html_url
stringlengths 48
51
| title
stringlengths 1
290
| comments
sequencelengths 0
30
| body
stringlengths 0
228k
⌀ | number
int64 2
7.08k
|
|---|---|---|---|---|
https://github.com/huggingface/datasets/issues/6396 | Issue with pyarrow 14.0.1 | [
"Looks like we should stop using `PyExtensionType` and use `ExtensionType` instead\r\n\r\nsee https://github.com/apache/arrow/commit/f14170976372436ec1d03a724d8d3f3925484ecf",
"https://github.com/huggingface/datasets-server/pull/2089#pullrequestreview-1724449532\r\n\r\n> Yes, I understand now: they have disabled their `PyExtensionType` and we use it in `datasets` for arrays... ",
"related?\r\n\r\nhttps://huggingface.co/datasets/ssbuild/tools_data/discussions/1#654e663b77c8ec680d10479c",
"> related?\r\n>\r\n> https://huggingface.co/datasets/ssbuild/tools_data/discussions/1#654e663b77c8ec680d10479c\r\n\r\nNo, related to https://github.com/huggingface/datasets/issues/5706",
"Running the following is a workaround:\r\n\r\n```\r\nimport pyarrow\r\npyarrow.PyExtensionType.set_auto_load(True)\r\n```"
] | See https://github.com/huggingface/datasets-server/pull/2089 for reference
```
from datasets import (Array2D, Dataset, Features)
feature_type = Array2D(shape=(2, 2), dtype="float32")
content = [[0.0, 0.0], [0.0, 0.0]]
features = Features({"col": feature_type})
dataset = Dataset.from_dict({"col": [content]}, features=features)
```
generates
```
/home/slesage/hf/datasets-server/libs/libcommon/.venv/lib/python3.9/site-packages/datasets/features/features.py:648: FutureWarning: pyarrow.PyExtensionType is deprecated and will refuse deserialization by default. Instead, please derive from pyarrow.ExtensionType and implement your own serialization mechanism.
pa.PyExtensionType.__init__(self, self.storage_dtype)
/home/slesage/hf/datasets-server/libs/libcommon/.venv/lib/python3.9/site-packages/datasets/features/features.py:1661: RuntimeWarning: pickle-based deserialization of pyarrow.PyExtensionType subclasses is disabled by default; if you only ingest trusted data files, you may re-enable this using `pyarrow.PyExtensionType.set_auto_load(True)`.
In the future, Python-defined extension subclasses should derive from pyarrow.ExtensionType (not pyarrow.PyExtensionType) and implement their own serialization mechanism.
obj = {field.name: generate_from_arrow_type(field.type) for field in pa_schema}
/home/slesage/hf/datasets-server/libs/libcommon/.venv/lib/python3.9/site-packages/datasets/features/features.py:1661: FutureWarning: pyarrow.PyExtensionType is deprecated and will refuse deserialization by default. Instead, please derive from pyarrow.ExtensionType and implement your own serialization mechanism.
obj = {field.name: generate_from_arrow_type(field.type) for field in pa_schema}
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/slesage/hf/datasets-server/libs/libcommon/.venv/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 924, in from_dict
return cls(pa_table, info=info, split=split)
File "/home/slesage/hf/datasets-server/libs/libcommon/.venv/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 693, in __init__
inferred_features = Features.from_arrow_schema(arrow_table.schema)
File "/home/slesage/hf/datasets-server/libs/libcommon/.venv/lib/python3.9/site-packages/datasets/features/features.py", line 1661, in from_arrow_schema
obj = {field.name: generate_from_arrow_type(field.type) for field in pa_schema}
File "/home/slesage/hf/datasets-server/libs/libcommon/.venv/lib/python3.9/site-packages/datasets/features/features.py", line 1661, in <dictcomp>
obj = {field.name: generate_from_arrow_type(field.type) for field in pa_schema}
File "/home/slesage/hf/datasets-server/libs/libcommon/.venv/lib/python3.9/site-packages/datasets/features/features.py", line 1381, in generate_from_arrow_type
return Value(dtype=_arrow_to_datasets_dtype(pa_type))
File "/home/slesage/hf/datasets-server/libs/libcommon/.venv/lib/python3.9/site-packages/datasets/features/features.py", line 111, in _arrow_to_datasets_dtype
raise ValueError(f"Arrow type {arrow_type} does not have a datasets dtype equivalent.")
ValueError: Arrow type extension<arrow.py_extension_type<pyarrow.lib.UnknownExtensionType>> does not have a datasets dtype equivalent.
``` | 6,396 |
https://github.com/huggingface/datasets/issues/6395 | Add ability to set lock type | [
"We've replaced our filelock implementation with the `filelock` package, so their repo is the right place to request this feature.\r\n\r\nIn the meantime, the following should work: \r\n```python\r\nimport filelock\r\nfilelock.FileLock = filelock.SoftFileLock\r\n\r\nimport datasets\r\n...\r\n```"
] | ### Feature request
Allow setting file lock type, maybe from an environment variable
Currently, it only depends on whether fnctl is available:
https://github.com/huggingface/datasets/blob/12ebe695b4748c5a26e08b44ed51955f74f5801d/src/datasets/utils/filelock.py#L463-L470C16
### Motivation
In my environment, flock isn't supported on a network attached drive
### Your contribution
I'll be happy to submit a pr. | 6,395 |
https://github.com/huggingface/datasets/issues/6394 | TorchFormatter images (H, W, C) instead of (C, H, W) format | [
"Here's a PR for that. https://github.com/huggingface/datasets/pull/6402\r\n\r\nIt's not backward compatible, unfortunately. ",
"Just ran into this working on data lib that's attempting to achieve common interfaces across hf datasets, webdataset, native torch style datasets. The defacto standards for image tensors are numpy == HWC, torch.Tensor == CHW. \r\n\r\nI had to drop use of 'torch' formatting because as is (H, W, C) makes it incompatible with pretty much all standard torch vision processing (torchvision, etc) including model inputs themselves... not sure what the breakage scope would be, but might be worth considering a breaking change since I'm not aware of many use cases where a torch.Tensor image is expected to be in HWC form. And if I set the format to 'torch', I'd expect to be able to apply torchvision transforms, etc directly to the output...\r\n\r\nEDIT: For 'torch' output to be compatible with torch conventions (namely torchvision for images), should follow this https://pytorch.org/vision/0.17/transforms.html#supported-input-types-and-conventions\r\n\r\nattn @lhoestq \r\n\r\n",
"We can define something like `.with_format(\"torch\", image_data_format=\"channels_first\")` and recommend using this in the docs maybe ? also cc @NielsRogge ",
"Sounds good to me. I guess it's not allowed to use the channels first format by default for backwards compatibility purposes?",
"This works, but am wondering how widespread the use of the function is for image datasets? My hunch would be that it's not used widely enough with image datasets to favour backwards compat (keeping default channels_last) over clumsiness of needing this to be 'correct' for typical use.. but don't have the data to back that up.",
"I see. I just checked in the HF libraries and it shouldn't break anything. And to be consistent with them we should actually use C H W. For example `transformers` image processors use C H W by default too.\r\n\r\nSo I'm ok with doing a breaking change to make it consistent with `transformers`, `torchvision`, etc.",
"Since it is quite connected, the proposed PR #6402 will not work for monochrome `PIL` images since they only have 2 dimensions as `numpy `arrays. [Torchvision ](https://pytorch.org/vision/stable/_modules/torchvision/transforms/functional.html#pil_to_tensor) adds a channel before permuting. Would that make sense here as well?",
"@Modexus yes, indeed that would make sense as torch expects 1, H, W for monochrome, not H,W as you'd often see in numpy (via PIL), OpenCV, etc.\r\n\r\nThe reference should be the torchvision fn https://pytorch.org/vision/main/_modules/torchvision/transforms/functional.html#pil_to_tensor",
"My PR now should handle monochrome PIL image. Thanks for the heads up :)"
] | ### Describe the bug
Using .set_format("torch") leads to images having shape (H, W, C), the same as in numpy.
However, pytorch normally uses (C, H, W) format.
Maybe I'm missing something but this makes the format a lot less useful as I then have to permute it anyways.
If not using the format it is possible to directly use torchvision transforms but any non-transformed value will not be a tensor.
Is there a reason for this choice?
