T09:49:05
| 2022-11-19T06:51:37
|
https://github.com/huggingface/datasets/issues/5258
| null |
albertvillanova
| false
|
[
"The bulk edit is running...\r\n\r\nSee for example: \r\n- A single config: https://huggingface.co/datasets/acronym_identification/discussions/2\r\n- Multiple configs: https://huggingface.co/datasets/babi_qa/discussions/1",
"TODO: Add \"dataset_info\" YAML metadata to:\r\n- [x] \"chr_en\" has no metadata JSON file, nor \"dataset_info\" YAML tag in its card\r\n - Fixing PR: https://huggingface.co/datasets/chr_en/discussions/1 \r\n- [x] \"conll2000\" has no metadata JSON file, but it has \"dataset_info\" YAML tag in its card\r\n- [x] \"crime_and_punish\" has no metadata JSON file, but it has \"dataset_info\" YAML tag in its card\r\n- [x] \"dart\" has no metadata JSON file, but it has \"dataset_info\" YAML tag in its card\r\n- [x] \"iwslt2017\" has no metadata JSON file, but it has \"dataset_info\" YAML tag in its card\r\n- [ ] \"mc4\" has no metadata JSON file, nor \"dataset_info\" YAML tag in its card\r\n- [ ] \"the_pile\" has no metadata JSON file, nor \"dataset_info\" YAML tag in its card\r\n- [ ] \"timit_asr\" has no metadata JSON file, nor \"dataset_info\" YAML tag in its card",
"The bulk edit is finished."
] |
1,452,656,891
| 5,257
|
remove an unused statement
|
closed
| 2022-11-17T04:00:50
| 2022-11-18T11:04:08
| 2022-11-18T11:04:08
|
https://github.com/huggingface/datasets/pull/5257
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5257",
"html_url": "https://github.com/huggingface/datasets/pull/5257",
"diff_url": "https://github.com/huggingface/datasets/pull/5257.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5257.patch",
"merged_at": "2022-11-18T11:04:08"
}
|
WrRan
| true
|
[] |
1,452,652,586
| 5,256
|
fix wrong print
|
closed
| 2022-11-17T03:54:26
| 2022-11-18T11:05:32
| 2022-11-18T11:05:32
|
https://github.com/huggingface/datasets/pull/5256
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5256",
"html_url": "https://github.com/huggingface/datasets/pull/5256",
"diff_url": "https://github.com/huggingface/datasets/pull/5256.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5256.patch",
"merged_at": "2022-11-18T11:05:32"
}
|
WrRan
| true
|
[] |
1,452,631,517
| 5,255
|
Add a Depth Estimation dataset - DIODE / NYUDepth / KITTI
|
closed
| 2022-11-17T03:22:22
| 2022-12-17T12:20:38
| 2022-12-17T12:20:37
|
https://github.com/huggingface/datasets/issues/5255
| null |
sayakpaul
| false
|
[
"Also cc @mariosasko and @lhoestq ",
"Cool ! Let us know if you have questions or if we can help :)\r\n\r\nI guess we'll also have to create the NYU CS Department on the Hub ?",
"> I guess we'll also have to create the NYU CS Department on the Hub ?\r\n\r\nYes, you're right! Let me add it to my profile first, and then we can transfer. Meanwhile, if it's recommended to loop the dataset author in here, let me know. \r\n\r\nAlso, the NYU Depth dataset seems big. Any example scripts for creating image datasets that I could refer? ",
"You can check the imagenet-1k one.\r\n\r\nPS: If the licenses allows it, it'b be nice to host the dataset as sharded TAR archives (like imagenet-1k) instead of the ZIP format they use:\r\n- it will make streaming much faster\r\n- ZIP compression is not well suited for images\r\n- it will allow parallel processing of the dataset (you can pass a subset of shards to each worker)\r\n\r\n> if it's recommended to loop the dataset author in here, let me know.\r\n\r\nIt's recommended indeed, you can send them an email once you have the dataset ready and invite them to the org on the Hub",
"> You can check the imagenet-1k one.\r\n\r\nWhere can I find the script? Are you referring to https://huggingface.co/docs/datasets/image_process ? Or is there anything more specific? ",
"You can find it here: https://huggingface.co/datasets/imagenet-1k/blob/main/imagenet-1k.py",
"Update: started working on it here: https://huggingface.co/datasets/sayakpaul/nyu_depth_v2. \r\n\r\nI am facing an issue and I have detailed it here: https://huggingface.co/datasets/sayakpaul/nyu_depth_v2/discussions/1\r\n\r\nEdit: The issue is gone. \r\n\r\nHowever, since the dataset is distributed as a single TAR archive (following the [URL used in TensorFlow Datasets](https://github.com/tensorflow/datasets/tree/master/tensorflow_datasets/datasets/nyu_depth_v2/nyu_depth_v2_dataset_builder.py)) the loading is taking longer. How would suggest to shard the single TAR archive? \r\n\r\n@lhoestq \r\n\r\n",
"A Colab Notebook demonstrating the dataset loading part: \r\n\r\nhttps://colab.research.google.com/gist/sayakpaul/aa0958c8d4ad8518d52a78f28044d871/scratchpad.ipynb\r\n\r\n@osanseviero @lhoestq \r\n\r\nI will work on a notebook to work with the dataset including data visualization.",
"@osanseviero @lhoestq things seem to work fine with the current version of the dataset [here](https://huggingface.co/datasets/sayakpaul/nyu_depth_v2). Here's a notebook I developed to help with visualization: https://colab.research.google.com/drive/1K3ZU8XUPRDOYD38MQS9nreQXJYitlKSW?usp=sharing. \r\n\r\n@lhoestq I need your help with the following:\r\n\r\n> However, since the dataset is distributed as a single TAR archive (following the [URL used in TensorFlow Datasets](https://github.com/tensorflow/datasets/tree/master/tensorflow_datasets/datasets/nyu_depth_v2/nyu_depth_v2_dataset_builder.py)) the loading is taking longer. How would suggest to shard the single TAR archive?\r\n\r\n@osanseviero @lhoestq question for you:\r\n\r\nWhere should we host the dataset? I think hosting it under hf.co/datasets (that is HF is the org) is fine as we have ImageNet-1k hosted similarly. We could then reach out to Diana Wofk (author of [Fast Depth](https://github.com/dwofk/fast-depth) and the owner of the repo on which TFDS NYU Depth V2 is based) for a review. WDYT? ",
"> However, since the dataset is distributed as a single TAR archive (following the [URL used in TensorFlow Datasets](https://github.com/tensorflow/datasets/tree/master/tensorflow_datasets/datasets/nyu_depth_v2/nyu_depth_v2_dataset_builder.py)) the loading is taking longer. How would suggest to shard the single TAR archive?\r\n\r\nFirst you can separate the train data and the validation data.\r\n\r\nThen since the dataset is quite big, you can even shard the train split and the validation split in multiple TAR archives. Something around 16 archives for train and 4 for validation would be fine for example.\r\n\r\nAlso no need to gzip the TAR archives, the images are already compressed in png or jpeg.",
"> Then since the dataset is quite big, you can even shard the train split and the validation split in multiple TAR archives. Something around 16 archives for train and 4 for validation would be fine for example.\r\n\r\nYes, I got you. But this process seems to be manual and should be tailored for the given dataset. Do you have any script that you used to create the ImageNet-1k shards? \r\n\r\n> Also no need to gzip the TAR archives, the images are already compressed in png or jpeg.\r\n\r\nI was not going to do that. Not sure what brought it up. ",
"> Yes, I got you. But this process seems to be manual and should be tailored for the given dataset. Do you have any script that you used to create the ImageNet-1k shards?\r\n\r\nI don't, but I agree it'd be nice to have a script for that !\r\n\r\n> I was not going to do that. Not sure what brought it up.\r\n\r\nThe original dataset is gzipped for some reason",
"Oh, I am using this URL for the download: https://github.com/tensorflow/datasets/blob/master/tensorflow_datasets/datasets/nyu_depth_v2/nyu_depth_v2_dataset_builder.py#L24. ",
"> Where should we host the dataset? I think hosting it under hf.co/datasets (that is HF is the org) is fine as we have ImageNet-1k hosted similarly.\r\n\r\nMaybe you can create an org for NYU Courant (this is the institute of the lab of the main author of the dataset if I'm not mistaken), and invite the authors to join.\r\n\r\nWe don't add datasets without namespace anymore",
"Updates: https://huggingface.co/datasets/sayakpaul/nyu_depth_v2/discussions/5\r\n\r\nThe entire process (preparing multiple archives, preparing data loading script, etc.) was fun and engaging, thanks to the documentation. I believe we could work on a small blog post that would work as a reference for the future contributors following this path. What say? \r\n\r\nCc: @lhoestq @osanseviero ",
"> I believe we could work on a small blog post that would work as a reference for the future contributors following this path. What say?\r\n\r\n@polinaeterna already mentioned it would be nice to present this process for audio (it's exactly the same), I believe it can be useful to many people",
"Cool. Let's work on that after the NYU Depth Dataset is fully in on Hub (under the appropriate org). 🤗",
"@lhoestq need to discuss something while I am adding the dataset card to https://huggingface.co/datasets/sayakpaul/nyu_depth_v2/. \r\n\r\nAs per [Papers With Code](https://paperswithcode.com/dataset/nyuv2), NYU Depth v2 is used for many different tasks:\r\n\r\n* Monocular depth estimation\r\n* Depth estimation \r\n* Semantic segmentation\r\n* Plane instance segmentation \r\n* ...\r\n\r\nSo, while writing the supported task part of the dataset card, should we focus on all these? IMO, we could focus on just depth estimation and semantic segmentation for now since we have supported models for these two. WDYT?\r\n\r\nAlso, I am getting: \r\n\r\n\r\n```\r\nremote: Your push was accepted, but with warnings:\r\nremote: - Warning: The task_ids \"depth-estimation\" is not in the official list: acceptability-classification, entity-linking-classification, fact-checking, intent-classification, multi-class-classification, multi-label-classification, multi-input-text-classification, natural-language-inference, semantic-similarity-classification, sentiment-classification, topic-classification, semantic-similarity-scoring, sentiment-scoring, sentiment-analysis, hate-speech-detection, text-scoring, named-entity-recognition, part-of-speech, parsing, lemmatization, word-sense-disambiguation, coreference-resolution, extractive-qa, open-domain-qa, closed-domain-qa, news-articles-summarization, news-articles-headline-generation, dialogue-generation, dialogue-modeling, language-modeling, text-simplification, explanation-generation, abstractive-qa, open-domain-abstractive-qa, closed-domain-qa, open-book-qa, closed-book-qa, slot-filling, masked-language-modeling, keyword-spotting, speaker-identification, audio-intent-classification, audio-emotion-recognition, audio-language-identification, multi-label-image-classification, multi-class-image-classification, face-detection, vehicle-detection, instance-segmentation, semantic-segmentation, panoptic-segmentation, image-captioning, grasping, task-planning, tabular-multi-class-classification, tabular-multi-label-classification, tabular-single-column-regression, rdf-to-text, multiple-choice-qa, multiple-choice-coreference-resolution, document-retrieval, utterance-retrieval, entity-linking-retrieval, fact-checking-retrieval, univariate-time-series-forecasting, multivariate-time-series-forecasting, visual-question-answering, document-question-answering\r\nremote: ----------------------------------------------------------\r\nremote: Please find the documentation at:\r\nremote: https://huggingface.co/docs/hub/model-cards#model-card-metadata\r\n```\r\n\r\nWhat should be the plan of action for this?\r\n\r\nCc: @osanseviero \r\n\r\n",
"> What should be the plan of action for this?\r\n\r\nWhen you merged https://github.com/huggingface/hub-docs/pull/488, there is a JS Interfaces GitHub Actions workflow that runs https://github.com/huggingface/hub-docs/actions/workflows/js-interfaces-tests.yml. It has a step called [export-task scripts](https://github.com/huggingface/hub-docs/actions/runs/3622479064/jobs/6107238948) which exports an interface you can use in `dataset`. If you look at the logs, it prints out a map. This map can replace https://github.com/huggingface/datasets/blob/main/src/datasets/utils/resources/tasks.json (tasks.json was generated with this script), which should add depth estimation\r\n",
"Thanks @osanseviero. \r\n\r\nhttps://github.com/huggingface/datasets/pull/5335",
"Closing the issue as the dataset has been successfully added: https://huggingface.co/datasets/sayakpaul/nyu_depth_v2"
] |
1,452,600,088
| 5,254
|
typo
|
closed
| 2022-11-17T02:39:57
| 2022-11-18T10:53:45
| 2022-11-18T10:53:45
|
https://github.com/huggingface/datasets/pull/5254
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5254",
"html_url": "https://github.com/huggingface/datasets/pull/5254",
"diff_url": "https://github.com/huggingface/datasets/pull/5254.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5254.patch",
"merged_at": "2022-11-18T10:53:45"
}
|
WrRan
| true
|
[] |
1,452,588,206
| 5,253
|
typo
|
closed
| 2022-11-17T02:22:58
| 2022-11-18T10:53:11
| 2022-11-18T10:53:10
|
https://github.com/huggingface/datasets/pull/5253
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5253",
"html_url": "https://github.com/huggingface/datasets/pull/5253",
"diff_url": "https://github.com/huggingface/datasets/pull/5253.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5253.patch",
"merged_at": "2022-11-18T10:53:10"
}
|
WrRan
| true
|
[] |
1,451,765,838
| 5,252
|
Support for decoding Image/Audio types in map when format type is not default one
|
closed
| 2022-11-16T15:02:13
| 2022-12-13T17:01:54
| 2022-12-13T16:59:04
|
https://github.com/huggingface/datasets/pull/5252
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5252",
"html_url": "https://github.com/huggingface/datasets/pull/5252",
"diff_url": "https://github.com/huggingface/datasets/pull/5252.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5252.patch",
"merged_at": "2022-12-13T16:59:04"
}
|
mariosasko
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5252). All of your documentation changes will be reflected on that endpoint.",
"Yes, if the image column is the first in the batch keys, it will decode the images because it reads the actual values. We could avoid this by checking the batch type, and if it's `LazyDict`, `num_examples` is equal to `len(batch.pa_table)`, which doesn't lead to decoding.",
"Good idea. This can be done in a subsequent PR btw, since it's out of scope of the original goal of this PR",
"Just fixed a small bug where it would show the pyarrow 10 warning about None -> empty lists conversions even with an Array2D with no nulls",
"Fixed another bug when your map function returns a mix of LazyDict or regular dict and added some tests"
] |
1,451,761,321
| 5,251
|
Docs are not generated after latest release
|
closed
| 2022-11-16T14:59:31
| 2022-11-22T16:27:50
| 2022-11-22T16:27:50
|
https://github.com/huggingface/datasets/issues/5251
| null |
albertvillanova
| false
|
[
"After a discussion with @mishig25:\r\n- He said that this action should be triggered if we call our release branch according to the regex `v*-release`, as transformers does\r\n- I said that our procedure is different: our release branch is *temporary* and it is deleted just after the release PR is merged to main\r\n - Indeed the release tag is not yet created when we make the release PR (not event when this is merged to main), but when we make the Release itself.\r\n\r\nI was thinking that maybe we could change the triggering event: use `release` instead of `push`.\r\n\r\nWhat do you think, @huggingface/datasets?",
"Why is it an issue if our branch is temporary ?",
"He says not; but the branch has no tag yet; does the doc building require the tag? Or just the version number in `__init__.py` or setup.py?",
"It uses `module.__version__` (i.e. the one defined in `__init__.py`) - no need to have a tag\r\n\r\nhttps://github.com/huggingface/doc-builder/blob/81575cf081964c30ea5fd39450f4820db963f18e/src/doc_builder/commands/build.py#L69",
"Thanks, @lhoestq.\r\n\r\n@mishig25 has manually forced the generation of the docs, that are live for 2.7.0 version: https://huggingface.co/docs/datasets/v2.7.0/en/index ",
"Cool ! this can be closed then ?",
"I was waiting for #5250 to be merged to close this.",
"just to confirm, is there anything I need to do from my side ? Or is everything good here ?"
] |
1,451,720,030
| 5,250
|
Change release procedure to use only pull requests
|
closed
| 2022-11-16T14:35:32
| 2022-11-22T16:30:58
| 2022-11-22T16:27:48
|
https://github.com/huggingface/datasets/pull/5250
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5250",
"html_url": "https://github.com/huggingface/datasets/pull/5250",
"diff_url": "https://github.com/huggingface/datasets/pull/5250.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5250.patch",
"merged_at": "2022-11-22T16:27:48"
}
|
albertvillanova
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5250). All of your documentation changes will be reflected on that endpoint.",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5250). All of your documentation changes will be reflected on that endpoint.",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5250). All of your documentation changes will be reflected on that endpoint.",
"Little recap:\r\n- The release-conda GH action was properly triggered by push-tag event: therefore I guess this event is also created when we publish a release and create a tag within it (as it is the case in the new procedure)\r\n - However, the package was only uploaded to huggingface channel and not to conda-forge channel\r\n - [x] Why? Need to address this.\r\n - Reply by @lhoestq: https://github.com/huggingface/datasets/pull/5250#discussion_r1025047531\r\n - We only maintain the huggingface channel\r\n - The conda-forge channel is maintained by the community; the 2.7.0 has been finally added as well to this channel \r\n- The generate-documentation GH action will be triggered by the push-to-branch event if we align the name of the release branch with the expected regex `v*-release`\r\n - [x] The naming has been aligned in the new procedure\r\n - [ ] Question: why do we have different triggering events for generate-doc and release-conda? Maybe we could set the same for both: either push-tag (when publishing the release), or push-to-branch\r\n - I think it will be better to use the push-tag event because in the new release procedure this happens later (when we publish the release), once we have already tested that everything works using the test-PyPI; on the contrary, the push-to-branch event happens before, even before opening the release PR: we could see afterwards that there is an issue, and cancel the Pull Request, but the docs and conda-package will already be published.\r\n- For the naming of the dev-version branch/PR, instead of having a complicated version naming, I'm proposing:\r\n - Using always the same branch name `dev-version`\r\n - Just include a step to delete this branch locally if it exists: `git branch -D dev-version`\r\n - The remote version will not exist because it is deleted once the PR is merged\r\n - This approach is approved by @lhoestq: https://github.com/huggingface/datasets/pull/5250#discussion_r1025048300",
"Just one question to be addressed: why do we have different triggering events for generate-doc and release-conda? Maybe we could set the same for both: either push-tag (when publishing the release), or push-to-branch\r\n\r\nI think it will be better to use the push-tag event because in the new release procedure this happens later (when we publish the release), once we have already tested that everything works using the test-PyPI; on the contrary, the push-to-branch event happens before, even before opening the release PR: we could see afterwards that there is an issue, and cancel the Pull Request, but the docs and conda-package will already be published.\r\n\r\nWe could even use the release-published event instead: [8694901](https://github.com/huggingface/datasets/pull/5250/commits/86949013c9dc59a07b55fad5b78104b8a03f60cd)\r\n",
"@lhoestq now that we have push-tag event for both build_documentation and release-conda, we have no constraint on the naming of the release branch:\r\n- we could name it simpler: maybe as you suggested above: https://github.com/huggingface/datasets/pull/5250#discussion_r1024119018\r\n `release-VERSION` instead of `vVERSION-release` (we do not use the prefix \"v\" anywhere in our repo)"
] |
1,451,692,247
| 5,249
|
Protect the main branch from inadvertent direct pushes
|
closed
| 2022-11-16T14:19:03
| 2023-12-21T10:28:27
| 2023-12-21T10:28:26
|
https://github.com/huggingface/datasets/issues/5249
| null |
albertvillanova
| false
|
[
"It seems all the tasks have been addressed, meaning this issue can be closed, no?"
] |
1,451,338,676
| 5,248
|
Complete doc migration
|
closed
| 2022-11-16T10:41:04
| 2022-11-16T15:06:50
| 2022-11-16T10:41:10
|
https://github.com/huggingface/datasets/pull/5248
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5248",
"html_url": "https://github.com/huggingface/datasets/pull/5248",
"diff_url": "https://github.com/huggingface/datasets/pull/5248.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5248.patch",
"merged_at": "2022-11-16T10:41:10"
}
|
mishig25
| true
|
[
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5248). All of your documentation changes will be reflected on that endpoint.",
"Thanks for the fix @mishig25.\r\n\r\nI guess this is the reason why the docs are not generated for the latest release version 2.7.0? https://huggingface.co/docs/datasets/index "
] |
1,451,297,749
| 5,247
|
Set dev version
|
closed
| 2022-11-16T10:17:31
| 2022-11-16T10:22:20
| 2022-11-16T10:17:50
|
https://github.com/huggingface/datasets/pull/5247
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5247",
"html_url": "https://github.com/huggingface/datasets/pull/5247",
"diff_url": "https://github.com/huggingface/datasets/pull/5247.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5247.patch",
"merged_at": "2022-11-16T10:17:50"
}
|
albertvillanova
| true
|
[
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5247). All of your documentation changes will be reflected on that endpoint."
] |
1,451,226,055
| 5,246
|
Release: 2.7.0
|
closed
| 2022-11-16T09:32:44
| 2022-11-16T09:39:42
| 2022-11-16T09:37:03
|
https://github.com/huggingface/datasets/pull/5246
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5246",
"html_url": "https://github.com/huggingface/datasets/pull/5246",
"diff_url": "https://github.com/huggingface/datasets/pull/5246.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5246.patch",
"merged_at": "2022-11-16T09:37:03"
}
|
albertvillanova
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,450,376,433
| 5,245
|
Unable to rename columns in streaming dataset
|
closed
| 2022-11-15T21:04:41
| 2022-11-28T12:53:24
| 2022-11-28T12:53:24
|
https://github.com/huggingface/datasets/issues/5245
| null |
peregilk
| false
|
[
"Hi @peregilk this bug is directly related to https://github.com/huggingface/datasets/issues/3888, and still not fixed... But I'll try to have a look!",
"Thanks @alvarobartt. It is great if you are able to fix it, but when reading the explanation it seems like it is possible to work around it.\r\n\r\nWe also tried keeping the 'info.features' and then adding a modified version back after the remove/rename. Unforutunately that leads to a dataset that is not possible to iterate over.",
"So if you iterate over the `IterableDataset` as `next(iter(ds))` and then run `rename_columns` when checking that data it will work, but in the end, it's just renaming the column one example/batch at a time, not renaming the column name for all the entries in the dataset, which is the ideal.",
"@alvarobartt Thanks. My use case was that I wanted to do multiple things, ie removing all unnecessary columns, renaming some valid columns, and then using cast (in my case checking if the audio is not 16K and casting it). It is just convenient to look into the info.features between each of these operations. Alternatively, I will just plan ahead...;) To me it seems like all the operations are working.\r\n\r\nThanks for the advice. It was very useful.",
"If we know the features before renaming, then we know the features after renaming, so we can pass the new features to the returned dataset in `rename_column` indeed ! If anyone is interested in contributing, feel free to open a PR and I'd be happy to help / give some pointers :)",
"Sure @lhoestq thanks! I’ll try to work on that",
"#self-assign"
] |
1,450,019,225
| 5,244
|
Allow dataset streaming from private a private source when loading a dataset with a dataset loading script
|
open
| 2022-11-15T16:02:10
| 2022-11-23T14:02:30
| null |
https://github.com/huggingface/datasets/issues/5244
| null |
bruno-hays
| false
|
[
"Hi ! What kind of private source ? We're exploring adding support for cloud storage and URIs like s3://, gs:// etc. with authentication in the download manager",
"Hello! It's a google cloud storage, so gs://, but I'm using it with https.\r\nBeing able to provide a file system like [here](https://huggingface.co/docs/datasets/main/filesystems#load-serialized-datasets) would be even more practical indeed.\r\nI've found a quite complicated workaround which consists of monkey patching all of the functions in streaming_download_manager.py to use my own _get_authentication_headers_for_url_ . \r\n\r\nA support for this use case would be greatly appreciated!\r\n\r\nFor reference my _get_authentication_headers_for_url_ looks like this:\r\n```\r\nimport os\r\nfrom typing import Optional, Union\r\n\r\nfrom datasets import config\r\nfrom huggingface_hub import HfFolder\r\nfrom gcsfs.credentials import GoogleCredentials\r\n\r\nDEFAULT_PROJECT = os.environ.get(\"GCSFS_DEFAULT_PROJECT\", \"\")\r\naccess = \"full_control\"\r\ngcs_token = os.environ.get(\"GCS_TOKEN\")\r\n\r\n\r\ndef get_authentication_headers_for_url(url: str, use_auth_token: Optional[Union[str, bool]] = None) -> dict:\r\n \"\"\"Handle the HF authentication\"\"\"\r\n headers = {}\r\n if url.startswith(config.HF_ENDPOINT):\r\n if use_auth_token is False:\r\n token = None\r\n elif isinstance(use_auth_token, str):\r\n token = use_auth_token\r\n else:\r\n token = HfFolder.get_token()\r\n elif url.startswith(\"https://storage.googleapis.com\"):\r\n credentials = GoogleCredentials(DEFAULT_PROJECT, access, gcs_token)\r\n credentials.maybe_refresh()\r\n token = credentials.credentials.token\r\n else:\r\n token = None\r\n if token:\r\n headers[\"authorization\"] = f\"Bearer {token}\"\r\n return headers\r\n```",
"I would be a big fan of this feature! @Hubert-Bonisseur if this doesn't become a supported feature, would you mind sharing your code? Thanks!",
"> I would be a big fan of this feature! @Hubert-Bonisseur if this doesn't become a supported feature, would you mind sharing your code? Thanks!\r\n\r\nI published it here:\r\nhttps://github.com/Hubert-Bonisseur/private-dataset-hub\r\n\r\nI modified the names of a lot of functions for privacy and I don't have time to test it again so you may get import errors, but you have the code. The custom_load_dataset is the function you are interested in I think.\r\n\r\nIt relies a lot on patching, if you find a better way to do this, I'd be interested.",
"Given the amount of patching it does, this is likely to break at one point. I'd encourage you to wait for a proper support in `datasets` directly if you can wait."
] |
1,449,523,962
| 5,243
|
Download only split data
|
open
| 2022-11-15T10:15:54
| 2025-02-25T14:47:03
| null |
https://github.com/huggingface/datasets/issues/5243
| null |
capsabogdan
| false
|
[
"Hi @capsabogdan! Unfortunately, it's hard to implement because quite often datasets data is being hosted in a single archive for all splits :( So we have to download the whole archive to split it into splits. This is the case for CommonVoice too. \r\n\r\nHowever, for cases when data is distributed in separate archives ащк different splits I suppose it can (and will) be implemented someday. \r\n\r\n\r\nBtw for quick check of the dataset you can use [streaming](https://huggingface.co/docs/datasets/stream):\r\n```python\r\ncv = load_dataset(\"mozilla-foundation/common_voice_11_0\", \"en\", split=\"test\", streaming=True)\r\ncv = iter(cv)\r\nprint(next(cv))\r\n\r\n>> {'client_id': 'a07b17f8234ded5e847443ea6f423cef745cbbc7537fb637d58326000aa751e829a21c4fd0a35fc17fb833aa7e95ebafce5efd19beeb8d843887b85e4eb35f5b',\r\n>> 'path': None,\r\n>> 'audio': {'path': 'cv-corpus-11.0-2022-09-21/en/clips/common_voice_en_100363.mp3',\r\n>> 'array': array([ 0.0000000e+00, 1.1748125e-14, 1.5450088e-14, ...,\r\n>> 1.3011958e-06, -6.3548953e-08, -9.9098514e-08], dtype=float32),\r\n>> ...}\r\n\r\n```",
"thank you for the answer but am not sure if this will not be helpful, as we\nneed maybe just 10% of the datasets for some experiment\n\ncan we get just a portion of the dataset with stream?\n\n\nis there really no solution? :(\n\nAm Di., 15. Nov. 2022 um 16:55 Uhr schrieb Polina Kazakova <\n***@***.***>:\n\n> Hi @capsabogdan <https://github.com/capsabogdan>! Unfortunately, it's\n> hard to implement because quite often datasets data is being hosted in a\n> single archive for all splits :( So we have to download the whole archive\n> to split it into splits. This is the case for CommonVoice too.\n>\n> However, for cases when data is distributed in separate archives in\n> different splits I suppose it can be implemented someday.\n>\n> Btw for quick check of the dataset you can use streaming\n> <https://huggingface.co/docs/datasets/stream>:\n>\n> cv = load_dataset(\"mozilla-foundation/common_voice_11_0\", \"en\", split=\"test\", streaming=True)cv = iter(cv)print(next(cv))\n> >> {'client_id': 'a07b17f8234ded5e847443ea6f423cef745cbbc7537fb637d58326000aa751e829a21c4fd0a35fc17fb833aa7e95ebafce5efd19beeb8d843887b85e4eb35f5b',>> 'path': None,>> 'audio': {'path': 'cv-corpus-11.0-2022-09-21/en/clips/common_voice_en_100363.mp3',>> 'array': array([ 0.0000000e+00, 1.1748125e-14, 1.5450088e-14, ...,>> 1.3011958e-06, -6.3548953e-08, -9.9098514e-08], dtype=float32),>> ...}\n>\n> —\n> Reply to this email directly, view it on GitHub\n> <https://github.com/huggingface/datasets/issues/5243#issuecomment-1315512887>,\n> or unsubscribe\n> <https://github.com/notifications/unsubscribe-auth/ALSIFOC3JYRCTH54OBRUJULWIOW6PANCNFSM6AAAAAASAYO2LY>\n> .\n> You are receiving this because you were mentioned.Message ID:\n> ***@***.***>\n>\n",
"maybe it would be nice if you guys ould do some sort of shard before\nloading the dataset, so users can download just chunks of data :)\n\nI think this would be very helpful\n\nAm Di., 15. Nov. 2022 um 19:24 Uhr schrieb Bogdan Capsa <\n***@***.***>:\n\n> thank you for the answer but am not sure if this will not be helpful, as\n> we need maybe just 10% of the datasets for some experiment\n>\n> can we get just a portion of the dataset with stream?\n>\n>\n> is there really no solution? :(\n>\n> Am Di., 15. Nov. 2022 um 16:55 Uhr schrieb Polina Kazakova <\n> ***@***.***>:\n>\n>> Hi @capsabogdan <https://github.com/capsabogdan>! Unfortunately, it's\n>> hard to implement because quite often datasets data is being hosted in a\n>> single archive for all splits :( So we have to download the whole archive\n>> to split it into splits. This is the case for CommonVoice too.\n>>\n>> However, for cases when data is distributed in separate archives in\n>> different splits I suppose it can be implemented someday.\n>>\n>> Btw for quick check of the dataset you can use streaming\n>> <https://huggingface.co/docs/datasets/stream>:\n>>\n>> cv = load_dataset(\"mozilla-foundation/common_voice_11_0\", \"en\", split=\"test\", streaming=True)cv = iter(cv)print(next(cv))\n>> >> {'client_id': 'a07b17f8234ded5e847443ea6f423cef745cbbc7537fb637d58326000aa751e829a21c4fd0a35fc17fb833aa7e95ebafce5efd19beeb8d843887b85e4eb35f5b',>> 'path': None,>> 'audio': {'path': 'cv-corpus-11.0-2022-09-21/en/clips/common_voice_en_100363.mp3',>> 'array': array([ 0.0000000e+00, 1.1748125e-14, 1.5450088e-14, ...,>> 1.3011958e-06, -6.3548953e-08, -9.9098514e-08], dtype=float32),>> ...}\n>>\n>> —\n>> Reply to this email directly, view it on GitHub\n>> <https://github.com/huggingface/datasets/issues/5243#issuecomment-1315512887>,\n>> or unsubscribe\n>> <https://github.com/notifications/unsubscribe-auth/ALSIFOC3JYRCTH54OBRUJULWIOW6PANCNFSM6AAAAAASAYO2LY>\n>> .\n>> You are receiving this because you were mentioned.Message ID:\n>> ***@***.***>\n>>\n>\n",
"+1 on this feature request - I am running into the same problem, where I only need the test set for a dataset that has a huge training set",
"Hey, I'm also interested in that as a feature. I'm having the same problem with Common Voice 13.0. The dataset is super big but I only want the test data to benchmark multilingual models, but I don't have much Terabytes to store all the dataset...",
"Consider this approach: Download and save individual audio files by streaming each split, then compile a CSV file that contains the file names and corresponding text.\r\n\r\n```python3\r\nimport os\r\nimport shutil\r\nfrom pathlib import Path\r\n\r\nimport datasets\r\nimport pandas as pd\r\nimport soundfile\r\nfrom datasets import Dataset, concatenate_datasets, load_dataset\r\n\r\n\r\ndataset = load_dataset(\"librispeech_asr\", 'clean', split=\"train.100\", streaming=True)\r\ndataset = iter(dataset)\r\n\r\ndownload_path = os.path.join(os.getcwd(), 'librispeech', 'clips')\r\ncsv_name = os.path.join(os.getcwd(), 'librispeech', 'clean_train_100.csv')\r\n\r\nrows = []\r\nfor i, row in enumerate(dataset):\r\n print(i)\r\n path = os.path.join(download_path, row['audio']['path'])\r\n soundfile.write(path, row['audio']['array'], row['audio']['sampling_rate'])\r\n\r\n del row['audio']\r\n rows.append(row)\r\n\r\ndf = pd.DataFrame(rows)\r\ndf.to_csv(csv_name, index=False, header=True)\r\n```",
"Faced this issue as well so wrote a short script that pulls a hub dataset, creates a small sample of it, and pushes the sample data to the hub as a new dataset.\n\n```python\ndef create_sample_dataset(full_dataset_name, sample_count=100, username=\"my-username\", cache_dir=\"./dataset\"):\n # Create a directory to save the sampled dataset\n os.makedirs(cache_dir, exist_ok=True)\n\n # Get the dataset name\n dataset_name = full_dataset_name.split(\"/\")[-1]\n dataset_name_sample = f\"{dataset_name}-sample-{sample_count}\"\n\n # Load the dataset\n dataset = datasets.load_dataset(full_dataset_name, cache_dir=cache_dir)\n\n # Sample 100 rows from the training split (or modify for other splits)\n train_sample = dataset[\"train\"].shuffle(seed=42).select(range(sample_count))\n test_sample = dataset[\"test\"].shuffle(seed=42).select(range(sample_count))\n\n # Push to hub\n train_sample.push_to_hub(dataset_name_sample, split=\"train\")\n print(\"INFO: Train split pushed to the hub successfully\")\n\n test_sample.push_to_hub(dataset_name_sample, split=\"test\")\n print(\"INFO: Test split pushed to the hub successfully\")\n```\n\nOnce sampled / pushed you have a smaller version of your dataset in the hub to pull from.\n\nThe [full gist is here](https://gist.github.com/neonwatty/2205277501928f5945726e9e90950ae9)."