### Steps to reproduce the bug
```python
from datasets import Dataset, Features, Audio, Image
images = ["path/to/image.png"] * 10
features = Features({"image": Image()})
ds = Dataset.from_dict({"image": images}, features=features)
ds = ds.with_format("torch")
ds[0]["image"].shape
```
```python
torch.Size([512, 512, 4])
```
### Expected behavior
```python
from datasets import Dataset, Features, Audio, Image
images = ["path/to/image.png"] * 10
features = Features({"image": Image()})
ds = Dataset.from_dict({"image": images}, features=features)
ds = ds.with_format("torch")
ds[0]["image"].shape
```
```python
torch.Size([4, 512, 512])
```
### Environment info
- `datasets` version: 2.14.6
- Platform: Linux-6.5.9-100.fc37.x86_64-x86_64-with-glibc2.31
- Python version: 3.11.6
- Huggingface_hub version: 0.18.0
- PyArrow version: 14.0.1
- Pandas version: 2.1.2 | 6,394 |
https://github.com/huggingface/datasets/issues/6393 | Filter occasionally hangs | [
"It looks like I may not be the first to encounter this: https://github.com/huggingface/datasets/issues/3172",
"Adding some more information, it seems to occur more frequently with large (millions of samples) datasets.",
"More information. My code is structured as (1) load (2) map (3) filter (4) filter. It was always the second filter that failed. Combining the two filters into one seems to reliably work.",
"@lhoestq it'd be great if someone had a chance to look at this. I suspect it is impacting many users given the other issue that I linked.",
"Hi ! Sorry for the late response. Was it happening after the first or the second filter ?\r\n\r\nIt looks like an issue with the garbage collector (which makes it random). Maybe datasets created with `filter` are not always handled properly ? cc @mariosasko",
"It was after the second filter (and combining the two filters into one seemingly resolved it). I obviously haven't tried all settings to know that these details are causal, but it did work for me.",
"Thanks, that's good to know.\r\n\r\nThe stacktrace suggests an issue when `del self._indices` is called, which happens when a filtered dataset falls out of scope. The indices are a PyArrow table memory mapped from disk, so I'm not quite sure how calling `del` on it can cause this issue. We do `del self._indices` to make sure the file on disk is not used anymore by the current process and avoid e.g. permission errors.\r\n\r\nHopefully we can find a way to reproduce this error, otherwise it will be quite hard to understand what happened",
"Yeah, I have a reliable repro, but it is not even close to minimal and uses a dataset I can't share. Perhaps you could try getting close to my setting.\r\n\r\n(1) make a large (~20GB) jsonl with prompt/response pairs\r\n(2) load it on a linux machine (`dataset = load_dataset(...)`)\r\n(3) map a tokenizer to it, with multiprocessing (`tokenized_dataset = dataset.map(...)`)\r\n(4) filter it once based on something, with multiprocessing (`filtered_1 = tokenized_dataset.filter(...)`)\r\n(5) filter it again based on something, with multiprocessing (`filtered_2 = filtered_1.filter(...)`)\r\n\r\nI included the variable names just in case it is relevant that I was creating new datasets each time, not overwriting the same variable.",
"@lhoestq I have another version of the repro that seems fairly reliably. I have lots of jsonl files, and I iteratively load each one with `load_dataset('json', data_files='path/to/my/file.jsonl', streaming=False, split='train')` and then `dataset.map(..., num_proc=<int>)`. This iteration hangs in a random place each time. So seems like there is a bug that hits with _some_ frequency.",
"With `num_proc=None` it works fine.",
"I am also having similar issue to #3172 when trying to tokenize the data. My dataset contains 10M samples. Is there anything that could be done without having to split up the processing into multiple datasets?"
] | ### Describe the bug
A call to `.filter` occasionally hangs (after the filter is complete, according to tqdm)
There is a trace produced
```
Exception ignored in: <function Dataset.__del__ at 0x7efb48130c10>
Traceback (most recent call last):
File "/usr/lib/python3/dist-packages/datasets/arrow_dataset.py", line 1366, in __del__
if hasattr(self, "_indices"):
File "/usr/lib/python3/dist-packages/composer/core/engine.py", line 123, in sigterm_handler
sys.exit(128 + signal)
SystemExit: 143
```
but I'm not sure if the trace is actually from `datasets`, or from surrounding code that is trying to clean up after datasets gets stuck.
Unfortunately I can't reproduce this issue anywhere close to reliably. It happens infrequently when using `num_procs > 1`. Anecdotally I started seeing it when using larger datasets (~10M samples).
### Steps to reproduce the bug
N/A see description
### Expected behavior
map/filter calls always complete sucessfully
### Environment info
- `datasets` version: 2.14.6
- Platform: Linux-5.4.0-137-generic-x86_64-with-glibc2.31
- Python version: 3.10.13
- Huggingface_hub version: 0.17.3
- PyArrow version: 13.0.0
- Pandas version: 2.1.2 | 6,393 |
https://github.com/huggingface/datasets/issues/6392 | `push_to_hub` is not robust to hub closing connection | [
"Hi! We made some improvements to `push_to_hub` to make it more robust a couple of weeks ago but haven't published a release in the meantime, so it would help if you could install `datasets` from `main` (`pip install https://github.com/huggingface/datasets`) and let us know if this improved version of `push_to_hub` resolves the issue (in case the `ConnectionError` happens, re-running `push_to_hub` should be faster now).\r\n\r\nAlso, note that the previous implementation retries the upload, but sometimes this is not enough, so re-running the op is the only option.",
"The update helped push more data.\r\nHowever it still crashed a little later:\r\n\r\n```\r\nTraceback (most recent call last):\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_errors.py\", line 270, in hf_raise_for_status\r\n response.raise_for_status()\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/requests/models.py\", line 1021, in raise_for_status\r\n raise HTTPError(http_error_msg, response=self)\r\nrequests.exceptions.HTTPError: 500 Server Error: Internal Server Error for url: https://hf-hub-lfs-us-east-1.s3.us-east-1.amazonaws.com/repos/6c/33/6c33b3be1463a656e43c7a4f2d43c4a1cdae6e9d81fff87f69167ef25ccb1b88/5f53cb57cf2a52ca0d4c2166a69a6714c64fcdbb7cb8936dfa5b11ac60058e5f?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Content-Sha256=UNSIGNED-PAYLOAD&X-Amz-Credential=AKIA2JU7TKAQFN2FTF47%2F20231110%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20231110T011254Z&X-Amz-Expires=86400&X-Amz-Signature=74e3e33c09ac4e7c6ac887aaee8d489f068869abbe1ee6d58a910fb18d0601d4&X-Amz-SignedHeaders=host&partNumber=13&uploadId=kQwunNkunfmT9D8GulQu_ufw1BTZtRA6wEUI4hnYOjytfdf.GKxDETgMr4wm8_0WNF2yGaNco_0h3JAGm4l9KV1N0nqr5XXyUCbs1ROmHP475fn9FIhc1umWQLEDc97V&x-id=UploadPart\r\n\r\nThe above exception was the direct cause of the following exception:\r\n\r\nTraceback (most recent call last):\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/_commit_api.py\", line 391, in _wrapped_lfs_upload\r\n lfs_upload(operation=operation, lfs_batch_action=batch_action, token=token)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/lfs.py\", line 223, in lfs_upload\r\n _upload_multi_part(operation=operation, header=header, chunk_size=chunk_size, upload_url=upload_action[\"href\"])\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/lfs.py\", line 319, in _upload_multi_part\r\n else _upload_parts_iteratively(operation=operation, sorted_parts_urls=sorted_parts_urls, chunk_size=chunk_size)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/lfs.py\", line 376, in _upload_parts_iteratively\r\n hf_raise_for_status(part_upload_res)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_errors.py\", line 330, in hf_raise_for_status\r\n raise HfHubHTTPError(str(e), response=response) from e\r\nhuggingface_hub.utils._errors.HfHubHTTPError: 500 Server Error: Internal Server Error for url: https://hf-hub-lfs-us-east-1.s3.us-east-1.amazonaws.com/repos/6c/33/6c33b3be1463a656e43c7a4f2d43c4a1cdae6e9d81fff87f69167ef25ccb1b88/5f53cb57cf2a52ca0d4c2166a69a6714c64fcdbb7cb8936dfa5b11ac60058e5f?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Content-Sha256=UNSIGNED-PAYLOAD&X-Amz-Credential=AKIA2JU7TKAQFN2FTF47%2F20231110%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20231110T011254Z&X-Amz-Expires=86400&X-Amz-Signature=74e3e33c09ac4e7c6ac887aaee8d489f068869abbe1ee6d58a910fb18d0601d4&X-Amz-SignedHeaders=host&partNumber=13&uploadId=kQwunNkunfmT9D8GulQu_ufw1BTZtRA6wEUI4hnYOjytfdf.GKxDETgMr4wm8_0WNF2yGaNco_0h3JAGm4l9KV1N0nqr5XXyUCbs1ROmHP475fn9FIhc1umWQLEDc97V&x-id=UploadPart\r\n\r\nThe above exception was the direct cause of the following exception:\r\n\r\nTraceback (most recent call last):\r\n File \"convert_to_hf.py\", line 121, in <module>\r\n main()\r\n File \"convert_to_hf.py\", line 109, in main\r\n audio_dataset.push_to_hub(\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/dataset_dict.py\", line 1699, in push_to_hub\r\n split_additions, uploaded_size, dataset_nbytes = self[split]._push_parquet_shards_to_hub(\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/arrow_dataset.py\", line 5215, in _push_parquet_shards_to_hub\r\n _retry(\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/utils/file_utils.py\", line 290, in _retry\r\n return func(*func_args, **func_kwargs)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/hf_api.py\", line 3665, in preupload_lfs_files\r\n _upload_lfs_files(\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_validators.py\", line 118, in _inner_fn\r\n return fn(*args, **kwargs)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/_commit_api.py\", line 401, in _upload_lfs_files\r\n _wrapped_lfs_upload(filtered_actions[0])\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/_commit_api.py\", line 393, in _wrapped_lfs_upload\r\n raise RuntimeError(f\"Error while uploading '{operation.path_in_repo}' to the Hub.\") from exc\r\nRuntimeError: Error while uploading 'batch_20/train-00206-of-00261.parquet' to the Hub.\r\n```",