] |
1,449,069,382
| 5,242
|
Failed Data Processing upon upload with zip file full of images
|
open
| 2022-11-15T02:47:52
| 2022-11-15T17:59:23
| null |
https://github.com/huggingface/datasets/issues/5242
| null |
scrambled2
| false
|
[
"cc @abhishekkrthakur @SBrandeis "
] |
1,448,510,407
| 5,241
|
Support hfh rc version
|
closed
| 2022-11-14T18:05:47
| 2022-11-15T16:11:30
| 2022-11-15T16:09:31
|
https://github.com/huggingface/datasets/pull/5241
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5241",
"html_url": "https://github.com/huggingface/datasets/pull/5241",
"diff_url": "https://github.com/huggingface/datasets/pull/5241.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5241.patch",
"merged_at": "2022-11-15T16:09:31"
}
|
lhoestq
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,448,478,617
| 5,240
|
Cleaner error tracebacks for dataset script errors
|
closed
| 2022-11-14T17:42:02
| 2022-11-15T18:26:48
| 2022-11-15T18:24:38
|
https://github.com/huggingface/datasets/pull/5240
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5240",
"html_url": "https://github.com/huggingface/datasets/pull/5240",
"diff_url": "https://github.com/huggingface/datasets/pull/5240.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5240.patch",
"merged_at": "2022-11-15T18:24:38"
}
|
mariosasko
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"@lhoestq Good catch! This currently leads to an AttributeError (due to `writer` being None) on this line:\r\nhttps://github.com/huggingface/datasets/blob/fed1628d49a91f9ae259ddf6edbb252c7972d9a3/src/datasets/builder.py#L1552\r\n"
] |
1,448,211,373
| 5,239
|
Add num_proc to from_csv/generator/json/parquet/text
|
closed
| 2022-11-14T14:53:00
| 2022-12-06T15:39:10
| 2022-12-06T15:39:09
|
https://github.com/huggingface/datasets/pull/5239
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5239",
"html_url": "https://github.com/huggingface/datasets/pull/5239",
"diff_url": "https://github.com/huggingface/datasets/pull/5239.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5239.patch",
"merged_at": "2022-12-06T15:39:09"
}
|
lhoestq
| true
|
[
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5239). All of your documentation changes will be reflected on that endpoint.",
"I ended up moving `num_proc` to `AbstractDatasetReader.__init__` :)\r\n\r\nLet me know if it sounds good to you now"
] |
1,448,211,251
| 5,238
|
Make `Version` hashable
|
closed
| 2022-11-14T14:52:55
| 2022-11-14T15:30:02
| 2022-11-14T15:27:35
|
https://github.com/huggingface/datasets/pull/5238
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5238",
"html_url": "https://github.com/huggingface/datasets/pull/5238",
"diff_url": "https://github.com/huggingface/datasets/pull/5238.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5238.patch",
"merged_at": "2022-11-14T15:27:35"
}
|
mariosasko
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,448,202,491
| 5,237
|
Encode path only for old versions of hfh
|
closed
| 2022-11-14T14:46:57
| 2022-11-14T17:38:18
| 2022-11-14T17:35:59
|
https://github.com/huggingface/datasets/pull/5237
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5237",
"html_url": "https://github.com/huggingface/datasets/pull/5237",
"diff_url": "https://github.com/huggingface/datasets/pull/5237.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5237.patch",
"merged_at": "2022-11-14T17:35:59"
}
|
lhoestq
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,448,190,801
| 5,236
|
Handle ArrowNotImplementedError caused by try_type being Image or Audio in cast
|
closed
| 2022-11-14T14:38:59
| 2022-11-14T16:04:29
| 2022-11-14T16:01:48
|
https://github.com/huggingface/datasets/pull/5236
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5236",
"html_url": "https://github.com/huggingface/datasets/pull/5236",
"diff_url": "https://github.com/huggingface/datasets/pull/5236.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5236.patch",
"merged_at": "2022-11-14T16:01:48"
}
|
mariosasko
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"> Not sure how we can have a test that is relevant for this though - feel free to add one if you have ideas\r\n\r\nYes, this was my reasoning for not adding a test. This change is pretty simple, so I think it's OK not to have a test for it."
] |
1,448,052,660
| 5,235
|
Pin `typer` version in tests to <0.5 to fix Windows CI
|
closed
| 2022-11-14T13:17:02
| 2022-11-14T15:43:01
| 2022-11-14T13:41:12
|
https://github.com/huggingface/datasets/pull/5235
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5235",
"html_url": "https://github.com/huggingface/datasets/pull/5235",
"diff_url": "https://github.com/huggingface/datasets/pull/5235.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5235.patch",
"merged_at": "2022-11-14T13:41:12"
}
|
polinaeterna
| true
|
[] |
1,447,999,062
| 5,234
|
fix: dataset path should be absolute
|
closed
| 2022-11-14T12:47:40
| 2022-12-07T23:49:22
| 2022-12-07T23:46:34
|
https://github.com/huggingface/datasets/pull/5234
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5234",
"html_url": "https://github.com/huggingface/datasets/pull/5234",
"diff_url": "https://github.com/huggingface/datasets/pull/5234.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5234.patch",
"merged_at": "2022-12-07T23:46:34"
}
|
vigsterkr
| true
|
[
"Good catch thanks ! Have you tried to use the absolue path in `MemoryMappedTable.__init__` in `table.py`?\r\n\r\nI think it can fix issues with relative paths at more levels than just fixing it `load_from_disk`. If it works I think it would be a more robust fix to this issue",
"@lhoestq right, that actually fixed it indeed. I've pushed the changes (one-liner). lemme know if there's anything else you need for this fix",
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,447,906,868
| 5,233
|
Fix shards in IterableDataset.from_generator
|
closed
| 2022-11-14T11:42:09
| 2022-11-14T14:16:03
| 2022-11-14T14:13:22
|
https://github.com/huggingface/datasets/pull/5233
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5233",
"html_url": "https://github.com/huggingface/datasets/pull/5233",
"diff_url": "https://github.com/huggingface/datasets/pull/5233.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5233.patch",
"merged_at": "2022-11-14T14:13:22"
}
|
lhoestq
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,446,294,165
| 5,232
|
Incompatible dill versions in datasets 2.6.1
|
closed
| 2022-11-12T06:46:23
| 2022-11-14T08:24:43
| 2022-11-14T08:07:59
|
https://github.com/huggingface/datasets/issues/5232
| null |
vinaykakade
| false
|
[
"Thanks for reporting, @vinaykakade.\r\n\r\nWe are discussing about making a release early this week.\r\n\r\nPlease note that in the meantime, in your specific case (as we also pointed out here: https://github.com/huggingface/datasets/issues/5162#issuecomment-1291720293), you can circumvent the issue by pinning `multiprocess` to 0.70.13 version (instead of using latest 0.70.14).\r\n\r\nDuplicate of:\r\n- https://github.com/huggingface/datasets/issues/5162",
"You can also make `pip-compile` work by using the backtracking resolver (instead of the legacy one): https://pip-tools.readthedocs.io/en/latest/#a-note-on-resolvers\r\n```\r\npip-compile --resolver=backtracking requirements.in\r\n```\r\nThis resolver will automatically use `multiprocess` 0.70.13 version.\r\n"
] |
1,445,883,267
| 5,231
|
Using `set_format(type='torch', columns=columns)` makes Array2D/3D columns stop formatting correctly
|
closed
| 2022-11-11T18:54:36
| 2022-11-11T20:42:29
| 2022-11-11T18:59:50
|
https://github.com/huggingface/datasets/issues/5231
| null |
plamb-viso
| false
|
[
"In case others find this, the problem was not with set_format, but my usages of `to_pandas()` and `from_pandas()` which I was using during dataset splitting; somewhere in the chain of converting to and from pandas the `Array2D/Array3D` types get converted to series of `Sequence()` types"
] |
1,445,507,580
| 5,230
|
dataclasses error when importing the library in python 3.11
|
closed
| 2022-11-11T13:53:49
| 2023-05-25T04:37:05
| 2022-11-14T15:27:37
|
https://github.com/huggingface/datasets/issues/5230
| null |
yonikremer
| false
|
[
"I opened [this issue](https://github.com/python/cpython/issues/99401).\r\nPython's maintainers say that the issue is caused by [this change](https://docs.python.org/3.11/whatsnew/3.11.html#dataclasses).\r\nI believe adding a `__hash__` method to `datasets.utils.version.Version` should solve (at least partially) this issue.",
"Has this been fixed? I am running into this issue now. \r\n\r\nIf this has been fixed, could have a new release with this?\r\n",
"Hi, I am getting error while training \r\n\r\n(tensorflow) C:\\tensorflow\\models\\research\\object_detection>python train.py --logtostderr --train_dir=training/ --pipeline_config_path=training/faster_rcnn_inception_v2_pets.config\r\nTraceback (most recent call last):\r\n File \"C:\\tensorflow\\models\\research\\object_detection\\train.py\", line 54, in <module>\r\n from object_detection.legacy import trainer\r\n File \"C:\\tensorflow\\models\\research\\object_detection\\legacy\\trainer.py\", line 27, in <module>\r\n from object_detection.builders import optimizer_builder\r\n File \"C:\\tensorflow\\models\\research\\object_detection\\builders\\optimizer_builder.py\", line 25, in <module>\r\n from official.modeling.optimization import ema_optimizer\r\n File \"C:\\tensorflow\\models\\official\\modeling\\optimization\\__init__.py\", line 19, in <module>\r\n from official.modeling.optimization.configs.optimization_config import *\r\n File \"C:\\tensorflow\\models\\official\\modeling\\optimization\\configs\\optimization_config.py\", line 31, in <module>\r\n @dataclasses.dataclass\r\n ^^^^^^^^^^^^^^^^^^^^^\r\n File \"C:\\Users\\x0133252\\AppData\\Local\\anaconda3\\envs\\tensorflow\\Lib\\dataclasses.py\", line 1223, in dataclass\r\n return wrap(cls)\r\n ^^^^^^^^^\r\n File \"C:\\Users\\x0133252\\AppData\\Local\\anaconda3\\envs\\tensorflow\\Lib\\dataclasses.py\", line 1213, in wrap\r\n return _process_class(cls, init, repr, eq, order, unsafe_hash,\r\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n File \"C:\\Users\\x0133252\\AppData\\Local\\anaconda3\\envs\\tensorflow\\Lib\\dataclasses.py\", line 958, in _process_class\r\n cls_fields.append(_get_field(cls, name, type, kw_only))\r\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n File \"C:\\Users\\x0133252\\AppData\\Local\\anaconda3\\envs\\tensorflow\\Lib\\dataclasses.py\", line 815, in _get_field\r\n raise ValueError(f'mutable default {type(f.default)} for field '\r\nValueError: mutable default <class 'official.modeling.optimization.configs.optimizer_config.SGDConfig'> for field sgd is not allowed: use default_factory",
"@Jayanth1812 and anyone else receiving a similar issue, it most likely has to do with your Python version. Downgrading to Python 3.9 works for me, but doing a downgrade might impact a lot of things. So to be safe and what worked for me was creating a new conda environment and following the installations here: https://tensorflow-object-detection-api-tutorial.readthedocs.io/en/latest/install.html\r\n\r\nAnd for Tensorflow GPU compatibility, after installing TensorFlow follow the instructions in section 4 'GPU Setup' in this document: https://www.tensorflow.org/install/pip",
"@Jayanth1812, you can see in your error stack trace, that the error is caused by the `tensorflow` library, not by the `datasets` library. See:\r\n```\r\nFile \"C:\\Users\\x0133252\\AppData\\Local\\anaconda3\\envs\\tensorflow\\Lib\\dataclasses.py\"\r\n```\r\n\r\nYou should open an issue in their repository instead: https://github.com/tensorflow/tensorflow "
] |
1,445,121,028
| 5,229
|
Type error when calling `map` over dataset containing 0-d tensors
|
closed
| 2022-11-11T08:27:28
| 2023-01-13T16:00:53
| 2023-01-13T16:00:53
|
https://github.com/huggingface/datasets/issues/5229
| null |
phipsgabler
| false
|
[
"Hi! \r\n\r\nWe could address this by calling `.item()` on such tensors to extract the value, but this would lose us the type, which could lead to storing the generated dataset in a suboptimal format. Considering this, I think the only proper fix would be implementing support for 0-D tensors on Apache Arrow's side (Arrow is the underlying format we use to store datasets on disk/in memory). WDYT @lhoestq?",
"I think we can just convert the item to a numpy typed scalar using `.numpy()` ?\r\n\r\nFor example this works:\r\n```python\r\nimport numpy as np\r\nimport pyarrow as pa\r\n\r\nassert pa.array([np.float64(1.0)]).type == pa.float64()\r\nassert pa.array([np.float32(1.0)]).type == pa.float32()\r\nassert pa.array([np.int32(1)]).type == pa.int32()\r\nassert pa.array([np.int64(1)]).type == pa.int64()\r\n```\r\n\r\nAnd therefore it would work the same as for PyTorch N-D Tensors: convert to Numpy Array to keep the type in `_cast_to_python_objects`, then convert to Arrow"
] |
1,444,763,105
| 5,228
|
Loading a dataset from the hub fails if you happen to have a folder of the same name
|
open
| 2022-11-11T00:51:54
| 2023-05-03T23:23:04
| null |
https://github.com/huggingface/datasets/issues/5228
| null |
dakinggg
| false
|
[
"`load_dataset` first checks for a local directory before checking for the Hub.\r\n\r\nTo make it explicit that it has to fetch the Hub, we could support the `hffs` syntax:\r\n```python\r\nload_dataset(\"hf://datasets/glue\")\r\n```\r\n\r\nwould that work for you ? Also cc @mariosasko who's leading the `hffs` project",
"yeah, that would be a fine solution.",
"This still has no proper solution in 2.11\r\n\r\nperhaps have a `download_config=\"force_remote\"` or just backtrack once you reach `EmptyDatasetError` locally and then try to load it from the hub (or a local cache, as that only gets checked if there is no local folder...?)"
] |
1,444,620,094
| 5,227
|
datasets.data_files.EmptyDatasetError: The directory at wikisql doesn't contain any data files
|
closed
| 2022-11-10T21:57:06
| 2023-10-07T05:04:41
| 2022-11-10T22:05:43
|
https://github.com/huggingface/datasets/issues/5227
| null |
ScottM-wizard
| false
|
[
"Fixed. Please close.",
"how to fix?i need your help"
] |
1,444,385,148
| 5,226
|
Q: Memory release when removing the column?
|
closed
| 2022-11-10T18:35:27
| 2022-11-29T15:10:10
| 2022-11-29T15:10:10
|
https://github.com/huggingface/datasets/issues/5226
| null |
bayartsogt-ya
| false
|
[
"Hi ! Datasets are memory mapped from your disk, i.e. they're not loaded in RAM. This is possible thanks to the Arrow data format.\r\n\r\nTherefore the column you remove is not in RAM, so removing it doesn't cause the RAM to decrease.",
"Thanks for the explanation! @lhoestq \r\nI wonder since it is memory mapped, can we reduce or remove this memory map?",
"Yes you can `del common_voice` for example or wait for it to be garbage collected"
] |
1,444,305,183
| 5,225
|
Add video feature
|
open
| 2022-11-10T17:36:11
| 2022-12-02T15:13:15
| null |
https://github.com/huggingface/datasets/issues/5225
| null |
nateraw
| false
|
[
"@NielsRogge @rwightman may have additional requirements regarding this feature.\r\n\r\nWhen adding a new (decodable) type, the hardest part is choosing the right decoding library. What I mean by \"right\" here is that it has all the features we need and is easy to install (with GPU support?).\r\n\r\nSome candidates/options:\r\n* [`decord`](https://github.com/dmlc/decord): no longer [maintained](https://github.com/dmlc/decord/issues/214), not trivial to install with GPU support\r\n* [`pyAV`](https://github.com/PyAV-Org/PyAV): used for CPU decoding in `torchvision`, GPU decoding not supported if I'm not mistaken, otherwise the best candidate probably\r\n* [`video_reader`](https://github.com/pytorch/vision/blob/de350bc01ad2193ea2888f0ce8a6a346d3cba5a9/torchvision/csrc/io/video_reader/video_reader.cpp): used for GPU decoding in `torchvision`, depends on `torch'\r\n* OpenCV: uses `ffmpeg` for video decoding under the hood\r\n* ...\r\n\r\nAnd the last resort is building our own library, which is the most flexible solution but also requires the most work.\r\n\r\nPS: I'm adding a link to an article that compares various video decoding libraries: https://towardsdatascience.com/lightning-fast-video-reading-in-python-c1438771c4e6",
"@mariosasko is GPU decoding a hard requirement here? Do we really need it? (I don't know)\r\n\r\nSomething to consider with `decord` is that it doesn't (AFAIK) support writing videos, so you'd still need something else for that. also I've noticed [issues](https://github.com/dmlc/decord/issues/242) with decord's ability to decode stereo audio streams along side the video (which you don't run into with PyAV).\r\n\r\n---\r\n\r\nI think PyAV should be able to do the job just fine to start. If we write the video io utilities as their own functions, we can hot swap them later if we find/write a different solution that's faster/better.",
"Video is still a bit of a mess, but I'd say pyAV is likely the best approach (or supporting all three via pytorchvideo, but that adds a middle man dependency).\r\n\r\nBeing able to decode on the GPU, into memory that could be passed off to a Tensor in whatever framework is being used would be the dream, I don't think there is any interop of that nature working right now. Number of decoder instances per GPU is limited so it's not clear if balancing load btw GPU decoders and CPUs would be needed in say large scale video training.\r\n\r\nAny of these solutions is less than ideal due to the nature of video, having a simple Python interface video / start -> end results in lots of extra memory (you need to decode whole range of the clips into a buffer before using anything). Any scalable video system would be streaming on the fly (issuing frames via callbacks as soon as the stream is far enough along to have re-ordered the frames and synced audio+video+other metadata (sensors, CC, etc).\r\n\r\n",
"For standalone usage, decoding on GPU could be ideal but isn't async processing of inputs on CPUs while letting the accelerator busy for training the de-facto? Of course, I am aware of other advanced mechanisms such as CPU offloading, but I think my point is conveyed. ",
"Here's a minimal implementation of the helper functions we'd need from PyAV, a lot of which I borrowed from `pytorchvideo`, stripping out the `torch` specific stuff:\r\n\r\n[](https://colab.research.google.com/gist/nateraw/c327cb6ff6b074e6ddc8068d19c0367d/pyav-io.ipynb)\r\n \r\nIt's not too much code...@mariosasko we could probably just maintain these helper fns within the `datasets` library, right? ",
"Also wanted to note I added a PR for video classification in `transformers` here, which uses `decord`. It's still open...should we make a decision now to align the libraries we are using between `datasets` and `transformers`? (CC @Narsil )\r\n\r\nhttps://github.com/huggingface/transformers/pull/20151",
"Fully agree on at least trying to unite things.\r\n\r\nMaking clear function boundaries to help us change dependency if needed seems like a good idea since there doesn't seem to be a clear winner.\r\n\r\nI also happen to like directly calling ffmpeg. For some reason it was a lot faster than pyav. "
] |
1,443,640,867
| 5,224
|
Seems to freeze when loading audio dataset with wav files from local folder
|
closed
| 2022-11-10T10:29:31
| 2023-04-25T09:54:05
| 2022-11-22T11:24:19
|
https://github.com/huggingface/datasets/issues/5224
| null |
uriii3
| false
|
[
"I just tried to do the same but changing the `.wav` files to `.mp3` files and that doesn't fix it.",
"I don't know if anyone will ever read this but I've tried to upload the same dataset with google colab and the output seems more clarifying. I didn't specify the train/test split so the dataset wasn't fully uploaded (or that is what I understood, might be wrong!!).\r\n\r\nNow, including the `drop_metadata` flag I can load the dataset normally (at least with colab notebook):\r\n\r\n```python\r\nfrom datasets import load_dataset\r\n\r\ndataset = load_dataset(\"audiofolder\", data_dir=\"../archive/Dataset\", , drop_metadata=True)\r\n```\r\n\r\nI'll close the issue.",
"@uriii3 Hello, I understand correctly that you converted your wav files to mp3?",
"Yes but it didn't matter. I don't remember which of them I ended up working with."