"I think the previous implementation was actually better: it pushes to the hub every shard. So if it fails, as long as the shards have the same checksum, it will skip the ones that have been pushed.\r\n\r\nThe implementation in `main` pushes commits at the end, so when it fails, there are no commits and therefore restarts from the beginning every time.\r\n\r\nBelow is the another error log from another run with `main`. I've reverting back to the current release as it does the job for me.\r\n\r\n```\r\nUploading the dataset shards: 86%|████████▌ | 224/261 [21:46<03:35, 5.83s/it]s]\r\nTraceback (most recent call last):\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_errors.py\", line 270, in hf_raise_for_status\r\n response.raise_for_status()\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/requests/models.py\", line 1021, in raise_for_status\r\n raise HTTPError(http_error_msg, response=self)\r\nrequests.exceptions.HTTPError: 500 Server Error: Internal Server Error for url: https://hf-hub-lfs-us-east-1.s3.us-east-1.amazonaws.com/repos/6c/33/6c33b3be1463a656e43c7a4f2d43c4a1cdae6e9d81fff87f69167ef25ccb1b88/97e68d7a5d4a747ffaa249fc09798e961d621fe4170599e6100197f7733f321d?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Content-Sha256=UNSIGNED-PAYLOAD&X-Amz-Credential=AKIA2JU7TKAQFN2FTF47%2F20231110%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20231110T145155Z&X-Amz-Expires=86400&X-Amz-Signature=5341e4b34dc325737f92dc9005c4a31e4d3f9a3d3d853b267e01915260acf629&X-Amz-SignedHeaders=host&partNumber=27&uploadId=NRD0izEWv7MPtC2bYrm5VJ4XgIbHctKNguR7zS1UhGOOrXwBJvigrOywBvQBnS9sxiy0J0ma9sNog8S13nIdTdE9p60MIITTstUFeKvLHSxpU.a527QED1JVYzJ.9xA0&x-id=UploadPart\r\n\r\nThe above exception was the direct cause of the following exception:\r\n\r\nTraceback (most recent call last):\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/_commit_api.py\", line 391, in _wrapped_lfs_upload\r\n lfs_upload(operation=operation, lfs_batch_action=batch_action, token=token)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/lfs.py\", line 223, in lfs_upload\r\n _upload_multi_part(operation=operation, header=header, chunk_size=chunk_size, upload_url=upload_action[\"href\"])\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/lfs.py\", line 319, in _upload_multi_part\r\n else _upload_parts_iteratively(operation=operation, sorted_parts_urls=sorted_parts_urls, chunk_size=chunk_size)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/lfs.py\", line 376, in _upload_parts_iteratively\r\n hf_raise_for_status(part_upload_res)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_errors.py\", line 330, in hf_raise_for_status\r\n raise HfHubHTTPError(str(e), response=response) from e\r\nhuggingface_hub.utils._errors.HfHubHTTPError: 500 Server Error: Internal Server Error for url: https://hf-hub-lfs-us-east-1.s3.us-east-1.amazonaws.com/repos/6c/33/6c33b3be1463a656e43c7a4f2d43c4a1cdae6e9d81fff87f69167ef25ccb1b88/97e68d7a5d4a747ffaa249fc09798e961d621fe4170599e6100197f7733f321d?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Content-Sha256=UNSIGNED-PAYLOAD&X-Amz-Credential=AKIA2JU7TKAQFN2FTF47%2F20231110%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20231110T145155Z&X-Amz-Expires=86400&X-Amz-Signature=5341e4b34dc325737f92dc9005c4a31e4d3f9a3d3d853b267e01915260acf629&X-Amz-SignedHeaders=host&partNumber=27&uploadId=NRD0izEWv7MPtC2bYrm5VJ4XgIbHctKNguR7zS1UhGOOrXwBJvigrOywBvQBnS9sxiy0J0ma9sNog8S13nIdTdE9p60MIITTstUFeKvLHSxpU.a527QED1JVYzJ.9xA0&x-id=UploadPart\r\n\r\nThe above exception was the direct cause of the following exception:\r\n\r\nTraceback (most recent call last):\r\n File \"convert_to_hf.py\", line 121, in <module>\r\n main()\r\n File \"convert_to_hf.py\", line 109, in main\r\n audio_dataset.push_to_hub(\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/dataset_dict.py\", line 1699, in push_to_hub\r\n p, glob_pattern_to_regex(PUSH_TO_HUB_WITHOUT_METADATA_CONFIGS_SPLIT_PATTERN_SHARDED)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/arrow_dataset.py\", line 5215, in _push_parquet_shards_to_hub\r\n token = token if token is not None else HfFolder.get_token()\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/utils/file_utils.py\", line 290, in _retry\r\n return func(*func_args, **func_kwargs)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/hf_api.py\", line 3665, in preupload_lfs_files\r\n _upload_lfs_files(\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_validators.py\", line 118, in _inner_fn\r\n return fn(*args, **kwargs)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/_commit_api.py\", line 401, in _upload_lfs_files\r\n _wrapped_lfs_upload(filtered_actions[0])\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/_commit_api.py\", line 393, in _wrapped_lfs_upload\r\n raise RuntimeError(f\"Error while uploading '{operation.path_in_repo}' to the Hub.\") from exc\r\nRuntimeError: Error while uploading 'batch_20/train-00224-of-00261.parquet' to the Hub.\r\n```",
"There's a new error from the hub now:\r\n```\r\nPushing dataset shards to the dataset hub: 49%|████▉ | 128/261 [11:38<12:05, 5.45s/it]\r\nTraceback (most recent call last):\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_errors.py\", line 270, in hf_raise_for_status\r\n response.raise_for_status()\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/requests/models.py\", line 1021, in raise_for_status\r\n raise HTTPError(http_error_msg, response=self)\r\nrequests.exceptions.HTTPError: 429 Client Error: Too Many Requests for url: https://huggingface.co/api/datasets/tarteel-ai/tawseem/commit/main\r\n\r\nThe above exception was the direct cause of the following exception:\r\n\r\nTraceback (most recent call last):\r\n File \"convert_to_hf.py\", line 121, in <module>\r\n main()\r\n File \"convert_to_hf.py\", line 109, in main\r\n audio_dataset.push_to_hub(\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/dataset_dict.py\", line 1641, in push_to_hub\r\n repo_id, split, uploaded_size, dataset_nbytes, _, _ = self[split]._push_parquet_shards_to_hub(\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/arrow_dataset.py\", line 5308, in _push_parquet_shards_to_hub\r\n _retry(\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/utils/file_utils.py\", line 293, in _retry\r\n raise err\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/utils/file_utils.py\", line 290, in _retry\r\n return func(*func_args, **func_kwargs)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_validators.py\", line 118, in _inner_fn\r\n return fn(*args, **kwargs)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/hf_api.py\", line 1045, in _inner\r\n return fn(self, *args, **kwargs)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/hf_api.py\", line 3850, in upload_file\r\n commit_info = self.create_commit(\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_validators.py\", line 118, in _inner_fn\r\n return fn(*args, **kwargs)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/hf_api.py\", line 1045, in _inner\r\n return fn(self, *args, **kwargs)\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/hf_api.py\", line 3237, in create_commit\r\n hf_raise_for_status(commit_resp, endpoint_name=\"commit\")\r\n File \"/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_errors.py\", line 330, in hf_raise_for_status\r\n raise HfHubHTTPError(str(e), response=response) from e\r\nhuggingface_hub.utils._errors.HfHubHTTPError: 429 Client Error: Too Many Requests for url: https://huggingface.co/api/datasets/tarteel-ai/tawseem/commit/main (Request ID: Root=1-654e48e6-598511b14413bb293fa67084;783522b4-66f9-4f8a-8a74-2accf7cabd17)\r\n\r\nYou have exceeded our hourly quotas for action: commit. We invite you to retry later.\r\n```\r\n\r\nAt least this is more explicit from the server side.",
"> think the previous implementation was actually better: it pushes to the hub every shard. So if it fails, as long as the shards have the same checksum, it will skip the ones that have been pushed.\r\n>\r\n>The implementation in main pushes commits at the end, so when it fails, there are no commits and therefore restarts from the beginning every time.\r\n>\r\n>Below is the another error log from another run with main. I've reverting back to the current release as it does the job for me.\r\n\r\nThe `preupload` step is instant for the already uploaded shards, so only the Parquet conversion is repeated without uploading the actual Parquet data (only to check the SHAs). The previous implementation manually checks the Parquet shard's fingerprint to resume uploading, so the current implementation is cleaner.\r\n\r\n> You have exceeded our hourly quotas for action: commit. We invite you to retry later.\r\n\r\nThis is the problem with the previous implementation. If the number of shards is large, it creates too many commits for the Hub in a short period.",
"But I agree that the `500 Server Error` returned by the Hub is annoying. Earlier today, I also got it on a small 5GB dataset (with 500 MB shards).\r\n\r\n@Wauplin @julien-c Is there something we can do about this?",
"@mariosasko can't do much if AWS raises a HTTP 500 unfortunately (we are simply pushing data to a S3 bucket).\r\nWhat we can do is to add a retry mechanism in the multi-part upload logic here: https://github.com/huggingface/huggingface_hub/blob/c972cba1fecb456a7b3325cdd1fdbcc425f21f94/src/huggingface_hub/lfs.py#L370 :confused: ",
"@Wauplin That code already retries the request using `http_backoff`, no?",
"> That code already retries the request using http_backoff, no?\r\n\r\nCurrently only on HTTP 503 by default. We should add 500 as well (and hope it is a transient error from AWS)",