] |
1,442,610,658
| 5,223
|
Add SQL guide
|
closed
| 2022-11-09T19:10:27
| 2022-11-15T17:40:25
| 2022-11-15T17:40:21
|
https://github.com/huggingface/datasets/pull/5223
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5223",
"html_url": "https://github.com/huggingface/datasets/pull/5223",
"diff_url": "https://github.com/huggingface/datasets/pull/5223.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5223.patch",
"merged_at": "2022-11-15T17:40:21"
}
|
stevhliu
| true
|
[
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5223). All of your documentation changes will be reflected on that endpoint.",
"I think we may want more content on this page that's not SQL related. Some of that content probably already lives in the main `load` docs page, but might be bad to remove major things like csv/pandas from there...WDYT we should do @lhoestq ?",
"Maybe the main load page can only show one example and redirect to this page for more details ?\r\n\r\nWe can do the same for pandas stuff: have one example in load, and redirect to this page for more details",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5223). All of your documentation changes will be reflected on that endpoint."
] |
1,442,412,507
| 5,222
|
HuggingFace website is incorrectly reporting that my datasets are pickled
|
closed
| 2022-11-09T16:41:16
| 2022-11-09T18:10:46
| 2022-11-09T18:06:57
|
https://github.com/huggingface/datasets/issues/5222
| null |
ProGamerGov
| false
|
[
"cc @McPatate maybe you know what's happening ?",
"Yes I think I know what is happening. We check in zips for pickles, and the UI must display the pickle jar when a scan has an associated list of imports, even when empty.\r\n~I'll fix ASAP !~",
"> I'll fix ASAP !\r\n\r\nActually I'd rather leave it like that for now, as it indicates that we checked for pickles and nothing dangerous appeared :)",
"Closing the issue with the typical \"feature not a bug\" "
] |
1,442,309,094
| 5,221
|
Cannot push
|
closed
| 2022-11-09T15:32:05
| 2022-11-10T18:11:21
| 2022-11-10T18:11:11
|
https://github.com/huggingface/datasets/issues/5221
| null |
bayartsogt-ya
| false
|
[
"Did you run `huggingface-cli lfs-enable-largefiles` before committing or before adding ? Maybe you can try before adding\r\n\r\nAnyway I'd encourage you to split your data into several TAR archives if possible, this way the dataset can loaded faster using multiprocessing (by giving each process a subset of shards to process)",
"@lhoestq \r\nThanks for the help!\r\n> Maybe you can try before adding\r\n\r\nIt did not help\r\n\r\nBut I totally got your point about split into multiple TAR archives. It really helped!"
] |
1,441,664,377
| 5,220
|
Implicit type conversion of lists in to_pandas
|
closed
| 2022-11-09T08:40:18
| 2022-11-10T16:12:26
| 2022-11-10T16:12:26
|
https://github.com/huggingface/datasets/issues/5220
| null |
sanderland
| false
|
[
"I think this behavior comes from PyArrow:\r\n```python\r\nimport pyarrow as pa\r\nt = pa.table({\"a\": [[0]]})\r\nt.to_pandas().a.values[0]\r\n# array([0])\r\n```\r\n\r\nI believe this has to do with zero-copy: you can get a pandas DataFrame without copying the buffers from arrow, and therefore end up with numpy arrays.",
"That's interesting, I guess not much to do here then."
] |
1,441,255,910
| 5,219
|
Delta Tables usage using Datasets Library
|
open
| 2022-11-09T02:43:56
| 2023-03-02T19:29:12
| null |
https://github.com/huggingface/datasets/issues/5219
| null |
reichenbch
| false
|
[
"Hi ! Interesting :) Can you provide concrete examples of cases where it can be useful ?",
"Few example blogs and posts that might help on this - \r\n\r\n1. https://hevodata.com/learn/databricks-delta-tables/\r\n2. https://docs.databricks.com/delta/index.html\r\n\r\nBasically, we are looking at utility of Datasets library with Delta Lake Tables.\r\n",
"`datasets` can already read/write from parquet from/to a cloud storage using fsspec, if I understand correctly it's should be possible to load parquet files as delat lake tables no ? :) Or is there someting missing ?",
"@lhoestq Per my understanding, delta lake table is a bunch of paruqet files together with the meta to support ACID. For example file 1 contains v0.1 of record A while file 2 contains v0.2 of record A. I am assuming the Hugging face dataset would delegate the read/write delta table to 3rd party lib, maybe pyarrow. Correct me if I was wrong @reichenbch \r\n\r\nAnd I am assuming, people are asking the versioning of Hugging face datasets. But I am assuming Hugging face delegate this function to github and it is not the key requirement for Public Data set. It actually the key function of ML Ops, I am not sure whether hugging face would like expand to that area."
] |
1,441,254,194
| 5,218
|
Delta Tables usage using Datasets Library
|
closed
| 2022-11-09T02:42:18
| 2022-11-09T02:42:36
| 2022-11-09T02:42:36
|
https://github.com/huggingface/datasets/issues/5218
| null |
rcv-koo
| false
|
[] |
1,441,252,740
| 5,217
|
Reword E2E training and inference tips in the vision guides
|
closed
| 2022-11-09T02:40:01
| 2022-11-10T01:38:09
| 2022-11-10T01:36:09
|
https://github.com/huggingface/datasets/pull/5217
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5217",
"html_url": "https://github.com/huggingface/datasets/pull/5217",
"diff_url": "https://github.com/huggingface/datasets/pull/5217.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5217.patch",
"merged_at": "2022-11-10T01:36:08"
}
|
sayakpaul
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,441,041,947
| 5,216
|
save_elasticsearch_index
|
open
| 2022-11-08T23:06:52
| 2022-11-09T13:16:45
| null |
https://github.com/huggingface/datasets/issues/5216
| null |
amobash2
| false
|
[
"Hi ! I think there exist tools to dump and reload an index in your elastic search but I'm not super familiar with it.\r\n\r\nAnyway after reloading an index in elastic search you can call `ds.load_elasticsearch_index` which will connect the index to the dataset without re-indexing"
] |
1,440,334,978
| 5,214
|
Update github pr docs actions
|
closed
| 2022-11-08T14:43:37
| 2022-11-08T15:39:58
| 2022-11-08T15:39:57
|
https://github.com/huggingface/datasets/pull/5214
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5214",
"html_url": "https://github.com/huggingface/datasets/pull/5214",
"diff_url": "https://github.com/huggingface/datasets/pull/5214.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5214.patch",
"merged_at": "2022-11-08T15:39:57"
}
|
mishig25
| true
|
[
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5214). All of your documentation changes will be reflected on that endpoint."
] |
1,440,037,534
| 5,213
|
Add support for different configs with `push_to_hub`
|
closed
| 2022-11-08T11:45:47
| 2022-12-02T16:48:23
| 2022-12-02T16:44:07
|
https://github.com/huggingface/datasets/pull/5213
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5213",
"html_url": "https://github.com/huggingface/datasets/pull/5213",
"diff_url": "https://github.com/huggingface/datasets/pull/5213.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5213.patch",
"merged_at": null
}
|
polinaeterna
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5213). All of your documentation changes will be reflected on that endpoint.",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5213). All of your documentation changes will be reflected on that endpoint.",
"Nice thanks !\r\n\r\nWould it be possible to have the new folders at the same level as \"data\" ? This way they're all separated\r\n```\r\n├─ config-v1/\r\n│ ├── train-00000-00002-...-.parquet\r\n│ └── train-00001-00002-...-.parquet\r\n└ config-v2/\r\n ├── train-00000-00002-...-.parquet\r\n └── train-00001-00002-...-.parquet\r\n```\r\nand if you don't provide a config name, it goes in a folder named \"default\" instead, that would be loaded by default.\r\n\r\nWe could also write in the YAML something like\r\n```yaml\r\nconfigs:\r\n- name: config-v1\r\n data_dir: config-v1\r\n- name: config-v2\r\n data_dir: config-v2\r\n```\r\nand loading `config-v1` would be equivalent to run `load_dataset(ds_name, \"config-v1\", data_dir=\"config-v1\")`\r\n\r\nDo you think it would make sense ?\r\n\r\nFor backward compatibility we can just keep the \"data/*\" pattern. It's ok to expect users to have an updated version of `datasets` to be able to load datasets with configurations.",
"@lhoestq thank you for the feedback! i'll reflect on this on Moday, my mind just melted because of the fever.\r\n\r\n@mariosasko @albertvillanova what do you think?",
"Thanks for addressing this, @polinaeterna. It is good:\r\n- we support configs for datasets without scripts\r\n- we align the behavior to datasets with scripts as much as possible\r\n\r\nMaybe adding some tests will help clarify what is the expected behavior...",
"After some discussion with @lhoestq we decided that it's better to rely on metadata file than on data files patterns. \r\n\r\nSo we decided to introduce a new field to yaml (like `configs` or smth like that) that would contain arbitrary configs kwargs to be passed to loader, including `data_dir` and `data_files`. \r\nThis is more aligned with datasets with custom scripts where we explicitly write all the supported configs and config parameters in the code and is extendable to all packaged modules.\r\nThis would solve https://github.com/huggingface/datasets/issues/5209\r\n\r\n(@lhoestq was right 21 days ago, this is a more general solution idk why i ignored this...)",
"closed in favor of https://github.com/huggingface/datasets/pull/5331"
] |
1,439,642,483
| 5,212
|
Fix CI require_beam maximum compatible dill version
|
closed
| 2022-11-08T07:30:01
| 2022-11-15T06:32:27
| 2022-11-15T06:32:26
|
https://github.com/huggingface/datasets/pull/5212
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5212",
"html_url": "https://github.com/huggingface/datasets/pull/5212",
"diff_url": "https://github.com/huggingface/datasets/pull/5212.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5212.patch",
"merged_at": "2022-11-15T06:32:26"
}
|
albertvillanova
| true
|
[
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5212). All of your documentation changes will be reflected on that endpoint."
] |
1,438,544,617
| 5,211
|
Update Overview.ipynb google colab
|
closed
| 2022-11-07T15:23:52
| 2022-11-29T15:59:48
| 2022-11-29T15:54:17
|
https://github.com/huggingface/datasets/pull/5211
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5211",
"html_url": "https://github.com/huggingface/datasets/pull/5211",
"diff_url": "https://github.com/huggingface/datasets/pull/5211.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5211.patch",
"merged_at": "2022-11-29T15:54:17"
}
|
lhoestq
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"WDYT @albertvillanova ?",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5211). All of your documentation changes will be reflected on that endpoint."
] |
1,438,492,507
| 5,210
|
Tweak readme
|
closed
| 2022-11-07T14:51:23
| 2022-11-24T11:35:07
| 2022-11-24T11:26:16
|
https://github.com/huggingface/datasets/pull/5210
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5210",
"html_url": "https://github.com/huggingface/datasets/pull/5210",
"diff_url": "https://github.com/huggingface/datasets/pull/5210.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5210.patch",
"merged_at": "2022-11-24T11:26:16"
}
|
lhoestq
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"Nit: We should also update the `Disclaimers` section to let the dataset owners know they should use Hub discussions rather than GH issues for removal requests/updates",
"Updated the disclaimers section, thanks !\r\n\r\nDoes it sound good to you @albertvillanova ?"
] |
1,438,367,678
| 5,209
|
Implement ability to define splits in metadata section of dataset card
|
closed
| 2022-11-07T13:27:16
| 2023-07-21T14:36:02
| 2023-07-21T14:36:01
|
https://github.com/huggingface/datasets/issues/5209
| null |
merveenoyan
| false
|
[
"@merveenoyan Do you want different files to be splits or configurations?\r\n\r\nFrom [what you specified in `Readme.md`](https://huggingface.co/datasets/inria-soda/tabular-benchmark/commit/fb4575853772c62a20203bdd6cc0202f5db4ce4e) I hypothesize that you want to have 4 **configs** corresponding to directories: `\"clf_cat\", \"clf_num\", \"reg_cat\", \"reg_num\"`. And inside each config you require to have as many splits as there are `csv` files\r\nso if you run \r\n```python\r\nload_dataset(\"inria-soda/tabular-benchmark\", \"clf_cat\", split=\"compass\")\r\n```\r\nyou will generate the data only from `compass.csv` file.\r\nIn this case, running `load_dataset(\"inria-soda/tabular-benchmark\", \"clf_cat\"`) without split parameter will return `DatasetDict` object with `\"KDDCup09_upselling\", \"cat_compass\", \"cat_covertype\", ... \"road_safety\"` keys (which values are splits - `Dataset` objects)\r\n\r\n**or**\r\ndo you want each file to be a separate config? Like:\r\n```python\r\nload_dataset(\"inria-soda/tabular-benchmark\", \"clf_cat_compass\") # returns DatasetDict with a single \"train\" split\r\n```\r\n**or**\r\nmaybe smth completely different? :smile: \r\n\r\nAnyway, now I have an impression that this is probably rather a matter of automatically inferring configs from repository structure rather than providing parameters in metadata yaml.\r\n",
"@polinaeterna I want the latter where you can think of every CSV file as a config, like MNLI from GLUE.",
"@merveenoyan @lhoestq I see two solutions to this case. \r\n1. Parse configurations automatically from directories names. That is, if you have data structure like:\r\n```\r\ntabular-benchmark\r\n └─clf_cat_compass\r\n └─compass.csv\r\n └─clf_cat_cat_covertype\r\n └─covertype.csv\r\n ...\r\n └─reg_cat_house_sales\r\n └─house_sales.csv\r\n```\r\nyou'll get \"clf_cat_compass\", \"clf_cat_cat_covertype\", ... \"reg_cat_house_sales\" configurations that would contain **only files from corresponding directories**. \r\n**\\+** this is a requested change and needed in general and would solve other problems, see https://github.com/huggingface/datasets/issues/4578, would also help with https://github.com/huggingface/datasets/pull/5213 which I'm working on currently\r\n**\\+** would allow users to do just `load_dataset(“inria-soda/tabular-benchmark”, “clf_cat_compass”)`, no `data_files` param required\r\n**\\-** in this specific case it would require restructuring of the data - putting each file in a directory named as a config name (to me personally it doesn't seem to be a big deal) \r\n\r\n2. More or less what we discussed before - add support for manually specifying parameters in the metadata. We can add new metadata yaml field (say, `\"custom_configs_info\"`), so that we can provide smth like:\r\n```yaml\r\n---\r\n...\r\ndataset_info:\r\n ... \r\ncustom_configs_info:\r\n- config_name: reg_cat_house_sales\r\n data_files:\r\n - reg_cat/house_sales.csv\r\n- config_name: clf_cat_compass\r\n data_files:\r\n - clf_cat/compass.csv\r\n...\r\n---\r\n```\r\n**\\+** Would be useful not only for tabular data and not only for `data_files` parameter - any packaged dataset’s viewer can be customized to use specific, non-default parameters. @merveenoyan do you maybe have any other examples/use cases in mind where you want to provide any specific parameters to the viewer? \r\n**\\-** I'm not sure here but assume that it might require changes in interaction with the viewer on the hub side - to parse these configurations, as they not default configurations (not in `BUILDER_CONFIGS` list). cc @severo But probably this can be solved on the `datasets` side too.\r\n\r\nOverall, I would start from implementing the first solution since it's related to what I'm doing now and is super useful for `datasets` in general. And then if we agree that having more flexibility in providing parameters to the viewer is required, I can implement the second one. Let me know what you think :) ",
"> We can add new metadata yaml field (say, \"custom_configs_info\"), so that we can provide smth like:\r\n\r\nLove it ! Some other ideas to name the \"custom_configs_info\" field: \"configs\", \"parameters\", \"config_args\", \"configurations\"\r\n\r\n> it might require changes in interaction with the viewer on the hub side - to parse these configurations, as they not default configurations (not in BUILDER_CONFIGS list)\r\n\r\nIf we update the `get_dataset_config_names()` function in `datasets` in inspect.py we should be fine - that's what the viewer is using\r\n\r\n> Overall, I would start from implementing the first solution since it's related to what I'm doing now and is super useful for datasets in general. And then if we agree that having more flexibility in providing parameters to the viewer is required, I can implement the second one. Let me know what you think :)\r\n\r\nActually I feel like the second solution includes the first use case you mentioned. If you implement the second solution, then users would just have to add a few lines of YAML and their directories would be considered configurations no ? Maybe there's no need to implement two different logics to do the same thing",
"is there any update on this? 🕵🏻",
"@merveenoyan I haven't started working on this yet, working on adding configs to packaged datasets instead: https://github.com/huggingface/datasets/pull/5213 because this both would allow you to solve your issue and is a frequently requested feature.\r\n\r\nadding arbitrary parameters to yaml would be my next task i think!",
"@merveenoyan ignore my comment above, I'm switching to this task now :D",
"I want to be able to create folders in a model.",
"Addressed in #5331 "
] |
1,438,035,707
| 5,208
|
Refactor CI hub fixtures to use monkeypatch instead of patch
|
closed
| 2022-11-07T09:25:05
| 2022-11-08T06:51:20
| 2022-11-08T06:49:17
|
https://github.com/huggingface/datasets/pull/5208
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5208",
"html_url": "https://github.com/huggingface/datasets/pull/5208",
"diff_url": "https://github.com/huggingface/datasets/pull/5208.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5208.patch",
"merged_at": "2022-11-08T06:49:17"
}
|
albertvillanova
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,437,858,506
| 5,207
|
Connection error of the HuggingFace's dataset Hub due to SSLError with proxy
|
open
| 2022-11-07T06:56:23
| 2025-03-08T09:04:10
| null |
https://github.com/huggingface/datasets/issues/5207
| null |
leemgs
| false
|
[
"Hi ! It looks like an issue with your python environment, can you make sure you're able to run GET requests to https://huggingface.co using `requests` in python ?",
"Thanks for your reply. Does this mean that I have to use the `do_dataset `function and the `requests `function to download the dataset from the company's proxy environment?\r\n\r\n\r\n* Reference: \r\n```bash\r\n### How to load this dataset directly with the [datasets](https://github.com/huggingface/datasets) library\r\n\r\n\r\n* https://huggingface.co/datasets/moyix/debian_csrc\r\n\r\n* from datasets import load_dataset\r\ndataset = load_dataset(\"moyix/debian_csrc\")\r\n\r\n\r\n\r\n### Or just clone the dataset repo\r\n\r\n\r\ngit lfs install\r\ngit clone https://huggingface.co/datasets/moyix/debian_csrc\r\n# if you want to clone without large files – just their pointers\r\n# prepend your git clone with the following env var:\r\nGIT_LFS_SKIP_SMUDGE=1\r\n```",
"You can use `requests` to see if downloading a file from the Hugging Face Hub works. If so, then `datasets` should work as well. If not, then you have to find another way using an internet connection that works",
"I resolved this issue by applying to \"unblock websites\" at https://huggingface.com in a corporate network environment with a firewall. \r\n",
"> Hi ! It looks like an issue with your python environment, can you make sure you're able to run GET requests to https://huggingface.co using `requests` in python ?\r\n\r\nyes，but still not work\r\n\r\n\r\n\r\n",
"I read https://github.com/huggingface/datasets/blob/main/src/datasets/load.py, it fail when get the dataset metadata, so download_config has not worked.\r\n```python\r\n hf_api = HfApi(config.HF_ENDPOINT)\r\n try:\r\n dataset_info = hf_api.dataset_info(\r\n repo_id=path,\r\n revision=revision,\r\n token=download_config.token,\r\n timeout=100.0,\r\n )\r\n except Exception as e: # noqa catch any exception of hf_hub and consider that the dataset doesn't exist\r\n if isinstance(\r\n e,\r\n (\r\n OfflineModeIsEnabled,\r\n requests.exceptions.ConnectTimeout,\r\n requests.exceptions.ConnectionError,\r\n ),\r\n ):\r\n raise ConnectionError(f\"Couldn't reach '{path}' on the Hub ({type(e).__name__})\")\r\n```\r\nI configure the huggingface_hub api, use configure_http_backend\r\n```python\r\nfrom huggingface_hub import configure_http_backend\r\ndef backend_factory() -> requests.Session:\r\n session = requests.Session()\r\n session.proxies = proxy\r\n session.verify = False\r\n return session\r\n\r\nconfigure_http_backend(backend_factory=backend_factory)\r\n```\r\nIt works.",
"Even tough it does not look like a certificate error in the error message, I had the same error and adding following lines to my code solved my problem.\r\n\r\nimport os\r\nos.environ['CURL_CA_BUNDLE'] = ''",
"@kuikuikuizzZ Could you please explain where the configuration code is added?",
"> Even tough it does not look like a certificate error in the error message, I had the same error and adding following lines to my code solved my problem.\r\n> \r\n> import os os.environ['CURL_CA_BUNDLE'] = ''\r\n\r\nWorked for as well!\r\nI faced the issue while submitting jobs through SLURM.",
"> Even tough it does not look like a certificate error in the error message, I had the same error and adding following lines to my code solved my problem.\r\n> \r\n> import os os.environ['CURL_CA_BUNDLE'] = ''\r\n\r\ndoesn't work , what does this code mean?",
"If you're working on a cluster, may be that they disabled remote connections for security purposes, you will have to download the files on your local machine and then transfer them to your cluster through scp or some other transfer protocol. I know you've probably resolved the issue, but that is for anyone in the future who might stumble across this thread and needs help because I struggled with that even after reading this thread.",
"> Even tough it does not look like a certificate error in the error message, I had the same error and adding following lines to my code solved my problem.\r\n> \r\n> import os os.environ['CURL_CA_BUNDLE'] = ''\r\n\r\nIf this not work, try this:\r\n```bash\r\nexport http_proxy=\"http://127.0.0.1:10810\"\r\nexport https_proxy=\"http://127.0.0.1:10810\"\r\ngit config --global http.proxy http://127.0.0.1:10810\r\ngit config --global https.proxy http://127.0.0.1:10810\r\n\r\njupyter notebook\r\n```\r\n\r\nset your proxy env first, then start notebook **in this session**\r\n",
"> If you're working on a cluster, may be that they disabled remote connections for security purposes, you will have to download the files on your local machine and then transfer them to your cluster through scp or some other transfer protocol. I know you've probably resolved the issue, but that is for anyone in the future who might stumble across this thread and needs help because I struggled with that even after reading this thread.\r\n\r\nThank you buddy!",
"@shafferjohn \nexport http_proxy=\"http://127.0.0.1:10810\"\nexport https_proxy=\"http://127.0.0.1:10810\"\ngit config --global http.proxy http://127.0.0.1:10810\ngit config --global https.proxy http://127.0.0.1:10810\n\njupyter notebook\n\nthis way worked for me"
] |
1,437,223,894
| 5,206
|
Use logging instead of printing to console
|
closed
| 2022-11-05T23:48:02
| 2022-11-06T00:06:00
| 2022-11-06T00:05:59
|
https://github.com/huggingface/datasets/issues/5206
| null |
bilelomrani1
| false
|
[
"Actually upon closer inspection, it is documented in the code that this behavior is intentional, so I'll close this."
] |
1,437,221,987
| 5,205
|
Add missing `DownloadConfig.use_auth_token` value
|
closed
| 2022-11-05T23:36:36
| 2022-11-08T08:13:00
| 2022-11-07T16:20:24
|
https://github.com/huggingface/datasets/pull/5205
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5205",
"html_url": "https://github.com/huggingface/datasets/pull/5205",
"diff_url": "https://github.com/huggingface/datasets/pull/5205.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5205.patch",
"merged_at": "2022-11-07T16:20:24"
}
|
alvarobartt
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,437,221,259
| 5,204
|
`push_to_hub` not propagating `token` through `DownloadConfig`
|
closed
| 2022-11-05T23:32:20
| 2022-11-08T10:12:09
| 2022-11-08T10:12:08
|
https://github.com/huggingface/datasets/issues/5204
| null |
alvarobartt
| false
|
[
"#self-assign",
"@lhoestq can you close this issue as part of the recent #5205 merge? Thanks 🤗 ",
"Thank you :)"
] |
1,436,710,518
| 5,203
|
Update canonical links to Hub links
|
closed
| 2022-11-04T22:50:50
| 2022-11-07T18:43:05
| 2022-11-07T18:40:19
|
https://github.com/huggingface/datasets/pull/5203
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5203",
"html_url": "https://github.com/huggingface/datasets/pull/5203",
"diff_url": "https://github.com/huggingface/datasets/pull/5203.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5203.patch",
"merged_at": "2022-11-07T18:40:19"
}
|
stevhliu
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,435,886,090
| 5,202
|
CI fails after bulk edit of canonical datasets
|
closed
| 2022-11-04T10:51:20
| 2023-02-16T09:11:10
| 2023-02-16T09:11:10
|
https://github.com/huggingface/datasets/issues/5202
| null |
albertvillanova
| false
|
[
"Fixed by: https://huggingface.co/datasets/paws/discussions/1"
] |
1,435,881,554
| 5,201
|
Do not sort splits in dataset info
|
closed
| 2022-11-04T10:47:21
| 2022-11-04T14:47:37
| 2022-11-04T14:45:09
|
https://github.com/huggingface/datasets/pull/5201
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5201",
"html_url": "https://github.com/huggingface/datasets/pull/5201",
"diff_url": "https://github.com/huggingface/datasets/pull/5201.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5201.patch",
"merged_at": "2022-11-04T14:45:09"
}
|
polinaeterna
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"It would be coherent with https://github.com/huggingface/datasets-server/issues/614#issuecomment-1290534153",
"I think we started working on this issue nearly at the same time... :sweat_smile: \r\n- CI was fixed with this: https://huggingface.co/datasets/paws/discussions/1\r\n\r\nRelated issue:\r\n- #5202",
"@albertvillanova yeah I noticed it right after the PR :smile: thank you! the fix of the dataset info yaml fixes tests on CI, but in general order of splits in yaml influences the order in which they are displayed in the viewer, if I understand it correctly. So I suggest not to sort splits in yaml initially to avoid this for other datasets in the future. I think [this change](https://github.com/huggingface/datasets/pull/5201/files#diff-198ba4fdf2f94cb3e1aba8a0170a43b08d4ab5636d682374321c5a383a8be24dR571) should work for it. \r\n\r\nChanges to tests here maybe can be reverted considering that order in yaml now corresponds to the one in tests, thanks to your change in the dataset info.",
"Hehe, @polinaeterna, we make comments nearly at the same time as well... :laughing: "
] |
1,435,831,559
| 5,200
|
Some links to canonical datasets in the docs are outdated
|
closed
| 2022-11-04T10:06:21
| 2022-11-07T18:40:20
| 2022-11-07T18:40:20
|
https://github.com/huggingface/datasets/issues/5200
| null |
polinaeterna
| false
|
[
"Thanks for catching this, I can go through the docs and replace the links to their corresponding datasets on the Hub!"
] |
1,434,818,836
| 5,199
|
Deprecate dummy data generation command
|
closed
| 2022-11-03T15:05:54
| 2022-11-04T14:01:50
| 2022-11-04T13:59:47
|
https://github.com/huggingface/datasets/pull/5199
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5199",
"html_url": "https://github.com/huggingface/datasets/pull/5199",
"diff_url": "https://github.com/huggingface/datasets/pull/5199.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5199.patch",
"merged_at": "2022-11-04T13:59:47"
}
|
mariosasko
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,434,699,165
| 5,198
|
Add note about the name of a dataset script
|
closed
| 2022-11-03T13:51:32
| 2022-11-04T12:47:59
| 2022-11-04T12:46:01
|
https://github.com/huggingface/datasets/pull/5198
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5198",
"html_url": "https://github.com/huggingface/datasets/pull/5198",
"diff_url": "https://github.com/huggingface/datasets/pull/5198.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5198.patch",
"merged_at": "2022-11-04T12:46:01"
}
|
polinaeterna
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,434,676,150
| 5,197
|
[zstd] Use max window log size
|
open
| 2022-11-03T13:35:58
| 2022-11-03T13:45:19
| null |
https://github.com/huggingface/datasets/pull/5197
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5197",
"html_url": "https://github.com/huggingface/datasets/pull/5197",
"diff_url": "https://github.com/huggingface/datasets/pull/5197.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5197.patch",
"merged_at": null
}
|
reyoung
| true
|
[
"@albertvillanova Please take a review.",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5197). All of your documentation changes will be reflected on that endpoint."
] |
1,434,401,646
| 5,196
|
Use hfh hf_hub_url function
|
closed
| 2022-11-03T10:08:09
| 2022-12-06T11:38:17
| 2022-11-09T07:15:12
|
https://github.com/huggingface/datasets/pull/5196
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5196",
"html_url": "https://github.com/huggingface/datasets/pull/5196",
"diff_url": "https://github.com/huggingface/datasets/pull/5196.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5196.patch",
"merged_at": "2022-11-09T07:15:12"
}
|
albertvillanova
| true
|
[
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5196). All of your documentation changes will be reflected on that endpoint.",
"@lhoestq I think we should first agree if `datasets` can introduce the breaking change of ignoring `config.HUB_DATASETS_URL`: some users may have override this.\r\n\r\nIf so, I then would suggest to initiate a deprecation cycle.",
"After a discussion with the rest of the datasets team, we agreed we can introduce the breaking change of ignoring `config.HUB_DATASETS_URL`: this will have minimal impact, only for **private Hubs**. We will address eventual possible impacts in the future.\r\n\r\nAdditionally, we also ignore `config.HUB_DEFAULT_VERSION`.\r\n\r\nSee explanation in this PR description: https://github.com/huggingface/datasets/pull/5196#issue-1434401646",
"I'm trying to upgrade datasets to 2.7.0 in https://github.com/huggingface/datasets-server, and the tests fail due to this change. I think it's a breaking change (that was not listed in https://github.com/huggingface/datasets/releases/tag/2.7.0) since code that previously worked (by setting `datasets.config.HUB_DATASETS_URL = CI_HUB_DATASETS_URL` for example) does not work anymore.\r\n\r\nI'm not sure what is the correct way to set up the tests; besides setting the env var \"HF_ENDPOINT\" before launching the tests (which, I think, is not a good way to do: the tests should not depend on the environment).",
"OK, I re-read this thread, and https://github.com/huggingface/datasets/pull/5196#issuecomment-1307430175 explicitely states that `config.HUB_DATASETS_URL` (as well as `config.HUB_DEFAULT_VERSION`) is now ignored. I was expecting the breaking changes to be listed in the release notes: https://github.com/huggingface/datasets/releases/tag/2.7.0.",
"> I'm not sure what is the correct way to set up the tests; besides setting the env var \"HF_ENDPOINT\" before launching the tests (which, I think, is not a good way to do: the tests should not depend on the environment).\r\n\r\nI think the current workaround of settings an env variable before launching the tests is \"not so bad\" when considering the fact that env variables are evaluated at import time in `huggingface_hub` (and most probable `datasets` as well). I think that when refactoring this in huggingface_hub (https://github.com/huggingface/huggingface_hub/issues/1172) I'll opt for instantiating a `Settings` object (or `Constants`) that contains all the settings variables. This way it will not be possible to import attributes individually + tests would be easier. As I see it, it would be similar to [what `Pydantic` does](https://pydantic-docs.helpmanual.io/usage/settings/) even though we most probably don't want Pydantic as a root dependency just for that. ",
"You can use fixtures in your tests:\r\n```python\r\nCI_HUB_ENDPOINT = \"https://hub-ci.huggingface.co\"\r\nCI_HUB_DATASETS_URL = CI_HUB_ENDPOINT + \"/datasets/{repo_id}/resolve/{revision}/{path}\"\r\nCI_HFH_HUGGINGFACE_CO_URL_TEMPLATE = CI_HUB_ENDPOINT + \"/{repo_id}/resolve/{revision}/{filename}\"\r\n\r\n@pytest.fixture\r\ndef ci_hfh_hf_hub_url(monkeypatch):\r\n monkeypatch.setattr(\r\n \"huggingface_hub.file_download.HUGGINGFACE_CO_URL_TEMPLATE\", CI_HFH_HUGGINGFACE_CO_URL_TEMPLATE\r\n )\r\n\r\n@pytest.fixture\r\ndef ci_hub_config(monkeypatch):\r\n monkeypatch.setattr(\"datasets.config.HF_ENDPOINT\", CI_HUB_ENDPOINT)\r\n monkeypatch.setattr(\"datasets.config.HUB_DATASETS_URL\", CI_HUB_DATASETS_URL)\r\n```\r\n\r\nand use `@pytest.fixture(autouse=True)` if you want to always use the CI endpoints.\r\n\r\nAnd when `huggingface-hub` and `datasets` change the way we can set the endpoint, we'll just need to update the fixtures.\r\nI think ultimately you'll only have to change the `huggingface-hub` endpoint settings\r\n",
"OK.\r\n\r\nIn fact, in datasets-server we set `config.HUB_DATASETS_URL` (https://github.com/huggingface/datasets-server/blob/35a30dbcd687b26db1f02502ea8305f70c064473/workers/splits/src/splits/config.py#L26) at config time, before starting the workers. It's not an issue with how to launch the tests, but with the app in itself.\r\n\r\nI understand that for now, the only way to fix this is to setup `HF_ENDPOINT` in the env when launching the app (currently, we set the endpoint with `COMMON_HF_ENDPOINT`, a custom env var I set to be sure not to have side-effects)",
"> You can use fixtures in your tests:\r\n\r\nThanks, used in https://github.com/huggingface/datasets-server/pull/644."
] |
1,434,290,689
| 5,195
|
[wip testing docs]
|
closed
| 2022-11-03T08:37:34
| 2023-04-04T15:10:37
| 2023-04-04T15:10:33
|
https://github.com/huggingface/datasets/pull/5195
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5195",
"html_url": "https://github.com/huggingface/datasets/pull/5195",
"diff_url": "https://github.com/huggingface/datasets/pull/5195.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5195.patch",
"merged_at": null
}
|
mishig25
| true
|
[
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5195). All of your documentation changes will be reflected on that endpoint."