"Opened a PR to retry in case S3 raises HTTP 500. Will also retry on any `ConnectionError` (connection reset by peer, connection lost,...). Hopefully this should make the upload process more robust to transient errors.",
"I still get the same error, using `push_to_hub`. Using `git lfs` and pushing the files solved it for me.",
"@BEpresent the fix has not been released yet. You can expect a release of `huggingface_hub` (with this fix) today or tomorrow :)"
] | ### Describe the bug
Like to #6172, `push_to_hub` will crash if Hub resets the connection and raise the following error:
```
Pushing dataset shards to the dataset hub: 32%|███▏ | 54/171 [06:38<14:23, 7.38s/it]
Traceback (most recent call last):
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/urllib3/connectionpool.py", line 715, in urlopen
httplib_response = self._make_request(
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/urllib3/connectionpool.py", line 467, in _make_request
six.raise_from(e, None)
File "<string>", line 3, in raise_from
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/urllib3/connectionpool.py", line 462, in _make_request
httplib_response = conn.getresponse()
File "/usr/lib/python3.8/http/client.py", line 1348, in getresponse
response.begin()
File "/usr/lib/python3.8/http/client.py", line 316, in begin
version, status, reason = self._read_status()
File "/usr/lib/python3.8/http/client.py", line 285, in _read_status
raise RemoteDisconnected("Remote end closed connection without"
http.client.RemoteDisconnected: Remote end closed connection without response
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/requests/adapters.py", line 486, in send
resp = conn.urlopen(
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/urllib3/connectionpool.py", line 799, in urlopen
retries = retries.increment(
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/urllib3/util/retry.py", line 550, in increment
raise six.reraise(type(error), error, _stacktrace)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/urllib3/packages/six.py", line 769, in reraise
raise value.with_traceback(tb)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/urllib3/connectionpool.py", line 715, in urlopen
httplib_response = self._make_request(
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/urllib3/connectionpool.py", line 467, in _make_request
six.raise_from(e, None)
File "<string>", line 3, in raise_from
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/urllib3/connectionpool.py", line 462, in _make_request
httplib_response = conn.getresponse()
File "/usr/lib/python3.8/http/client.py", line 1348, in getresponse
response.begin()
File "/usr/lib/python3.8/http/client.py", line 316, in begin
version, status, reason = self._read_status()
File "/usr/lib/python3.8/http/client.py", line 285, in _read_status
raise RemoteDisconnected("Remote end closed connection without"
urllib3.exceptions.ProtocolError: ('Connection aborted.', RemoteDisconnected('Remote end closed connection without response'))
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/_commit_api.py", line 383, in _wrapped_lfs_upload
lfs_upload(operation=operation, lfs_batch_action=batch_action, token=token)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/lfs.py", line 223, in lfs_upload
_upload_multi_part(operation=operation, header=header, chunk_size=chunk_size, upload_url=upload_action["href"])
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/lfs.py", line 319, in _upload_multi_part
else _upload_parts_iteratively(operation=operation, sorted_parts_urls=sorted_parts_urls, chunk_size=chunk_size)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/lfs.py", line 375, in _upload_parts_iteratively
part_upload_res = http_backoff("PUT", part_upload_url, data=fileobj_slice)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_http.py", line 258, in http_backoff
response = session.request(method=method, url=url, **kwargs)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/requests/sessions.py", line 589, in request
resp = self.send(prep, **send_kwargs)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/requests/sessions.py", line 703, in send
r = adapter.send(request, **kwargs)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_http.py", line 63, in send
return super().send(request, *args, **kwargs)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/requests/adapters.py", line 501, in send
raise ConnectionError(err, request=request)
requests.exceptions.ConnectionError: (ProtocolError('Connection aborted.', RemoteDisconnected('Remote end closed connection without response')), '(Request ID: 2bab8c06-b701-4266-aead-fe2e0dc0e3ed)')
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "convert_to_hf.py", line 116, in <module>
main()
File "convert_to_hf.py", line 108, in main
audio_dataset.push_to_hub(
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/dataset_dict.py", line 1641, in push_to_hub
repo_id, split, uploaded_size, dataset_nbytes, _, _ = self[split]._push_parquet_shards_to_hub(
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 5308, in _push_parquet_shards_to_hub
_retry(
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/datasets/utils/file_utils.py", line 290, in _retry
return func(*func_args, **func_kwargs)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_validators.py", line 118, in _inner_fn
return fn(*args, **kwargs)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/hf_api.py", line 828, in _inner
return fn(self, *args, **kwargs)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/hf_api.py", line 3221, in upload_file
commit_info = self.create_commit(
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_validators.py", line 118, in _inner_fn
return fn(*args, **kwargs)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/hf_api.py", line 828, in _inner
return fn(self, *args, **kwargs)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/hf_api.py", line 2695, in create_commit
upload_lfs_files(
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/utils/_validators.py", line 118, in _inner_fn
return fn(*args, **kwargs)
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/_commit_api.py", line 393, in upload_lfs_files
_wrapped_lfs_upload(filtered_actions[0])
File "/admin/home-piraka9011/.virtualenvs/w2v2/lib/python3.8/site-packages/huggingface_hub/_commit_api.py", line 385, in _wrapped_lfs_upload
raise RuntimeError(f"Error while uploading '{operation.path_in_repo}' to the Hub.") from exc
RuntimeError: Error while uploading 'batch_19/train-00054-of-00171-932beb4082c034bf.parquet' to the Hub.
```
The function should retry if the operations fails, or at least offer a way to recover after such a failure.
Right now, calling the function again will start sending all the parquets files leading to duplicates in the repository, with no guarantee that it will actually be pushed.
Previously, it would crash with an error 400 #4677 .
### Steps to reproduce the bug
Any large dataset pushed the hub:
```py
audio_dataset.push_to_hub(
repo_id="org/dataset",
)
```
### Expected behavior
`push_to_hub` should have an option for max retries or resume.
### Environment info
- `datasets` version: 2.14.6
- Platform: Linux-5.15.0-1044-aws-x86_64-with-glibc2.29
- Python version: 3.8.10
- Huggingface_hub version: 0.16.4
- PyArrow version: 13.0.0
- Pandas version: 2.0.3 | 6,392 |
https://github.com/huggingface/datasets/issues/6389 | Index 339 out of range for dataset of size 339 <-- save_to_file() | [
"Hi! Can you make the above reproducer self-contained by adding code that generates the data?",
"I managed a workaround eventually but I don't know what it was (I made a lot of changes to seq2seq). I'll try to include generating code in the future. (If I close, I don't know if you see it. Feel free to close; I'll re-open if I encounter it again (if I can))."
] | ### Describe the bug
When saving out some Audio() data.
The data is audio recordings with associated 'sentences'.
(They use the audio 'bytes' approach because they're clips within audio files).
Code is below the traceback (I can't upload the voice audio/text (it's not even me)).
```
Traceback (most recent call last):
File "/mnt/ddrive/prj/voice/voice-training-dataset-create/./dataset.py", line 156, in <module>
create_dataset(args)
File "/mnt/ddrive/prj/voice/voice-training-dataset-create/./dataset.py", line 138, in create_dataset
hf_dataset.save_to_disk(args.outds, max_shard_size='50MB')
File "/home/j/src/py/datasets/src/datasets/arrow_dataset.py", line 1531, in save_to_disk
for kwargs in kwargs_per_job:
File "/home/j/src/py/datasets/src/datasets/arrow_dataset.py", line 1508, in <genexpr>
"shard": self.shard(num_shards=num_shards, index=shard_idx, contiguous=True),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/j/src/py/datasets/src/datasets/arrow_dataset.py", line 4609, in shard
return self.select(
^^^^^^^^^^^^
File "/home/j/src/py/datasets/src/datasets/arrow_dataset.py", line 556, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/j/src/py/datasets/src/datasets/fingerprint.py", line 511, in wrapper
out = func(dataset, *args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/j/src/py/datasets/src/datasets/arrow_dataset.py", line 3797, in select
return self._select_contiguous(start, length, new_fingerprint=new_fingerprint)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/j/src/py/datasets/src/datasets/arrow_dataset.py", line 556, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/j/src/py/datasets/src/datasets/fingerprint.py", line 511, in wrapper
out = func(dataset, *args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/j/src/py/datasets/src/datasets/arrow_dataset.py", line 3857, in _select_contiguous
_check_valid_indices_value(start, len(self))
File "/home/j/src/py/datasets/src/datasets/arrow_dataset.py", line 648, in _check_valid_indices_value
raise IndexError(f"Index {index} out of range for dataset of size {size}.")
IndexError: Index 339 out of range for dataset of size 339.