] |
1,434,206,951
| 5,194
|
Fix docs about dataset_info in YAML
|
closed
| 2022-11-03T07:10:23
| 2022-11-03T13:31:27
| 2022-11-03T13:29:21
|
https://github.com/huggingface/datasets/pull/5194
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5194",
"html_url": "https://github.com/huggingface/datasets/pull/5194",
"diff_url": "https://github.com/huggingface/datasets/pull/5194.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5194.patch",
"merged_at": "2022-11-03T13:29:21"
}
|
albertvillanova
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,433,883,780
| 5,193
|
"One or several metadata. were found, but not in the same directory or in a parent directory"
|
closed
| 2022-11-02T22:46:25
| 2022-11-03T13:39:16
| 2022-11-03T13:35:44
|
https://github.com/huggingface/datasets/issues/5193
| null |
lambda-science
| false
|
[
"Also unrelated but still: https://huggingface.co/docs/datasets/image_dataset#generate-the-dataset\r\n```If your loading script passed the test, you should now have a dataset_infos.json file in your dataset folder.```\r\nIt's not the case anymore as it's now in the readme.md, it was confusing to me",
"And here is my data loader script: https://huggingface.co/datasets/corentinm7/MyoQuant-SDH-Data/blob/main/SDH_16k.py\r\nI have one file archive to download that contains the images for all splits and one `metadata.jsonl` to download that contains the informations about what image goes into what split.",
"Hi @lambda-science! It seems that your repo is recognized as a packaged module [ImageFolder](https://huggingface.co/docs/datasets/main/en/image_dataset#imagefolder), not as a dataset with the custom loading script, because loader looks for a script that has the same name as the dataset repo. So please try to rename your script to `MyoQuant-SDH-Data.py`, this should help.",
"> Hi @lambda-science! It seems that your repo is recognized as a packaged module [ImageFolder](https://huggingface.co/docs/datasets/main/en/image_dataset#imagefolder), not as a dataset with the custom loading script, because loader looks for a script that has the same name as the dataset repo. So please try to rename your script to `MyoQuant-SDH-Data.py`, this should help.\r\n\r\nHi !\r\n\r\nThank you for your answer. That was... embarrassingly easy, sorry for this issue, everything is fixed now ! \r\n\r\nHave a nice day ! :)",
"@lambda-science that's not embarrassing at all! it's actually not clear from the documentation that the script should have the same name, so thank you for the issue, we'll add this information to the docs :) "
] |
1,433,199,790
| 5,192
|
Drop labels in Image and Audio folders if files are on different levels in directory or if there is only one label
|
closed
| 2022-11-02T14:01:41
| 2022-11-15T16:32:53
| 2022-11-15T16:31:07
|
https://github.com/huggingface/datasets/pull/5192
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5192",
"html_url": "https://github.com/huggingface/datasets/pull/5192",
"diff_url": "https://github.com/huggingface/datasets/pull/5192.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5192.patch",
"merged_at": "2022-11-15T16:31:07"
}
|
polinaeterna
| true
|
[
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5192). All of your documentation changes will be reflected on that endpoint.",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5192). All of your documentation changes will be reflected on that endpoint.",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5192). All of your documentation changes will be reflected on that endpoint.",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5192). All of your documentation changes will be reflected on that endpoint.",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5192). All of your documentation changes will be reflected on that endpoint.",
"> Nit: maybe we can use the count_path_segments function from this file for counting (updated with your logic to make it faster).\r\n\r\n@mariosasko just to make sure I understood you correctly - are you okay with this change? (actually `os.path.normpath` is redundant here as paths from `data_files` should be already normalized but just in case)\r\nhttps://github.com/huggingface/datasets/pull/5192/files#diff-1f09f7a178211f7539b1499b64b69793bd53b30c8b7b34cfcc5835e25d31929fR33\r\nIf you are, we can merge.",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5192). All of your documentation changes will be reflected on that endpoint.",
"The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_5192). All of your documentation changes will be reflected on that endpoint.",
"awesome ! :D"
] |
1,433,191,658
| 5,191
|
Make torch.Tensor and spacy models cacheable
|
closed
| 2022-11-02T13:56:18
| 2022-11-02T17:20:48
| 2022-11-02T17:18:42
|
https://github.com/huggingface/datasets/pull/5191
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5191",
"html_url": "https://github.com/huggingface/datasets/pull/5191",
"diff_url": "https://github.com/huggingface/datasets/pull/5191.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5191.patch",
"merged_at": "2022-11-02T17:18:42"
}
|
mariosasko
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,433,014,626
| 5,190
|
`path` is `None` when downloading a custom audio dataset from the Hub
|
closed
| 2022-11-02T11:51:25
| 2022-11-02T12:55:02
| 2022-11-02T12:55:02
|
https://github.com/huggingface/datasets/issues/5190
| null |
lewtun
| false
|
[
"Hi! Yes, this is expected behavior - we do this as a security measure to not leak local paths (this info would be useless on other users' machines anyways) and only push audio bytes. \r\n"
] |
1,432,769,143
| 5,189
|
Reduce friction in tabular dataset workflow by eliminating having splits when dataset is loaded
|
open
| 2022-11-02T09:15:02
| 2022-12-06T12:13:17
| null |
https://github.com/huggingface/datasets/issues/5189
| null |
merveenoyan
| false
|
[
"I have to admit I'm not a fan of this idea, as this would result in a non-consistent behavior between tabular and non-tabular datasets, which is confusing if done without the context you provided. Instead, we could consider returning a `Dataset` object rather than `DatasetDict` if there is only one split in the generated dataset. But then again, I think this lib is a bit too old to make such changes. @lhoestq @albertvillanova WDYT?\r\n\r\n",
"We can brainstorm here to see how we could make it happen ? And then depending on the options we see if it's a change we can do.\r\n\r\nI'm starting with a first reasoning\r\n\r\nCurrently not passing `split=` in `load_dataset` means \"return a dict with each split\".\r\n\r\nNow what would happen if a dataset has no split ? Ideally it should return one Dataset. And passing `split=` would have no sense. So depending on the dataset content, not passing `split=` should return a dict or a Dataset. In particular, those two cases should work:\r\n```python\r\n# case 1: dataset without split\r\nds = load_dataset(\"dataset_without_split\")\r\nds[0], ds[\"column_name\"], list(ds) # we want this\r\n\r\n# case 2: dataset with splits\r\nds = load_dataset(\"dataset_with_splits\")\r\nds[\"train\"] # this works and can't be changed\r\nds = load_dataset(\"dataset_with_splits\", split=\"train\")\r\nds[0], ds[\"column_name\"], list(ds) # this works and can't be changed\r\n```\r\n\r\nI can see several ideas:\r\n1. allowing `load_dataset` to return a different object based on the dataset content - either a Dataset or a DatasetDict\r\n - we can update `get_dataset_split_names` to return None or a list if users want to know in advance what object will be returned. They can also use `isinstance` _a posteriori_\r\n - but in this case we expect users to be careful when loading datasets and always to extra steps to check if they got a Dataset or DatasetDict\r\n2. merge Dataset and DatasetDict objects\r\n - they already share many functions: map, filter, push_to_hub etc.\r\n - we can define `ds[0]` to be the first item of the first split, and consider that the uses accesses rows from the full table of all the splits concatenated\r\n - however there is a collision when doing `ds[\"column_name\"]` or `ds[\"train\"]` that we need to address: the first returns a list, while the other returns a Dataset.\r\n\r\nWhat are your opinions on those two ideas ? Do you have other ideas in mind ?",
"I like the first idea more (concatenating splits doesn't seem useful, no?). This is a significant breaking change, so I think we should do a poll (or something similar) to gather more info on the actual \"expected behavior\" and wait for Datasets 3.0 if we decide to implement it.\r\n\r\nPS: @thomwolf also suggested the same thing a while ago (https://github.com/huggingface/datasets/issues/743#issuecomment-746074641).",
"I think it's an interesting improvement to the user experience for a case that comes often (no split) so I would definitively support it.\r\n\r\nI would be more in favor of option 2 rather than returning various types of objects from load_dataset and handling carefully the possible collisions indeed",
"Related: if a dataset only has one split, we don't show the splits select control in the dataset viewer on the Hub, eg. compare https://huggingface.co/datasets/hf-internal-testing/fixtures_image_utils/viewer/image/test with https://huggingface.co/datasets/glue/viewer/mnli/test.\r\n\r\nSee https://github.com/huggingface/moon-landing/pull/3858 for more details (internal)",
"I feel like the second idea is a bit more overkill. \r\n@severo I would say it's a bit irrelevant to the problem we have but is a separate problem @polinaeterna is solving at the moment. 😅 (also discussed on slack)",
"OK, sorry for polluting the thread. The relation I saw with the dataset viewer is that from a UX point of view, we hide the concepts of split and configuration whenever possible -> this issue feels like doing the same in the datasets library.",
"I would agree that returning different types based on the content of the dataset might be confusing.\r\n\r\nWe can do something similar to what `fetch_*` or `load_*` from `sklearn.datasets` do, which is to have an arg which changes the type of the returned type. For instance, `load_iris` would return a dict, but `load_iris(..., return_X_y=True)` would return a tuple.\r\n\r\nHere we can have a similar arg such as `return_X` which would then only return a single `DataSet` or an array.",
"> I feel like the second idea is a bit more overkill.\r\n\r\nOverkill in what sense ?\r\n\r\n> Here we can have a similar arg such as return_X which would then only return a single DataSet or an array.\r\n\r\nRight now one can already pass `split=\"all\"` to get one `Dataset` object with all the data in it (unsplit). We could also have something like `return_all=True` so make the API clearer.\r\n\r\n> I would be more in favor of option 2 rather than returning various types of objects from load_dataset and handling carefully the possible collisions indeed\r\n\r\nI think it would be ok to handle the collision by allowing both `ds[\"train\"]` and `ds[\"column_name\"]` (and maybe adding something like `ds.splits` for those who want to iterate over the splits or add new ones)",
"Would it make sense to remove the notion of \"split\" in `load_dataset`? I feel a lof of it comes from the want to have some sort of group of more or less similar dataset. \"train\"/\"test\"/\"validation\" are the traditional ones, but there are some datasets that have much more splits.\r\n\r\nWould it make sense to force `load_dataset` to only load a single `Dataset` object, and fail if it doesn't point to one. And have another method that's like `load_dataset_group_info` that can return a very arbitrary info class (Dict, List whatever), but you need to pass individual infos to `load_dataset` to run anything? Typically I don't think `DatasetDict.map` is really that helpful, but that's my personal opinion. This would help make things more readable (typically knowing if an object is a `Dataset` or a `DatasetDict`)",
"> Would it make sense to remove the notion of \"split\" in load_dataset?\r\n\r\nI think we need to keep it - though in practice people can name the splits whatever they want anyway.\r\n\r\n> Would it make sense to force load_dataset to only load a single Dataset object, and fail if it doesn't point to one.\r\n\r\nWe need to keep backward compatibility ideally - in particular the load_dataset + ds[\"train\"] one",
"> I think we need to keep it - though in practice people can name the splits whatever they want anyway.\r\n\r\nIt was my understanding that the whole issue was that `load_dataset` returned multiple types of objects.\r\n\r\n> We need to keep backward compatibility ideally - in particular the load_dataset + ds[\"train\"] one\r\n\r\nYeah sorry I meant ideally. One can always start developing `load_dataset_v2` can deprecate the first one and remove it in the longer term.",
"> It was my understanding that the whole issue was that load_dataset returned multiple types of objects.\r\n\r\nYes indeed, but we still want to keep a way to load the train/val/test/whatever splits alone ;)",
"@thomasw21's solution is good but it will break backwards compatibility. 😅",
"Started to experiment with merging Dataset and DatasetDict. My plan is to define the splits of a Dataset in Dataset.info.splits (already exists, but never used). A Dataset would then be the concatenation of its splits if they exist.\r\n\r\nNot sure yet this is the way to go. My plan is to play with it and see and share it with you, so we can see if it makes sense from a UX point of view.",
"So just to make sure that I understand the current direction, people will have to be extra careful when handling splits right?\r\nImagine \"potato\" a dataset containing train/validation split:\r\n```\r\nload_dataset(\"potato\") # returns the concatenation of all the splits\r\n```\r\nPreviously the design would force you to choose a split (it would raise otherwise), or manually concat them if you really wanted to play with concatenated splits. Now it would potentially run without raising for a bit of time until you figure out that you've been training on both train and validation split.\r\n\r\nWould it make sense to use a dataset specific default instead of using the concatenation, typically \"potato\" dataset's default would be train?\r\n```\r\nload_dataset(\"potato\") # returns \"train\" split\r\nload_dataset(\"potato\", split=\"train\") # returns \"train\" split\r\nload_dataset(\"potato\", split=\"validation\") # returns \"validation\" split\r\nconcatenate_datasets([load_dataset(\"potato\", split=\"train\"), load_dataset(\"potato\", split=\"validation\")]) # returns concatenation\r\n```",
"> load_dataset(\"potato\") # returns \"train\" split\r\n\r\nTo avoid a breaking change we need to be able to do `load_dataset(\"potato\")[\"validation\"]` as well.\r\n\r\nIn that case I'd wonder where the validation split comes from, since the rows of the dataset wouldn't contain the validation split according to your example. That's why I'm more in favor of concatenating.\r\n\r\nA dataset is one table, that optionally has some split info about subsets (e.g. for training an evaluation)\r\n\r\nThis also allows anyone to re-split the dataset the way they want if they're not happy with the default:\r\n\r\n```python\r\nds = load_dataset(\"potato\").train_test_split(test_size=0.2)\r\ntrain_ds = ds[\"train\"]\r\ntest_ds = ds[\"test\"]\r\n```",
"Just thinking about this, we could just have `to_dataframe()` as `load_dataset(\"blah\").to_dataframe()` to get the whole dataset, and not change anything else.",
"I have a first implementation of option 2 (merging Dataset and DatasetDict) in this PR: https://github.com/huggingface/datasets/pull/5301/\r\n\r\nFeel free to play with it if you're interested, and let me know what you think. In this PR, a dataset is one table that optionally has some split info about subsets.",
"@adrinjalali we already have [to_pandas](https://huggingface.co/docs/datasets/package_reference/main_classes#datasets.Dataset.to_pandas) AFAIK that essentially does the same thing (for a dataset, not for a dataset dict), I was wondering if it makes sense to have this as I don't know portion of people who load non-tabular datasets into dataframes. @lhoestq I saw your PR and it will break a lot of things imo, WDYT of this option? ",
"> we already have [to_pandas](https://huggingface.co/docs/datasets/package_reference/main_classes#datasets.Dataset.to_pandas) AFAIK that essentially does the same thing (for a dataset, not for a dataset dict)\r\n\r\nyes correct :)\r\n\r\n> I saw your PR and it will break a lot of things imo\r\n\r\nDo you have concrete examples you can share ?\r\n\r\n> WDYT of this option?\r\n\r\nThe to_dataframe option ? I think it not enough, since you'd still get a `DatasetDict({\"train\": Dataset()})` if you load a dataset with no splits (e.g. one CSV), and this doesn't really make sense.\r\n\r\nNote that in the PR I opened you can do\r\n```python\r\nds = load_dataset(\"dataset_with_just_one_csv\") # Dataset type\r\ndf = load_dataset(\"dataset_with_just_one_csv\").to_pandas() # DataFrame type\r\n```",
"@lhoestq no I think @adrinjalali and I meant when user calls `to_dataframe` if there's only train split in `DatasetDict` we could directly load that into dataframe. This might cause a confusion given there's to_pandas but I think it's more intuitive and least breaking change. (given people -who use `datasets` for tabular workflows- will eventually call `to_pandas` anyway) ",
"So in that case it would be fine to still end up with a dataset dict with a \"train\" split ?",
"yeah what I mean is this:\r\n\r\n```py\r\ndataset = load_dataset(\"blah\")\r\n\r\n# deal with a split of the dataset\r\ntrain = dataset[\"train\"]\r\ntrain_df = dataset[\"train\"].to_dataframe()\r\n\r\n# deal with the whole dataset\r\ndataset_df = dataset.to_dataframe()\r\n```\r\n\r\nSo we do two things to improve tabular experience:\r\n- allow datasets to have a single split\r\n- add `to_dataframe` to the root dict level so that users can simply call `df = load_dataset(\"blah\").to_dataframe()` and have it in their `pandas.DataFrame` object.",
"Ok ! Note that we already have `Dataset.to_pandas()` so for consistency I'd call it `DatasetDict.to_pandas()` as well, does it sound good to you ? This is something we can add pretty easily",
"yeah that sounds perfect @lhoestq !",
"> So just to make sure that I understand the current direction, people will have to be extra careful when handling splits right?\r\n\r\nWe can raise an error if someone does `load_dataset(...)[0]` if the dataset is made of several splits, and return the first example if there's one or zero splits (i.e. when it's not ambiguous). Had this idea from the dicussions in #5312 WDYT @thomasw21 ?",
"> We can raise an error if someone does load_dataset(...)[0] if the dataset is made of several splits,\r\n\r\nBut then how is that different to have the distinction between DatasetDict and Dataset then? Is it just that \"default behaviour when there are no splits or single split, it returns directly the split when there's no ambiguity\".\r\n\r\nAlso I was wondering how the concatenation could have heavy impacts when running mapping functions/filtering in batch? Typically can batch be somehow mixed?",
"> But then how is that different to have the distinction between DatasetDict and Dataset then?\r\n\r\nBecause it doesn't make sense to be able to do `example = ds[0]` or `examples = list(ds)` on a class named `DatasetDict` of type `Dict[str, Dataset]`.\r\n\r\n> Also I was wondering how the concatenation could have heavy impacts when running mapping functions/filtering in batch? Typically can batch be somehow mixed?\r\n\r\nNo, we run each function on each split separated",
"> Because it doesn't make sense to be able to do example = ds[0] or examples = list(ds) on a class named DatasetDict of type Dict[str, Dataset].\r\n\r\nHum but you're still going to raise an exception in both those cases with your current change no? (actually list(ds) would return the name of the splits no?)\r\n\r\n> No, we run each function on each split separated\r\n\r\nNice!"
] |
1,432,477,139
| 5,188
|
add: segmentation guide.
|
closed
| 2022-11-02T04:34:36
| 2022-11-04T18:25:57
| 2022-11-04T18:23:34
|
https://github.com/huggingface/datasets/pull/5188
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5188",
"html_url": "https://github.com/huggingface/datasets/pull/5188",
"diff_url": "https://github.com/huggingface/datasets/pull/5188.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5188.patch",
"merged_at": "2022-11-04T18:23:34"
}
|
sayakpaul
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"Thanks @osanseviero. Am I good to merge? ",
"I would wait for a second approval just in case :) ",
"Sure :) ",
"Merging since the images have been pushed as LFS files ([PR](https://huggingface.co/datasets/huggingface/documentation-images/discussions/8)). "
] |
1,432,375,375
| 5,187
|
chore: add notebook links to img cls and obj det.
|
closed
| 2022-11-02T02:30:09
| 2022-11-03T01:52:24
| 2022-11-03T01:49:56
|
https://github.com/huggingface/datasets/pull/5187
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5187",
"html_url": "https://github.com/huggingface/datasets/pull/5187",
"diff_url": "https://github.com/huggingface/datasets/pull/5187.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5187.patch",
"merged_at": "2022-11-03T01:49:56"
}
|
sayakpaul
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"@nateraw I guess the failing test is unrelated. ",
"@sayakpaul Yea failures are unrelated. ",
"Alright. Will wait for @osanseviero's take and then merge. ",
"FYI @stevhliu ",
"@osanseviero @stevhliu @nateraw thank you for your comments. Acted on them.",
"Thanks! Can I merge? Or should we wait for approvals from the others?",
"Since @stevhliu approved as well, I think you're good to go",
"Alright!\r\n\r\nMerging as a Member for the first time 🫀"
] |
1,432,045,011
| 5,186
|
Incorrect error message when Dataset.from_sql fails and sqlalchemy not installed
|
closed
| 2022-11-01T20:25:51
| 2022-11-15T18:24:39
| 2022-11-15T18:24:39
|
https://github.com/huggingface/datasets/issues/5186
| null |
nateraw
| false
|
[
"Hi! The first `Dataset.from_sql` call also outputs the \"ImportError: Using URI string without sqlalchemy installed.\" message, but you also get \"During handling of the above exception another exception occurred: ...\" after which the ValueError is printed. I agree that this behavior makes it easy to miss the original error. \r\n\r\nI think we can improve this by not throwing the writer's ValueError if the error from a dataset script is already being handled to make debugging easier. @lhoestq @albertvillanova wdyt?",
"Yup ! Alternatively the error can be raised in sql.py before generating the examples ? In `_info` for example",
"yea @lhoestq that would probably be good. The 2nd error is useless if the 1st error is the real reason it failed. "
] |
1,432,021,611
| 5,185
|
Allow passing a subset of output features to Dataset.map
|
open
| 2022-11-01T20:07:20
| 2022-11-01T20:07:34
| null |
https://github.com/huggingface/datasets/issues/5185
| null |
sanderland
| false
|
[] |
1,431,418,066
| 5,183
|
Loading an external dataset in a format similar to conll2003
|
closed
| 2022-11-01T13:18:29
| 2022-11-02T11:57:50
| 2022-11-02T11:57:50
|
https://github.com/huggingface/datasets/issues/5183
| null |
Taghreed7878
| false
|
[] |
1,431,029,547
| 5,182
|
Add notebook / other resource links to the task-specific data loading guides
|
closed
| 2022-11-01T07:57:26
| 2022-11-03T01:49:57
| 2022-11-03T01:49:57
|
https://github.com/huggingface/datasets/issues/5182
| null |
sayakpaul
| false
|
[
"Yea this would be great! We would need an object detection tutorial notebook too if it doesn't already exist there. ",
"There is one: https://huggingface.co/docs/datasets/object_detection.\r\n\r\nI will start the work. "
] |
1,431,027,102
| 5,181
|
Add a guide for semantic segmentation
|
closed
| 2022-11-01T07:54:50
| 2022-11-04T18:23:36
| 2022-11-04T18:23:36
|
https://github.com/huggingface/datasets/issues/5181
| null |
sayakpaul
| false
|
[
"Sure this sounds great! Would this be pure torchvision, albumentations, or something else?",
"I am considering `torchvision` and `albumentations`. Also [works with TensorFlow](https://github.com/deep-diver/segformer-tf-transformers/blob/main/notebooks/TFSegFormer_Finetune.ipynb). \r\n\r\nI am assigning the issue to myself then. "
] |
1,431,012,438
| 5,180
|
An example or recommendations for creating large image datasets?
|
open
| 2022-11-01T07:38:38
| 2022-11-02T10:17:11
| null |
https://github.com/huggingface/datasets/issues/5180
| null |
sayakpaul
| false
|
[
"The beam utilities allow to prepare a dataset as parquet in your cloud storage. From my perspective this CLI is not super easy to use, but we've been working on a new python API to prepare a dataset in your cloud storage:\r\n```python\r\nfrom datasets import load_dataset_builder\r\n\r\nbuilder = load_dataset_builder(\"c4\", \"en\")\r\nbuilder.download_and_prepapre(\"s3://my-bucket/c4\", file_format=\"parquet\")\r\n```\r\n\r\nAnd to use Beam you can do:\r\n```python\r\nbeam_runner = ... # one of \"SparkRunner\", \"DataFlowRunner\", \"DirectRunner\", etc.\r\nbeam_options = ...\r\n\r\nbuilder.download_and_prepapre(\r\n \"s3://my-bucket/c4\",\r\n file_format=\"parquet\",\r\n beam_runner=beam_runner,\r\n beam_options=beam_options\r\n)\r\n```\r\n\r\nThough Beam can be used ONLY if there is a dataset script based on the `BeamBasedBuilder` right now - it doesn't work on an arbitrary dataset (see [wikipedia.py](https://huggingface.co/datasets/wikipedia/blob/main/wikipedia.py) for example).",
"Thanks! \r\n\r\nWould be nice to have something similar for creating large image datasets. "
] |
1,430,826,100
| 5,179
|
`map()` fails midway due to format incompatibility
|
closed
| 2022-11-01T03:57:59
| 2022-11-08T11:35:26
| 2022-11-08T11:35:26
|
https://github.com/huggingface/datasets/issues/5179
| null |
sayakpaul
| false
|
[
"Cc: @lhoestq ",
"You can end up with a list instead of a tensor if all the tensors inside the list can't be stacked together - can you make sure all your inputs are tensors with the same shape ?",
"Is there an easy way to ensure it?",
"You can make sure your `tokenize` function always return tensors of the same shape",
"I modified my `tokenize()` function to be like so:\r\n\r\n```py\r\ndef tokenize(batch):\r\n return tokenizer(batch[\"text\"], padding=\"longest\")\r\n```\r\n\r\nso that the padding always happens w.r.t to the length of the longest sequence in a batch. The issue still persists. Is there any other way? ",
"tbh I though your first implementation was fine\r\n```python\r\ndef tokenize(batch):\r\n return tokenizer(batch[\"text\"], padding=True, truncation=True)\r\n```\r\n\r\nMaybe you can try to see what the erroring data looks like by adding a try/except in `get_test_accuracy` ?",
"This is what I got. \r\n\r\nFor the non-erroring data, it looks like (without the labels):\r\n\r\n```\r\ntensor([[ 101, 10047, 3110, ..., 0, 0, 0],\r\n [ 101, 1045, 2514, ..., 0, 0, 0],\r\n [ 101, 1045, 2514, ..., 0, 0, 0],\r\n ...,\r\n [ 101, 1045, 2005, ..., 0, 0, 0],\r\n [ 101, 1045, 2572, ..., 0, 0, 0],\r\n [ 101, 10047, 7481, ..., 0, 0, 0]]) 128\r\ntensor([[1, 1, 1, ..., 0, 0, 0],\r\n [1, 1, 1, ..., 0, 0, 0],\r\n [1, 1, 1, ..., 0, 0, 0],\r\n ...,\r\n [1, 1, 1, ..., 0, 0, 0],\r\n [1, 1, 1, ..., 0, 0, 0],\r\n [1, 1, 1, ..., 0, 0, 0]]) 128\r\n```\r\n\r\nFor the erroring part:\r\n\r\n```\r\n[tensor([ 101, 1045, 2064, 2102, 2393, 3110, 2066, 2242, 6355, 3047, 2004, 2574,\r\n 2004, 1996, 8629, 2357, 2125, 4299, 1045, 2071, 2424, 2009, 2006, 7858,\r\n 102, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0]), tensor([ 101, 10047, 5458, 1997, 3110, 11654, 1998, 11055, 102, 0,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0]), tensor([ 101, 1045, 2074, 2064, 2102, 6073, 1996, 3110, 2008, 2026,\r\n 14982, 2000, 5587, 2203, 16650, 29563, 2030, 2569, 4506, 2052,\r\n 2191, 1037, 2738, 11552, 2208, 17044, 14540, 2100, 3375, 102,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0]),\r\n...\r\n\r\n[tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\r\n 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]), tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\r\n 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]),\r\n...\r\n```\r\n\r\nI also tried investigating the shapes of the individual entries within a `batch` without the labels:\r\n\r\n```py\r\ndef get_test_accuracy(model):\r\n def fn(batch): \r\n try:\r\n inputs = {k:v.to(device) for k,v in batch.items() \r\n if k in tokenizer.model_input_names}\r\n with torch.no_grad():\r\n output = model(**inputs)\r\n pred_label = torch.argmax(output.logits, axis=-1)\r\n return {\"predicted_label\": pred_label.cpu().numpy()}\r\n except:\r\n for k in batch:\r\n if k != \"label\":\r\n for i in range(len(batch[k])):\r\n print(batch[k][i].shape)\r\n return fn\r\n```\r\n\r\nThey are:\r\n\r\n```\r\n...\r\ntorch.Size([66])\r\ntorch.Size([66])\r\ntorch.Size([66])\r\ntorch.Size([66])\r\ntorch.Size([66])\r\ntorch.Size([66])\r\ntorch.Size([66])\r\ntorch.Size([66])\r\ntorch.Size([66])\r\ntorch.Size([66])\r\ntorch.Size([66])\r\ntorch.Size([66])\r\ntorch.Size([66])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\ntorch.Size([69])\r\n```\r\n\r\nThere are differing shapes. I understand if I set `batch_size=None` in `emotions_encoded = emotions.map(tokenize, batched=True)` the problem should be fixed as the whole dataset would be treated as a single batch. But is there a way to do that in batches? ",
"If you use the same batch_size for your two maps, you should get the exact same batches - therefore all containing the same shapes",
"Oh I see. Thanks. Closing this issue. "
] |
1,430,800,810
| 5,178
|
Unable to download the Chinese `wikipedia`, the dumpstatus.json not found!
|
closed
| 2022-11-01T03:17:55
| 2022-11-02T08:27:15
| 2022-11-02T08:24:29
|
https://github.com/huggingface/datasets/issues/5178
| null |
beyondguo
| false
|
[
"In the dumps page of the wiki (https://dumps.wikimedia.org/zhwiki/), I found the following dumps:\r\n```\r\nIndex of /zhwiki/\r\n[../](https://dumps.wikimedia.org/)\r\n[20220701/](https://dumps.wikimedia.org/zhwiki/20220701/) 21-Aug-2022 01:48 -\r\n[20220720/](https://dumps.wikimedia.org/zhwiki/20220720/) 02-Sep-2022 01:48 -\r\n[20220801/](https://dumps.wikimedia.org/zhwiki/20220801/) 21-Sep-2022 01:44 -\r\n[20220820/](https://dumps.wikimedia.org/zhwiki/20220820/) 01-Oct-2022 09:39 -\r\n[20220901/](https://dumps.wikimedia.org/zhwiki/20220901/) 20-Oct-2022 09:44 -\r\n[20220920/](https://dumps.wikimedia.org/zhwiki/20220920/) 23-Sep-2022 12:06 -\r\n[20221001/](https://dumps.wikimedia.org/zhwiki/20221001/) 04-Oct-2022 15:10 -\r\n[20221020/](https://dumps.wikimedia.org/zhwiki/20221020/) 01-Nov-2022 03:15 -\r\n[latest/](https://dumps.wikimedia.org/zhwiki/latest/) 01-Nov-2022 03:15 -\r\n```\r\n\r\nMaybe the older dumps are not available which caused the downloading failure? \r\n\r\nHowever, when I changed to the newer version:\r\n```\r\ndata = load_dataset('wikipedia', '20220701.zh', beam_runner='DirectRunner')\r\n```\r\n\r\nit shows:\r\n```\r\nValueError: BuilderConfig 20220701.zh not found. Available: ['20220301.aa', '20220301.ab', '20220301.ace', '20220301.ady', '20220301.af', '20220301.ak', '20220301.als', '20220301.am', '20220301.an', '20220301.ang', '20220301.ar', '20220301.arc', '20220301.arz', '20220301.as', '20220301.ast', '20220301.atj', '20220301.av', '20220301.ay', '20220301.az', '20220301.azb', '20220301.ba', '20220301.bar', '20220301.bat-smg', '20220301.bcl', '20220301.be', '20220301.be-x-old', '20220301.bg', '20220301.bh', '20220301.bi', '20220301.bjn', '20220301.bm', '20220301.bn', '20220301.bo', '20220301.bpy', '20220301.br', '20220301.bs', '20220301.bug', '20220301.bxr', '20220301.ca', '20220301.cbk-zam', '20220301.cdo', '20220301.ce', '20220301.ceb', '20220301.ch', '20220301.cho', '20220301.chr', '20220301.chy', '20220301.ckb', '20220301.co', '20220301.cr', '20220301.crh', '20220301.cs', '20220301.csb', '20220301.cu', '20220301.cv', '20220301.cy', '20220301.da', '20220301.de', '20220301.din', '20220301.diq', '20220301.dsb', '20220301.dty', '20220301.dv', '20220301.dz', '20220301.ee', '20220301.el', '20220301.eml', '20220301.en', '20220301.eo', '20220301.es', '20220301.et', '20220301.eu', '20220301.ext', '20220301.fa', '20220301.ff', '20220301.fi', '20220301.fiu-vro', '20220301.fj', '20220301.fo', '20220301.fr', '20220301.frp', '20220301.frr', '20220301.fur', '20220301.fy', '20220301.ga', '20220301.gag', '20220301.gan', '20220301.gd', '20220301.gl', '20220301.glk', '20220301.gn', '20220301.gom', '20220301.gor', '20220301.got', '20220301.gu', '20220301.gv', '20220301.ha', '20220301.hak', '20220301.haw', '20220301.he', '20220301.hi', '20220301.hif', '20220301.ho', '20220301.hr', '20220301.hsb', '20220301.ht', '20220301.hu', '20220301.hy', '20220301.ia', '20220301.id', '20220301.ie', '20220301.ig', '20220301.ii', '20220301.ik', '20220301.ilo', '20220301.inh', '20220301.io', '20220301.is', '20220301.it', '20220301.iu', '20220301.ja', '20220301.jam', '20220301.jbo', '20220301.jv', '20220301.ka', '20220301.kaa', '20220301.kab', '20220301.kbd', '20220301.kbp', '20220301.kg', '20220301.ki', '20220301.kj', '20220301.kk', '20220301.kl', '20220301.km', '20220301.kn', '20220301.ko', '20220301.koi', '20220301.krc', '20220301.ks', '20220301.ksh', '20220301.ku', '20220301.kv', '20220301.kw', '20220301.ky', '20220301.la', '20220301.lad', '20220301.lb', '20220301.lbe', '20220301.lez', '20220301.lfn', '20220301.lg', '20220301.li', '20220301.lij', '20220301.lmo', '20220301.ln', '20220301.lo', '20220301.lrc', '20220301.lt', '20220301.ltg', '20220301.lv', '20220301.mai', '20220301.map-bms', '20220301.mdf', '20220301.mg', '20220301.mh', '20220301.mhr', '20220301.mi', '20220301.min', '20220301.mk', '20220301.ml', '20220301.mn', '20220301.mr', '20220301.mrj', '20220301.ms', '20220301.mt', '20220301.mus', '20220301.mwl', '20220301.my', '20220301.myv', '20220301.mzn', '20220301.na', '20220301.nah', '20220301.nap', '20220301.nds', '20220301.nds-nl', '20220301.ne', '20220301.new', '20220301.ng', '20220301.nl', '20220301.nn', '20220301.no', '20220301.nov', '20220301.nrm', '20220301.nso', '20220301.nv', '20220301.ny', '20220301.oc', '20220301.olo', '20220301.om', '20220301.or', '20220301.os', '20220301.pa', '20220301.pag', '20220301.pam', '20220301.pap', '20220301.pcd', '20220301.pdc', '20220301.pfl', '20220301.pi', '20220301.pih', '20220301.pl', '20220301.pms', '20220301.pnb', '20220301.pnt', '20220301.ps', '20220301.pt', '20220301.qu', '20220301.rm', '20220301.rmy', '20220301.rn', '20220301.ro', '20220301.roa-rup', '20220301.roa-tara', '20220301.ru', '20220301.rue', '20220301.rw', '20220301.sa', '20220301.sah', '20220301.sat', '20220301.sc', '20220301.scn', '20220301.sco', '20220301.sd', '20220301.se', '20220301.sg', '20220301.sh', '20220301.si', '20220301.simple', '20220301.sk', '20220301.sl', '20220301.sm', '20220301.sn', '20220301.so', '20220301.sq', '20220301.sr', '20220301.srn', '20220301.ss', '20220301.st', '20220301.stq', '20220301.su', '20220301.sv', '20220301.sw', '20220301.szl', '20220301.ta', '20220301.tcy', '20220301.te', '20220301.tet', '20220301.tg', '20220301.th', '20220301.ti', '20220301.tk', '20220301.tl', '20220301.tn', '20220301.to', '20220301.tpi', '20220301.tr', '20220301.ts', '20220301.tt', '20220301.tum', '20220301.tw', '20220301.ty', '20220301.tyv', '20220301.udm', '20220301.ug', '20220301.uk', '20220301.ur', '20220301.uz', '20220301.ve', '20220301.vec', '20220301.vep', '20220301.vi', '20220301.vls', '20220301.vo', '20220301.wa', '20220301.war', '20220301.wo', '20220301.wuu', '20220301.xal', '20220301.xh', '20220301.xmf', '20220301.yi', '20220301.yo', '20220301.za', '20220301.zea', '20220301.zh', '20220301.zh-classical', '20220301.zh-min-nan', '20220301.zh-yue', '20220301.zu']\r\n```\r\n\r\nSo I guess adding the latest dumps versions to the `BuilderConfig` may solve the problem? But how to add it?",
"Hi, @beyondguo, thanks for reporting.\r\n\r\nYou have all the information in the dataset card: https://huggingface.co/datasets/wikipedia\r\n\r\n> Then, you can load any subset of Wikipedia per language and per date this way:\r\n> ```python\r\n> from datasets import load_dataset\r\n> \r\n> load_dataset(\"wikipedia\", language=\"sw\", date=\"20220120\", beam_runner=...) \r\n> ```\r\n> where you can pass as beam_runner any Apache Beam supported runner for (distributed) data processing (see [here](https://beam.apache.org/documentation/runners/capability-matrix/)). Pass \"DirectRunner\" to run it on your machine.\r\n> \r\n> You can find the full list of languages and dates [here](https://dumps.wikimedia.org/backup-index.html).\r\n\r\nNote that you have to pass the language and date as keyword arguments, and the available dates depend on the language and can be found on Wikimedia website.",
"Also:\r\n> Some subsets of Wikipedia have already been processed by HuggingFace, and you can load them just with:\r\n> ```python\r\n> load_dataset(\"wikipedia\", \"20220301.en\")\r\n> ```\r\n> The list of pre-processed subsets is:\r\n> - \"20220301.de\"\r\n> - \"20220301.en\"\r\n> - \"20220301.fr\"\r\n> - \"20220301.frr\"\r\n> - \"20220301.it\"\r\n> - \"20220301.simple\""
] |
1,430,238,556
| 5,177
|
Update create image dataset docs
|
closed
| 2022-10-31T17:45:56
| 2022-11-02T17:15:22
| 2022-11-02T17:13:02
|
https://github.com/huggingface/datasets/pull/5177
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5177",
"html_url": "https://github.com/huggingface/datasets/pull/5177",
"diff_url": "https://github.com/huggingface/datasets/pull/5177.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5177.patch",
"merged_at": "2022-11-02T17:13:02"
}
|
stevhliu
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,430,214,539
| 5,176
|
prepare dataset for cloud storage doesn't work
|
closed
| 2022-10-31T17:28:57
| 2023-03-28T09:11:46
| 2023-03-28T09:11:45
|