```
### Steps to reproduce the bug
(I had to set the default max batch size down due to a different bug... or maybe it's related: https://github.com/huggingface/datasets/issues/5717)
```python3
#!/usr/bin/env python3
import argparse
import os
from pathlib import Path
import soundfile as sf
import datasets
datasets.config.DEFAULT_MAX_BATCH_SIZE=35
from datasets import Features, Array2D, Value, Dataset, Sequence, Audio
import numpy as np
import librosa
import sys
import soundfile as sf
import io
import logging
logging.basicConfig(level=logging.DEBUG, filename='debug.log', filemode='w',
format='%(name)s - %(levelname)s - %(message)s')
# Define the arguments for the command-line interface
def parse_args():
parser = argparse.ArgumentParser(description="Create a Huggingface dataset from labeled audio files.")
parser.add_argument("--indir_labeled", action="append", help="Directory containing labeled audio files.", required=True)
parser.add_argument("--outds", help="Path to save the dataset file.", required=True)
parser.add_argument("--max_clips", type=int, help="Max count of audio samples to add to the dataset.", default=None)
parser.add_argument("-r", "--sr", type=int, help="Sample rate for the audio files.", default=16000)
parser.add_argument("--no-resample", action="store_true", help="Disable resampling of the audio files.")
parser.add_argument("--max_clip_secs", type=float, help="Max length of audio clips in seconds.", default=3.0)
parser.add_argument("-v", "--verbose", action='count', default=1, help="Increase verbosity")
return parser.parse_args()
# Convert the NumPy arrays to audio bytes in WAV format
def numpy_to_bytes(audio_array, sampling_rate=16000):
with io.BytesIO() as bytes_io:
sf.write(bytes_io, audio_array, samplerate=sampling_rate,
format='wav', subtype='FLOAT') # float32
return bytes_io.getvalue()
# Function to find audio and label files in a directory
def find_audio_label_pairs(indir_labeled):
audio_label_pairs = []
for root, _, files in os.walk(indir_labeled):
for file in files:
if file.endswith(('.mp3', '.wav', '.aac', '.flac')):
audio_path = Path(root) / file
if args.verbose>1:
print(f'File: {audio_path}')
label_path = audio_path.with_suffix('.labels.txt')
if label_path.exists():
if args.verbose>0:
print(f' Pair: {audio_path}')
audio_label_pairs.append((audio_path, label_path))
return audio_label_pairs
def process_audio_label_pair(audio_path, label_path, sampling_rate, no_resample, max_clip_secs):
# Read the label file
with open(label_path, 'r') as label_file:
labels = label_file.readlines()
# Load the full audio file
full_audio, current_sr = sf.read(audio_path)
if not no_resample and current_sr != sampling_rate:
# You can use librosa.resample here if librosa is available
full_audio = librosa.resample(full_audio, orig_sr=current_sr, target_sr=sampling_rate)
audio_segments = []
sentences = []
# Process each label
for label in labels:
start_secs, end_secs, label_text = label.strip().split('\t')
start_sample = int(float(start_secs) * sampling_rate)
end_sample = int(float(end_secs) * sampling_rate)
# Extract segment and truncate or pad to max_clip_secs
audio_segment = full_audio[start_sample:end_sample]
max_samples = int(max_clip_secs * sampling_rate)
if len(audio_segment) > max_samples: # Truncate
audio_segment = audio_segment[:max_samples]
elif len(audio_segment) < max_samples: # Pad
padding = np.zeros(max_samples - len(audio_segment), dtype=audio_segment.dtype)
audio_segment = np.concatenate((audio_segment, padding))
audio_segment = numpy_to_bytes(audio_segment)
audio_data = {
'path': str(audio_path),
'bytes': audio_segment,
}
audio_segments.append(audio_data)
sentences.append(label_text)
return audio_segments, sentences
# Main function to create the dataset
def create_dataset(args):
audio_label_pairs = []
for indir in args.indir_labeled:
audio_label_pairs.extend(find_audio_label_pairs(indir))
# Initialize our dataset data
dataset_data = {
'path': [], # This will be a list of strings
'audio': [], # This will be a list of dictionaries
'sentence': [], # This will be a list of strings
}
# Process each audio-label pair and add the data to the dataset
for audio_path, label_path in audio_label_pairs[:args.max_clips]:
audio_segments, sentences = process_audio_label_pair(audio_path, label_path, args.sr, args.no_resample, args.max_clip_secs)
if audio_segments and sentences:
for audio_data, sentence in zip(audio_segments, sentences):
if args.verbose>1:
print(f'Appending {audio_data["path"]}')
dataset_data['path'].append(audio_data['path'])
dataset_data['audio'].append({
'path': audio_data['path'],
'bytes': audio_data['bytes'],
})
dataset_data['sentence'].append(sentence)
features = Features({
'path': Value('string'), # Path is redundant in common voice set also
'audio': Audio(sampling_rate=16000),
'sentence': Value('string'),
})
hf_dataset = Dataset.from_dict(dataset_data, features=features)
for key in dataset_data:
for i, item in enumerate(dataset_data[key]):
if item is None or (isinstance(item, bytes) and len(item) == 0):
logging.error(f"Invalid {key} at index {i}: {item}")
import ipdb; ipdb.set_trace(context=16); pass
hf_dataset.save_to_disk(args.outds, max_shard_size='50MB')
# try:
# hf_dataset.save_to_disk(args.outds)
# except TypeError as e:
# # If there's a TypeError, log the exception and the dataset data that might have caused it
# logging.exception("An error occurred while saving the dataset.")
# import ipdb; ipdb.set_trace(context=16); pass
# for key in dataset_data:
# logging.debug(f"{key} length: {len(dataset_data[key])}")
# if key == 'audio':
# # Log the first 100 bytes of the audio data to avoid huge log files
# for i, audio in enumerate(dataset_data[key]):
# logging.debug(f"Audio {i}: {audio['bytes'][:100]}")
# raise
# Run the script
if __name__ == "__main__":
args = parse_args()
create_dataset(args)
```
### Expected behavior
It shouldn't fail.
### Environment info
- `datasets` version: 2.14.7.dev0
- Platform: Linux-6.1.0-13-amd64-x86_64-with-glibc2.36
- Python version: 3.11.2
- `huggingface_hub` version: 0.17.3
- PyArrow version: 13.0.0
- Pandas version: 2.1.2
- `fsspec` version: 2023.9.2
| 6,389 |
https://github.com/huggingface/datasets/issues/6388 | How to create 3d medical imgae dataset? | [] | ### Feature request
I am newer to huggingface, after i look up `datasets` docs, I can't find how to create the dataset contains 3d medical image (ends with '.mhd', '.dcm', '.nii')
### Motivation
help us to upload 3d medical dataset to huggingface!
### Your contribution
I'll submit a PR if I find a way to add this feature | 6,388 |
https://github.com/huggingface/datasets/issues/6387 | How to load existing downloaded dataset ? | [
"Feel free to use `dataset.save_to_disk(...)`, then scp the directory containing the saved dataset and reload it on your other machine using `dataset = load_from_disk(...)`"
] | Hi @mariosasko @lhoestq @katielink
Thanks for your contribution and hard work.
### Feature request
First, I download a dataset as normal by:
```
from datasets import load_dataset
dataset = load_dataset('username/data_name', cache_dir='data')
```
The dataset format in `data` directory will be:
```
-data
|-data_name
|-test-00000-of-00001-bf4c733542e35fcb.parquet
|-train-00000-of-00001-2a1df75c6bce91ab.parquet
```
Then I use SCP to clone this dataset into another machine, and then try:
```
from datasets import load_dataset
dataset = load_dataset('data/data_name') # load from local path
```
This leads to re-generating training and validation split for each time, and the disk quota will be duplicated occupation.
How can I just load the dataset without generating and saving these splits again?
### Motivation
I do not want to download the same dataset in two machines, scp is much faster and better than HuggingFace API. I hope we can directly load the downloaded datasets (.parquest)
### Your contribution
Please refer to the feature | 6,387 |
https://github.com/huggingface/datasets/issues/6386 | Formatting overhead | [
"Ah I think the `line-profiler` log is off-by-one and it is in fact the `extract_batch` method that's taking forever. Will investigate further.",
"I tracked it down to a quirk of my setup. Apologies."
] | ### Describe the bug
Hi! I very recently noticed that my training time is dominated by batch formatting. Using Lightning's profilers, I located the bottleneck within `datasets.formatting.formatting` and then narrowed it down with `line-profiler`. It turns out that almost all of the overhead is due to creating new instances of `self.python_arrow_extractor`. I admit I'm confused why that could be the case - as far as I can tell there's no complex `__init__` logic to execute.
### Steps to reproduce the bug
1. Set up a dataset `ds` with potentially several (4+) columns (not sure if this is necessary, but it did at one point of the investigation make overhead worse)
2. Process it using a custom transform, `ds = ds.with_transform(transform_func)`
3. Decorate this function https://github.com/huggingface/datasets/blob/main/src/datasets/formatting/formatting.py#L512 with `@profile` from https://pypi.org/project/line-profiler/
4. Profile with `$ kernprof -l script_to_profile.py`
### Expected behavior
Batch formatting should have acceptable overhead.