https://github.com/huggingface/datasets/issues/5176
| null |
araonblake
| false
|
[
"It looks like an issue with `gcsfs`, are you able to instantiate a `GCSFileSystem` manually ?",
"closing since it was probably due to gcsfs"
] |
1,428,696,231
| 5,175
|
Loading an external NER dataset
|
closed
| 2022-10-30T09:31:55
| 2022-11-01T13:15:49
| 2022-11-01T13:15:49
|
https://github.com/huggingface/datasets/issues/5175
| null |
Taghreed7878
| false
|
[] |
1,427,216,416
| 5,174
|
Preserve None in list type cast in PyArrow 10
|
closed
| 2022-10-28T12:48:30
| 2022-10-28T13:15:33
| 2022-10-28T13:13:18
|
https://github.com/huggingface/datasets/pull/5174
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5174",
"html_url": "https://github.com/huggingface/datasets/pull/5174",
"diff_url": "https://github.com/huggingface/datasets/pull/5174.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5174.patch",
"merged_at": "2022-10-28T13:13:18"
}
|
mariosasko
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,425,880,441
| 5,173
|
Raise ffmpeg warnings only once
|
closed
| 2022-10-27T15:58:33
| 2022-10-28T16:03:05
| 2022-10-28T16:00:51
|
https://github.com/huggingface/datasets/pull/5173
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5173",
"html_url": "https://github.com/huggingface/datasets/pull/5173",
"diff_url": "https://github.com/huggingface/datasets/pull/5173.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5173.patch",
"merged_at": "2022-10-28T16:00:51"
}
|
polinaeterna
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,425,523,114
| 5,172
|
Inconsistency behavior between handling local file protocol and other FS protocols
|
open
| 2022-10-27T12:03:20
| 2024-05-08T19:31:13
| null |
https://github.com/huggingface/datasets/issues/5172
| null |
leoleoasd
| false
|
[] |
1,425,355,111
| 5,171
|
Add PB and TB in convert_file_size_to_int
|
closed
| 2022-10-27T09:50:31
| 2022-10-27T12:14:27
| 2022-10-27T12:12:30
|
https://github.com/huggingface/datasets/pull/5171
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5171",
"html_url": "https://github.com/huggingface/datasets/pull/5171",
"diff_url": "https://github.com/huggingface/datasets/pull/5171.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5171.patch",
"merged_at": "2022-10-27T12:12:30"
}
|
lhoestq
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,425,301,835
| 5,170
|
[Caching] Deterministic hashing of torch tensors
|
closed
| 2022-10-27T09:15:15
| 2022-11-02T17:18:43
| 2022-11-02T17:18:43
|
https://github.com/huggingface/datasets/issues/5170
| null |
lhoestq
| false
|
[] |
1,425,075,254
| 5,169
|
Add "ipykernel" to list of `co_filename`s to remove
|
closed
| 2022-10-27T05:56:17
| 2022-11-02T15:46:00
| 2022-11-02T15:43:20
|
https://github.com/huggingface/datasets/pull/5169
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5169",
"html_url": "https://github.com/huggingface/datasets/pull/5169",
"diff_url": "https://github.com/huggingface/datasets/pull/5169.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5169.patch",
"merged_at": "2022-11-02T15:43:20"
}
|
gpucce
| true
|
[
"I don't know how I could add some tests for this, although jupyter is not among the dependencies so at least that would need to be added. If someone can tell a recommended way I will try to do it!",
"So testing by myself and looking around the jupyter codebase it looks like the `co_filename` of objects created within jupyter is of the form `f\"{tempfile.tempdir}/ipykernel_{id1}/{id2}.py\"` however I can't find the exact command setting it so I [asked in discourse](https://discourse.jupyter.org/t/co-filename-within-notebooks/16538). For now adapted the `co_filename` filter and added tests according to this I hope to get an answer and possibly fix based on that.",
"Ok ! I think it's fine to just check if the parent folder is named like `ipykernel_*` then\r\n\r\nsee the source code of how it's created:\r\n\r\nhttps://github.com/ipython/ipykernel/blob/7f73ff705510b35d1e2faad7f5a676c620ce08d4/ipykernel/compiler.py#L72-L75",
"Should look better now didn't notice the duplicated tests",
"_The documentation is not available anymore as the PR was closed or merged._",
"Should work now on windows too",
"I did the changes you suggested and tried to rebase, the first part went fine, the second less so :( \r\n\r\nIf you have time to spare, can you tell me what should I do now to fix this? thanks",
"Instead of rebasing you can just merge `main` into your branch, otherwise the GitHub preview of your PR shows changes of from `main`.\r\n\r\nFeel free to close this PR and create a new one. Or alternatively your can force push to this PR with a new clean git history.",
"I have force-pushed and merged main, only shows the right changes, if you can run CI one more time it should be ok now",
"Hi, sorry I have been busy, the thing is I can't really understand why the test fail, besides the ugly thing I had done in the last commit to check if within CI smth stange happened with `os`, locally tests pass",
"The CI wasn't passing when using the latest version `dill==0.3.6`. We have a separate function to dump CodeType objects for 0.3.6\r\n\r\nI applied the same changes you did to this other function as well - it should be all good now",
"> The CI wasn't passing when using the latest version `dill==0.3.6`. We have a separate function to dump CodeType objects for 0.3.6\r\n> \r\n> I applied the same changes you did to this other function as well - it should be all good now\r\n\r\nThanks, it would have taken a long time to figure out :)"
] |
1,424,368,572
| 5,168
|
Fix CI require beam
|
closed
| 2022-10-26T16:49:33
| 2022-10-27T09:25:19
| 2022-10-27T09:23:26
|
https://github.com/huggingface/datasets/pull/5168
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5168",
"html_url": "https://github.com/huggingface/datasets/pull/5168",
"diff_url": "https://github.com/huggingface/datasets/pull/5168.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5168.patch",
"merged_at": "2022-10-27T09:23:26"
}
|
albertvillanova
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"I'm merging this PR because it is quite a trivial fix and this is required by:\r\n- #5166"
] |
1,424,124,477
| 5,167
|
Add ffmpeg4 installation instructions in warnings
|
closed
| 2022-10-26T14:21:14
| 2022-10-27T09:01:12
| 2022-10-27T08:58:58
|
https://github.com/huggingface/datasets/pull/5167
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5167",
"html_url": "https://github.com/huggingface/datasets/pull/5167",
"diff_url": "https://github.com/huggingface/datasets/pull/5167.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5167.patch",
"merged_at": "2022-10-27T08:58:58"
}
|
polinaeterna
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"To make it warn only once, feel free to use a global counter in python - and if the warning has already been done, you don't do it again",
"> Added the same formatting for the error message :)\r\n\r\nnice!! thank you! \r\n\r\n> Oh and regarding the warning counter, you can do it in another PR maybe ?\r\n\r\nYes, more warnings is better then no warnings.... I'll merge when the CI passes"
] |
1,423,629,582
| 5,166
|
Support dill 0.3.6
|
closed
| 2022-10-26T08:24:59
| 2022-10-28T05:41:05
| 2022-10-28T05:38:14
|
https://github.com/huggingface/datasets/pull/5166
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5166",
"html_url": "https://github.com/huggingface/datasets/pull/5166",
"diff_url": "https://github.com/huggingface/datasets/pull/5166.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5166.patch",
"merged_at": "2022-10-28T05:38:14"
}
|
albertvillanova
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"I think it hasn't been merged ? https://github.com/uqfoundation/dill/pull/501\r\n\r\nThough I can see that the CI is green because it uses dill 0.3.1.1 - we should probably fix the dill version in both CIs:\r\n- use 0.3.1.1 for the CI with the minimum requirements\r\n- use latest for the CI with the latest requirements",
"I have noticed our CI uses `dill-0.3.1.1`, so not really testing dill 0.3.6...",
"The dill version in our CI is due to `apache-beam`...",
"I've tested locally: we need a specific fix for 0.3.6 (different from the previous ones)...",
"I think we can force the version of dill to be whatever we want in the CI - no matter what beam says. The alternative would be to run beam tests separately but it's more work",
"@lhoestq I tried the easiest solution: force dill==0.3.6 ignoring the requirement of apache-beam. But it doesn't work:\r\n- For example, for `tests/test_builder.py::test_beam_based_builder_download_and_prepare_as_parquet`:\r\n```\r\n @dill.dill.register(dill.dill.ModuleType)\r\n def save_module(pickler, obj):\r\n if dill.dill.is_dill(pickler) and obj is pickler._main:\r\n return old_save_module(pickler, obj)\r\n else:\r\n> dill.dill.log.info('M2: %s' % obj)\r\nE AttributeError: module 'dill._dill' has no attribute 'log'\r\n\r\nvenv/lib/python3.9/site-packages/apache_beam/internal/dill_pickler.py:170: AttributeError\r\n```\r\n - Apache Beam registers some dill functions (`save_module`) which are incompatible with dill 0.3.6 (in 0.3.6 'dill._dill' has no attribute 'log' but 'logger')\r\n - This has an impact in CI tests using either Apache Beam or `multiprocess` (even without using Apache Beam!):\r\n```\r\nFAILED tests/test_beam.py::BeamBuilderTest::test_download_and_prepare - AttributeError: module 'dill._dill' has no attribute 'log'\r\nFAILED tests/test_beam.py::BeamBuilderTest::test_nested_features - AttributeError: module 'dill._dill' has no attribute 'log'\r\nFAILED tests/test_arrow_dataset.py::BaseDatasetTest::test_filter_multiprocessing_in_memory - AttributeError: module 'dill._dill' has no attribute 'log'\r\nFAILED tests/test_arrow_dataset.py::BaseDatasetTest::test_filter_multiprocessing_on_disk - AttributeError: module 'dill._dill' has no attribute 'log'\r\nFAILED tests/test_builder.py::test_beam_based_download_and_prepare - AttributeError: module 'dill._dill' has no attribute 'log'\r\nFAILED tests/test_arrow_dataset.py::BaseDatasetTest::test_map_caching_in_memory - AttributeError: module 'dill._dill' has no attribute 'log'\r\nFAILED tests/test_arrow_dataset.py::BaseDatasetTest::test_map_caching_on_disk - AttributeError: module 'dill._dill' has no attribute 'log'\r\nFAILED tests/test_builder.py::test_beam_based_as_dataset - AttributeError: module 'dill._dill' has no attribute 'log'\r\nFAILED tests/test_arrow_dataset.py::BaseDatasetTest::test_map_multiprocessing_in_memory - AttributeError: module 'dill._dill' has no attribute 'log'\r\nFAILED tests/test_arrow_dataset.py::BaseDatasetTest::test_map_multiprocessing_on_disk - AttributeError: module 'dill._dill' has no attribute 'log'\r\nFAILED tests/test_builder.py::test_beam_based_builder_download_and_prepare_as_parquet - AttributeError: module 'dill._dill' has no attribute 'log'\r\n```\r\n\r\nI guess we should implement the other option: run beam tests separately.\r\n\r\nI'm opening another PR for the CI refactoring.",
"Ah crap >< maybe only install apache_beam for the \"minimum requirements\" CI",
"@lhoestq if we install apache-beam only in the \"minimum requirements\" CI, then this other PR should be merged first:\r\n- #5168 \r\n\r\nOtherwise, our CI for \"latest\" will fail because it will try to run the beam tests (because PyTorch is installed but indeed apache-beam is not installed).",
"One of the test is failing because we set \r\n```python\r\n# google colab doesn't allow to pickle loggers\r\n# so we want to make sure each tests passes without pickling the logger\r\ndef reduce_ex(self):\r\n raise pickle.PicklingError()\r\n\r\ndatasets.arrow_dataset.logger.__reduce_ex__ = reduce_ex\r\n```\r\nin `test_arrow_dataset.py` to avoid pickling the logger because it used to fail on google colab.\r\n\r\nNow pickling the logger seems to be working on google colab again - so you can remove it, and it should fix some tests",
"For the other 2 errors:\r\n- FAILED tests/test_arrow_dataset.py::BaseDatasetTest::test_map_caching_in_memory - _pickle.PicklingError: Can't pickle <class 'unittest.mock.MagicMock'>: it's not the same object as unittest.mock.MagicMock\r\n- FAILED tests/test_arrow_dataset.py::BaseDatasetTest::test_map_caching_on_disk - _pickle.PicklingError: Can't pickle <class 'unittest.mock.MagicMock'>: it's not the same object as unittest.mock.MagicMock\r\n\r\nI have implemented a pickable MagicMock."
] |
1,423,616,677
| 5,165
|
Memory explosion when trying to access 4d tensors in datasets cast to torch or np
|
open
| 2022-10-26T08:14:47
| 2022-10-26T08:14:47
| null |
https://github.com/huggingface/datasets/issues/5165
| null |
clefourrier
| false
|
[] |
1,422,813,247
| 5,164
|
WIP: drop labels in Image and Audio folders by default
|
closed
| 2022-10-25T17:21:49
| 2022-11-16T14:21:16
| 2022-11-02T14:03:02
|
https://github.com/huggingface/datasets/pull/5164
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5164",
"html_url": "https://github.com/huggingface/datasets/pull/5164",
"diff_url": "https://github.com/huggingface/datasets/pull/5164.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5164.patch",
"merged_at": null
}
|
polinaeterna
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._",
"close in favor of https://github.com/huggingface/datasets/pull/5192"
] |
1,422,540,337
| 5,163
|
Reduce default max `writer_batch_size`
|
closed
| 2022-10-25T14:14:52
| 2022-10-27T12:19:27
| 2022-10-27T12:16:47
|
https://github.com/huggingface/datasets/pull/5163
|
{
"url": "https://api.github.com/repos/huggingface/datasets/pulls/5163",
"html_url": "https://github.com/huggingface/datasets/pull/5163",
"diff_url": "https://github.com/huggingface/datasets/pull/5163.diff",
"patch_url": "https://github.com/huggingface/datasets/pull/5163.patch",
"merged_at": "2022-10-27T12:16:47"
}
|
mariosasko
| true
|
[
"_The documentation is not available anymore as the PR was closed or merged._"
] |
1,422,461,112
| 5,162
|
Pip-compile: Could not find a version that matches dill<0.3.6,>=0.3.6
|
closed
| 2022-10-25T13:23:50
| 2022-11-14T08:25:37
| 2022-10-28T05:38:15
|
https://github.com/huggingface/datasets/issues/5162
| null |
Rijgersberg
| false
|
[
"Thanks for reporting, @Rijgersberg.\r\n\r\nWe were waiting for the release of `dill` 0.3.6, that happened 2 days ago (24 Oct 2022): https://github.com/uqfoundation/dill/releases/tag/dill-0.3.6\r\n- See comment: https://github.com/huggingface/datasets/pull/4397#discussion_r880629543\r\n\r\nAlso `multiprocess` 0.70.14 was released 2 days ago: https://github.com/uqfoundation/multiprocess/releases/tag/multiprocess-0.70.14\r\n\r\nWe are addressing this issue to align dependencies.",
"In your specific setup, I guess the compatible configuration is with `multiprocess` 0.70.13 (instead of 0.70.14).",
"@Rijgersberg this issue is fixed. It will be available in our next `datasets` release.",
"Thanks!",
"> @Rijgersberg this issue is fixed. It will be available in our next `datasets` release.\n\nAny chance you have a eta? ",
"@StefanSamba we are disussing about making a release early this week.",
"@Rijgersberg, please also that you can make `pip-compile` work by using the backtracking resolver (instead of the legacy one): https://pip-tools.readthedocs.io/en/latest/#a-note-on-resolvers\r\n```\r\npip-compile --resolver=backtracking requirements.in\r\n```\r\nThis resolver will automatically use `multiprocess` 0.70.13 version. "
] |
1,422,371,748
| 5,161
|
Dataset can’t cache model’s outputs
|
closed
| 2022-10-25T12:19:00
| 2022-11-03T16:12:52
| 2022-11-03T16:12:51
|
https://github.com/huggingface/datasets/issues/5161
| null |
jongjyh
| false
|
[
"Addressed in https://github.com/huggingface/datasets/pull/5191 (torch.Tensor objects now produce deterministic hashes)"
] |
1,422,193,938
| 5,160
|
Automatically add filename for image/audio folder
|
open
| 2022-10-25T09:56:49
| 2022-10-26T16:51:46
| null |
https://github.com/huggingface/datasets/issues/5160
| null |
patrickvonplaten
| false
|
[
"Also cc @anton-l ",
"BTW the exact same holds true for the audio folder",
"I'm fine with adding a new column with the file name personally. Not sure how breaking this is though",
"@patrickvonplaten do you mean just filename or full relative path inside the repo?\r\nI think it shouldn't be breaking, at least I cannot come up with any case where it is. Maybe @mariosasko can?\r\n\r\nalso I think that the problem here and in general is that Image/AudioFolder has default configuration which implies automatic label creation if there is not metadata file. It can be changed when you load the dataset with `load_dataset` but not on it's Hub page. \r\n\r\n",
"> also I think that the problem here and in general Image/AudioFolder has default configuration which implies automatic label creation if there is not metadata file\r\n\r\nYea I agree it's often the wrong default. We can also imagine adding the builder's parameters as YAML in the repo.",
"@lhoestq yes I also got the idea of some YAML config! not sure of what priority it is though.",
"but it would actually also solve this issue: https://github.com/huggingface/datasets/issues/5153",
"I meant just the file name (no path) that would already be super helpful IMO :-) (maybe dir+filename if there are dirs in the folder)",
"@patrickvonplaten one more time, to be sure I understand you.\r\nFor example, we have data structure like this:\r\n```\r\n├─ data/\r\n│ └─ subdir/\r\n│ └── cats/\r\n│ ├── 0.jpg\r\n│ ├── 1.jpg\r\n│ └── 2.jpg\r\n│ └── dogs/\r\n│ ├── 0.jpg\r\n│ ├── 1.jpg\r\n│ └── 2.jpg\r\n└── another_subdir/\r\n ├── 10.jpg\r\n ├── 11.jpg\r\n └── 12.jpg\r\n```\r\nIs it okay to provide `\"data/subdir/cats/0.jpg\"`, `\"data/subdir/dogs/0.jpg\"`, `\"data/another_subdir/10.jpg\"`?\r\nI think providing just filenames might be confusing if they are not unique, as in this example. ",
"Yes I think the relative path as you proposed makes a lot of sense :-) "
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.