### Environment info
```
datasets=2.14.6
pyarrow=14.0.0
``` | 6,386 |
https://github.com/huggingface/datasets/issues/6385 | Get an error when i try to concatenate the squad dataset with my own dataset | [
"The `answers.text` field in the JSON dataset needs to be a list of strings, not a string.\r\n\r\nSo, here is the fixed code:\r\n```python\r\nfrom huggingface_hub import notebook_login\r\nfrom datasets import load_dataset\r\n\r\n\r\n\r\nnotebook_login(\"mymailadresse\", \"mypassword\")\r\nsquad = load_dataset(\"squad\", split=\"train[:5000]\")\r\nsquad = squad.train_test_split(test_size=0.2)\r\ndataset1 = squad[\"train\"]\r\n\r\n\r\n\r\n\r\nimport json\r\n\r\nmybase = [\r\n {\r\n \"id\": \"1\",\r\n \"context\": \"She lives in Nantes\",\r\n \"question\": \"Where does she live?\",\r\n \"answers\": {\r\n \"text\": [\"Nantes\"],\r\n \"answer_start\": [13],\r\n }\r\n }\r\n]\r\n\r\n\r\n\r\n\r\n# Save the data to a JSON file\r\njson_file_path = r\"data\"\r\nwith open(json_file_path, \"w\", encoding= \"utf-8\") as json_file:\r\n json.dump(mybase, json_file, indent=4)\r\n\r\n\r\n\r\n\r\n# Load the JSON file as a dataset\r\ncustom_dataset = load_dataset(\"json\", data_files=json_file_path, features=dataset1.features)\r\n\r\n\r\n# Access the train split\r\ntrain_dataset = custom_dataset[\"train\"]\r\n\r\n\r\nfrom datasets import concatenate_datasets\r\n\r\n\r\n# Concatenate the datasets\r\nconcatenated_dataset = concatenate_datasets([train_dataset, dataset1])\r\n```",
"Thank you @mariosasko for your help ! It works !"
] | ### Describe the bug
Hello,
I'm new here and I need to concatenate the squad dataset with my own dataset i created. I find the following error when i try to do it: Traceback (most recent call last):
Cell In[9], line 1
concatenated_dataset = concatenate_datasets([train_dataset, dataset1])
File ~\anaconda3\Lib\site-packages\datasets\combine.py:213 in concatenate_datasets
return _concatenate_map_style_datasets(dsets, info=info, split=split, axis=axis)
File ~\anaconda3\Lib\site-packages\datasets\arrow_dataset.py:6002 in _concatenate_map_style_datasets
_check_if_features_can_be_aligned([dset.features for dset in dsets])
File ~\anaconda3\Lib\site-packages\datasets\features\features.py:2122 in _check_if_features_can_be_aligned
raise ValueError(
ValueError: The features can't be aligned because the key answers of features {'id': Value(dtype='string', id=None), 'title': Value(dtype='string', id=None), 'context': Value(dtype='string', id=None), 'question': Value(dtype='string', id=None), 'answers': Sequence(feature={'text': Value(dtype='string', id=None), 'answer_start': Value(dtype='int32', id=None)}, length=-1, id=None)} has unexpected type - Sequence(feature={'text': Value(dtype='string', id=None), 'answer_start': Value(dtype='int32', id=None)}, length=-1, id=None) (expected either {'answer_start': Sequence(feature=Value(dtype='int64', id=None), length=-1, id=None), 'text': Value(dtype='string', id=None)} or Value("null").
### Steps to reproduce the bug
```python
from huggingface_hub import notebook_login
from datasets import load_dataset
notebook_login("mymailadresse", "mypassword")
squad = load_dataset("squad", split="train[:5000]")
squad = squad.train_test_split(test_size=0.2)
dataset1 = squad["train"]
import json
mybase = [
{
"id": "1",
"context": "She lives in Nantes",
"question": "Where does she live?",
"answers": {
"text": "Nantes",
"answer_start": [13],
}
}
]
# Save the data to a JSON file
json_file_path = r"C:\Users\mypath\thefile.json"
with open(json_file_path, "w", encoding= "utf-8") as json_file:
json.dump(mybase, json_file, indent=4)
# Load the JSON file as a dataset
custom_dataset = load_dataset("json", data_files=json_file_path)
# Access the train split
train_dataset = custom_dataset["train"]
from datasets import concatenate_datasets
# Concatenate the datasets
concatenated_dataset = concatenate_datasets([train_dataset, dataset1])
```
### Expected behavior
I would expect the two datasets to be concatenated without error. The len(dataset1) is equal to 4000 and the len(train_dataset) is equal to 1 so I would exepect concatenated_dataset to be created and having lenght 4001.
### Environment info
Python 3.11.4 and using windows
Thank you for your help | 6,385 |
https://github.com/huggingface/datasets/issues/6384 | Load the local dataset folder from other place | [
"Solved"
] | This is from https://github.com/huggingface/diffusers/issues/5573
| 6,384 |
https://github.com/huggingface/datasets/issues/6383 | imagenet-1k downloads over and over | [
"Have you solved this problem?"
] | ### Describe the bug
What could be causing this?
```
$ python3
Python 3.8.13 (default, Mar 28 2022, 11:38:47)
[GCC 7.5.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from datasets import load_dataset
>>> load_dataset("imagenet-1k")
Downloading builder script: 100%|██████████| 4.72k/4.72k [00:00<00:00, 7.51MB/s]
Downloading readme: 100%|███████████████████| 85.4k/85.4k [00:00<00:00, 510kB/s]
Downloading extra modules: 100%|████████████| 46.4k/46.4k [00:00<00:00, 300kB/s]
Downloading data: 100%|████████████████████| 29.1G/29.1G [19:36<00:00, 24.8MB/s]
Downloading data: 100%|████████████████████| 29.3G/29.3G [08:38<00:00, 56.5MB/s]
Downloading data: 100%|████████████████████| 29.0G/29.0G [09:26<00:00, 51.2MB/s]
Downloading data: 100%|████████████████████| 29.2G/29.2G [09:38<00:00, 50.6MB/s]
Downloading data: 100%|███████████████████▉| 29.2G/29.2G [09:37<00:00, 44.1MB/s^Downloading data: 0%| | 106M/29.1G [00:05<23:49, 20.3MB/s]
```
### Steps to reproduce the bug
See above commands/code
### Expected behavior
imagenet-1k is downloaded
### Environment info
- `datasets` version: 2.14.6
- Platform: Linux-6.2.0-34-generic-x86_64-with-glibc2.17
- Python version: 3.8.13
- Huggingface_hub version: 0.15.1
- PyArrow version: 14.0.0
- Pandas version: 1.5.2 | 6,383 |
https://github.com/huggingface/datasets/issues/6382 | Add CheXpert dataset for vision | [
"Hey @SauravMaheshkar ! Just responded to your email.\r\n\r\n_For transparency, copying part of my response here:_\r\nI agree, it would be really great to have this and other BenchMD datasets easily accessible on the hub.\r\n\r\nI think the main limiting factor is that the ChexPert dataset is currently hosted on the Stanford AIMI Shared Datasets website, with a license that does not permit redistribution IIRC. Thus, I believe we would need to create a [dataset loading script](https://huggingface.co/docs/datasets/image_dataset#loading-script) that would check authentication with the Stanford AIMI site before downloading and extracting the data. \r\n\r\nI've started a HF dataset repo [here](https://huggingface.co/datasets/katielink/CheXpert), in case you want to collaborate on writing up this loading script! I'm also happy to take a stab when I have some more time next week.",
"Hey @katielink I would love to try this out. Please guide me.",
"Hi @katielink , I would also love to be on board and contribute to this loading script/project if it is still being developed. I'm interested because I personally would like to gain access to the CheXpert dataset and am facing some weird issues, so I'd like to sort it out for me, and potentially others. Please keep me updated and guide me on this as well!!!"
] | ### Feature request
### Name
**CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison**
### Paper
https://arxiv.org/abs/1901.07031
### Data
https://stanfordaimi.azurewebsites.net/datasets/8cbd9ed4-2eb9-4565-affc-111cf4f7ebe2
### Motivation
CheXpert is one of the fundamental models in medical image classification and can serve as a viable pre-training dataset for radiology classification or low-scale ablation / exploratory studies.
This could also serve as a good pre-training dataset for Kaggle competitions.
### Your contribution
Would love to make a PR and pre-process / get this into 🤗 | 6,382 |
https://github.com/huggingface/datasets/issues/6377 | Support pyarrow 14.0.0 | [] | Support pyarrow 14.0.0 by fixing the root cause of:
- #6374
and revert:
- #6375 | 6,377 |
https://github.com/huggingface/datasets/issues/6376 | Caching problem when deleting a dataset | [
"Thanks for reporting! Can you also share the error message printed in step 5?",
"I did not store it at the time but I'll try to re-do a mwe next week to get it again",
"I haven't managed to reproduce this issue using a [notebook](https://colab.research.google.com/drive/1m6eduYun7pFTkigrCJAFgw0BghlbvXIL?usp=sharing) that follows the steps to reproduce the bug. So, I'm closing it.\r\n\r\nBut feel free to re-open it if you have a better reproducer."
] | ### Describe the bug
Pushing a dataset with n + m features to a repo which was deleted, but contained n features, will fail.
### Steps to reproduce the bug
1. Create a dataset with n features per row
2. `dataset.push_to_hub(YOUR_PATH, SPLIT, token=TOKEN)`
3. Go on the hub, delete the repo at `YOUR_PATH`
4. Update your local dataset to have n + m features per row
5. `dataset.push_to_hub(YOUR_PATH, SPLIT, token=TOKEN)` will fail because of a mismatch in features number
### Expected behavior
Step 5 should work or display a message to indicate the cache has not been cleared
### Environment info
- `datasets` version: 2.12.0
- Platform: Linux-5.15.0-88-generic-x86_64-with-glibc2.31
- Python version: 3.10.10
- Huggingface_hub version: 0.16.4
- PyArrow version: 11.0.0
- Pandas version: 2.0.0
| 6,376 |
https://github.com/huggingface/datasets/issues/6374 | CI is broken: TypeError: Couldn't cast array | [] | See: https://github.com/huggingface/datasets/actions/runs/6730567226/job/18293518039
```
FAILED tests/test_table.py::test_cast_sliced_fixed_size_array_to_features - TypeError: Couldn't cast array of type
fixed_size_list<item: int32>[3]
to
Sequence(feature=Value(dtype='int64', id=None), length=3, id=None)
``` | 6,374 |
https://github.com/huggingface/datasets/issues/6371 | `Dataset.from_generator` should not try to download from HF GCS | [
"Indeed, setting `try_from_gcs` to `False` makes sense for `from_generator`.\r\n\r\nWe plan to deprecate and remove `try_from_hf_gcs` soon, as we can use Hub for file hosting now, but this is a good temporary fix.\r\n"
] | ### Describe the bug
When using [`Dataset.from_generator`](https://github.com/huggingface/datasets/blob/c9c1166e1cf81d38534020f9c167b326585339e5/src/datasets/arrow_dataset.py#L1072) with `streaming=False`, the internal logic will call [`download_and_prepare`](https://github.com/huggingface/datasets/blob/main/src/datasets/io/generator.py#L47) which will attempt to download from HF GCS which is redundant, because user has already provided the generator from which the data should be drawn.
If someone attempts to call `Dataset.from_generator` from an environment that doesn't have external internet access (for example internal production machine) and doesn't set `HF_DATASETS_OFFLINE=1`, this will result in process being stuck at building connection.
### Steps to reproduce the bug
```python
import datasets
def gen():
for _ in range(100):
yield {"text": "dummy text"}
dataset = datasets.Dataset.from_generator(gen)
```
A minimum example executed on any environment that doesn't have access to HF GCS can result in the error
### Expected behavior
`try_from_hf_gcs` should be set to False here https://github.com/huggingface/datasets/blob/c9c1166e1cf81d38534020f9c167b326585339e5/src/datasets/io/generator.py#L51
### Environment info
- `datasets` version: 2.14.4
- Platform: Linux-3.10.0-1160.90.1.el7.x86_64-x86_64-with-glibc2.17
- Python version: 3.10.12
- Huggingface_hub version: 0.17.1
- PyArrow version: 12.0.1
- Pandas version: 2.0.3 | 6,371 |
https://github.com/huggingface/datasets/issues/6370 | TensorDataset format does not work with Trainer from transformers | [
"I figured it out. I found that `Trainer` does not work with TensorDataset even though the document says it uses it. Instead, I ended up creating a dictionary and converting it to a dataset using `dataset.Dataset.from_dict()`.\r\n\r\nI will leave this post open for a while. If someone knows a better approach, please leave a comment.",
"Only issues directly related to the HF datasets library should be reported here. ~So, I'm transferring this issue to the `transformers` repo.~ I'm not a `transformers` maintainer, so GitHub doesn't let me transfer it there :(. This means you need to do it manually."
] | ### Describe the bug
The model was built to do fine tunning on BERT model for relation extraction.
trainer.train() returns an error message ```TypeError: vars() argument must have __dict__ attribute``` when it has `train_dataset` generated from `torch.utils.data.TensorDataset`
However, in the document, the required data format is `torch.utils.data.TensorDataset`.
Transformers trainer is supposed to accept the train_dataset in the format of torch.utils.data.TensorDataset, but it returns error message *"TypeError: vars() argument must have __dict__ attribute"*
```
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
<ipython-input-30-5df728c929a2> in <cell line: 1>()
----> 1 trainer.train()
2 trainer.evaluate(test_dataset)
9 frames
/usr/local/lib/python3.10/dist-packages/transformers/data/data_collator.py in <listcomp>(.0)
107
108 if not isinstance(features[0], Mapping):
--> 109 features = [vars(f) for f in features]
110 first = features[0]
111 batch = {}
TypeError: vars() argument must have __dict__ attribute
```
### Steps to reproduce the bug
Create train_dataset using `torch.utils.data.TensorDataset`, for instance,
```train_dataset = torch.utils.data.TensorDataset(train_input_ids, train_attention_masks, train_labels)```
Feed this `train_dataset` to your trainer and run trainer.train
```
trainer = Trainer(model,
training_args,
train_dataset=train_dataset,
eval_dataset=dev_dataset,
compute_metrics=compute_metrics,
)
```
### Expected behavior
Trainer should start training
### Environment info
It is running on Google Colab
- `datasets` version: 2.14.6
- Platform: Linux-5.15.120+-x86_64-with-glibc2.35
- Python version: 3.10.12
- Huggingface_hub version: 0.17.3
- PyArrow version: 9.0.0
- Pandas version: 1.5.3 | 6,370 |
https://github.com/huggingface/datasets/issues/6369 | Multi process map did not load cache file correctly | [
"The inconsistency may be caused by the usage of \"update_fingerprint\" and setting \"trust_remote_code\" to \"True.\"\r\nWhen the tokenizer employs \"trust_remote_code,\" the behavior of the map function varies with each code execution. Even if the remote code of the tokenizer remains the same, the result of \"asher.hexdigest()\" is found to be inconsistent each time.\r\nThis may result in different processes executing multiple maps\r\n\r\n\r\n\r\n",
"The issue may be related to problems previously discussed in GitHub issues [#3847](https://github.com/huggingface/datasets/issues/3847) and [#6318](https://github.com/huggingface/datasets/pull/6318). \r\nThis arises from the fact that tokenizer.tokens_trie._tokens is an unordered set, leading to varying hash results:\r\n`value = hash_bytes(dumps(tokenizer.tokens_trie._tokens))`\r\nConsequently, this results in different outcomes each time for:\r\n`new_fingerprint = update_fingerprint(datasets._fingerprint, transform, kwargs_for_fingerprint)`\r\n\r\nTo address this issue, it's essential to make `Trie._tokens` a deterministic set while ensuring a consistent order after the final update of `_tokens`.\r\n",
"We now sort `set` and `dict` items to make their hashes deterministic (install from `main` with `pip install git+https://github.com/huggingface/datasets` to test this). Consequently, this should also make the `tokenizer.tokens_trie`'s hash deterministic. Feel free to re-open the issue if this is not the case."
] | ### Describe the bug
When I was training model on Multiple GPUs by DDP, the dataset is tokenized multiple times after main process.
Code is modified from [run_clm.py](https://github.com/huggingface/transformers/blob/7d8ff3629b2725ec43ace99c1a6e87ac1978d433/examples/pytorch/language-modeling/run_clm.py#L484)
### Steps to reproduce the bug
```
block_size = data_args.block_size
IGNORE_INDEX = -100
Ignore_Input = False
def tokenize_function(examples):
sources = []
targets = []
for instruction, inputs, output in zip(examples['instruction'], examples['input'], examples['output']):
source = instruction + inputs
target = f"{output}{tokenizer.eos_token}"
sources.append(source)
targets.append(target)
tokenized_sources = tokenizer(sources, return_attention_mask=False)
tokenized_targets = tokenizer(targets, return_attention_mask=False,
add_special_tokens=False
)
all_input_ids = []
all_labels = []
for s, t in zip(tokenized_sources['input_ids'], tokenized_targets['input_ids']):
if len(s) > block_size and Ignore_Input == False:
# print(s)
continue
input_ids = torch.LongTensor(s + t)[:block_size]
if Ignore_Input:
labels = torch.LongTensor([IGNORE_INDEX] * len(s) + t)[:block_size]
else:
labels = input_ids
assert len(input_ids) == len(labels)
all_input_ids.append(input_ids)
all_labels.append(labels)
results = {
'input_ids': all_input_ids,
'labels': all_labels,
}
return results
with training_args.main_process_first(desc="dataset map tokenization ", local=False):
# print('local_rank',training_args.local_rank)
if not data_args.streaming:
tokenized_datasets = raw_datasets.map(
tokenize_function,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
desc="Running tokenizer on dataset ",
)
else:
tokenized_datasets = raw_datasets.map(
tokenize_function,
batched=True,
remove_columns=column_names,
desc="Running tokenizer on dataset "
)
```
### Expected behavior
This code should only tokenize the dataset in the main process, and the other processes load the dataset after waiting
### Environment info
transformers == 4.34.1
datasets == 2.14.5 | 6,369 |
https://github.com/huggingface/datasets/issues/6366 | with_format() function returns bytes instead of PIL images even when image column is not part of "columns" | [
"Thanks for reporting! I've opened a PR with a fix."
] | ### Describe the bug
When using the with_format() function on a dataset containing images, even if the image column is not part of the columns provided in the function, its type will be changed to bytes.
Here is a minimal reproduction of the bug:
https://colab.research.google.com/drive/1hyaOspgyhB41oiR1-tXE3k_gJCdJUQCf?usp=sharing
### Steps to reproduce the bug
1. Load the image dataset
2. apply with_format(columns=["text"])
3. Check the type of images in the "image" column before and after applying with_format
### Expected behavior
The type should stay the same, but it does not
### Environment info
datasets==2.14.6
| 6,366 |
https://github.com/huggingface/datasets/issues/6365 | Parquet size grows exponential for categorical data | [
"Wrong repo."
] | ### Describe the bug
It seems that when saving a data frame with a categorical column inside the size can grow exponentially.
This seems to happen because when we save the categorical data to parquet, we are saving the data + all the categories existing in the original data. This happens even when the categories are not present in the original data.
### Steps to reproduce the bug
To reproduce the bug, it is enough to run this script:
```
import pandas as pd
import os
if __name__ == "__main__":
for n in [10, 1e2, 1e3, 1e4, 1e5]:
for n_col in [1, 10, 100, 1000, 10000]:
input = pd.DataFrame([{"{i}": f"{i}_cat" for col in range(n_col)} for i in range(int(n))])
input.iloc[0:100].to_parquet("a.parquet")
for col in input.columns:
input[col] = input[col].astype("category")
input.iloc[0:100].to_parquet("b.parquet")
a_size_mb = os.stat("a.parquet").st_size / (1024 * 1024)
b_size_mb = os.stat("b.parquet").st_size / (1024 * 1024)
print(f"{n} {n_col} {a_size_mb} {b_size_mb} {100*b_size_mb/a_size_mb:.2f}")
```
That produces this output:
<img width="464" alt="Screenshot 2023-10-31 at 11 25 25" src="https://github.com/huggingface/datasets/assets/82567957/2b8a9284-7f9e-4c10-a006-0a27236ebd15">
### Expected behavior
In my opinion either:
1. The two file should have (almost) the same size
2. There should be warning telling the user that such difference in size is possible
### Environment info
Python 3.8.18
pandas==2.0.3
numpy==1.24.4 | 6,365 |
https://github.com/huggingface/datasets/issues/6364 | ArrowNotImplementedError: Unsupported cast from string to list using function cast_list | [
"You can use the following code to load this CSV with the list values preserved:\r\n```python\r\nfrom datasets import load_dataset\r\nimport ast\r\n\r\nconverters = {\r\n \"contexts\" : ast.literal_eval,\r\n \"ground_truths\" : ast.literal_eval,\r\n}\r\n\r\nds = load_dataset(\"csv\", data_files=\"golden_dataset.csv\", converters=converters)\r\n```",
"Thank you! it worked :)"
] | Hi,
I am trying to load a local csv dataset(similar to explodinggradients_fiqa) using load_dataset. When I try to pass features, I am facing the mentioned issue.
CSV Data sample(golden_dataset.csv):
Question | Context | answer | groundtruth
"what is abc?" | "abc is this and that" | "abc is this " | "abc is this and that"
```
import csv
# built it based on https://huggingface.co/datasets/explodinggradients/fiqa/viewer/ragas_eval?row=0
mydict = [
{'question' : "what is abc?", 'contexts': ["abc is this and that"], 'answer': "abc is this " , 'groundtruth': ["abc is this and that"]},
{'question' : "what is abc?", 'contexts': ["abc is this and that"], 'answer': "abc is this " , 'groundtruth': ["abc is this and that"]},
{'question' : "what is abc?", 'contexts': ["abc is this and that"], 'answer': "abc is this " , 'groundtruth': ["abc is this and that"]}
]
fields = ['question', 'contexts', 'answer', 'ground_truths']
with open('golden_dataset.csv', 'w', newline='\n') as file:
writer = csv.DictWriter(file, fieldnames = fields)
writer.writeheader()
for row in mydict:
writer.writerow(row)
```
Retrieved dataset:
DatasetDict({
train: Dataset({
features: ['question', 'contexts', 'answer', 'ground_truths'],
num_rows: 1
})
})
Code to reproduce issue:
```
from datasets import load_dataset, Features, Sequence, Value
encode_features = Features(
{
"question": Value(dtype='string', id=0),
"contexts": Sequence(feature=Value(dtype='string', id=1)),
"answer": Value(dtype='string', id=2),
"ground_truths": Sequence(feature=Value(dtype='string',id=3)),
}
)
eval_dataset = load_dataset('csv', data_files='/golden_dataset.csv', features = encode_features )
```
Error trace:
```
---------------------------------------------------------------------------
ArrowNotImplementedError Traceback (most recent call last)
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/datasets/builder.py:1925, in ArrowBasedBuilder._prepare_split_single(self, gen_kwargs, fpath, file_format, max_shard_size, job_id)
1924 _time = time.time()
-> 1925 for _, table in generator:
1926 if max_shard_size is not None and writer._num_bytes > max_shard_size:
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/datasets/packaged_modules/csv/csv.py:192, in Csv._generate_tables(self, files)
189 # Uncomment for debugging (will print the Arrow table size and elements)
190 # logger.warning(f"pa_table: {pa_table} num rows: {pa_table.num_rows}")
191 # logger.warning('\n'.join(str(pa_table.slice(i, 1).to_pydict()) for i in range(pa_table.num_rows)))
--> 192 yield (file_idx, batch_idx), self._cast_table(pa_table)
193 except ValueError as e:
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/datasets/packaged_modules/csv/csv.py:167, in Csv._cast_table(self, pa_table)
165 if all(not require_storage_cast(feature) for feature in self.config.features.values()):
166 # cheaper cast
--> 167 pa_table = pa.Table.from_arrays([pa_table[field.name] for field in schema], schema=schema)
168 else:
169 # more expensive cast; allows str <-> int/float or str to Audio for example
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/pyarrow/table.pxi:3781, in pyarrow.lib.Table.from_arrays()
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/pyarrow/table.pxi:1449, in pyarrow.lib._sanitize_arrays()
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/pyarrow/array.pxi:354, in pyarrow.lib.asarray()
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/pyarrow/table.pxi:551, in pyarrow.lib.ChunkedArray.cast()
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/pyarrow/compute.py:400, in cast(arr, target_type, safe, options, memory_pool)
399 options = CastOptions.safe(target_type)
--> 400 return call_function("cast", [arr], options, memory_pool)
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/pyarrow/_compute.pyx:572, in pyarrow._compute.call_function()
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/pyarrow/_compute.pyx:367, in pyarrow._compute.Function.call()
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/pyarrow/error.pxi:144, in pyarrow.lib.pyarrow_internal_check_status()
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/pyarrow/error.pxi:121, in pyarrow.lib.check_status()
ArrowNotImplementedError: Unsupported cast from string to list using function cast_list
The above exception was the direct cause of the following exception:
DatasetGenerationError Traceback (most recent call last)
Cell In[57], line 1
----> 1 eval_dataset = load_dataset('csv', data_files='/golden_dataset.csv', features = encode_features )
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/datasets/load.py:2153, in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, verification_mode, ignore_verifications, keep_in_memory, save_infos, revision, token, use_auth_token, task, streaming, num_proc, storage_options, **config_kwargs)
2150 try_from_hf_gcs = path not in _PACKAGED_DATASETS_MODULES
2152 # Download and prepare data
-> 2153 builder_instance.download_and_prepare(
2154 download_config=download_config,
2155 download_mode=download_mode,
2156 verification_mode=verification_mode,
2157 try_from_hf_gcs=try_from_hf_gcs,
2158 num_proc=num_proc,
2159 storage_options=storage_options,
2160 )
2162 # Build dataset for splits
2163 keep_in_memory = (
2164 keep_in_memory if keep_in_memory is not None else is_small_dataset(builder_instance.info.dataset_size)
2165 )
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/datasets/builder.py:954, in DatasetBuilder.download_and_prepare(self, output_dir, download_config, download_mode, verification_mode, ignore_verifications, try_from_hf_gcs, dl_manager, base_path, use_auth_token, file_format, max_shard_size, num_proc, storage_options, **download_and_prepare_kwargs)
952 if num_proc is not None:
953 prepare_split_kwargs["num_proc"] = num_proc
--> 954 self._download_and_prepare(
955 dl_manager=dl_manager,
956 verification_mode=verification_mode,
957 **prepare_split_kwargs,
958 **download_and_prepare_kwargs,
959 )
960 # Sync info
961 self.info.dataset_size = sum(split.num_bytes for split in self.info.splits.values())
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/datasets/builder.py:1049, in DatasetBuilder._download_and_prepare(self, dl_manager, verification_mode, **prepare_split_kwargs)
1045 split_dict.add(split_generator.split_info)
1047 try:
1048 # Prepare split will record examples associated to the split
-> 1049 self._prepare_split(split_generator, **prepare_split_kwargs)
1050 except OSError as e:
1051 raise OSError(
1052 "Cannot find data file. "
1053 + (self.manual_download_instructions or "")
1054 + "\nOriginal error:\n"
1055 + str(e)
1056 ) from None
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/datasets/builder.py:1813, in ArrowBasedBuilder._prepare_split(self, split_generator, file_format, num_proc, max_shard_size)
1811 job_id = 0
1812 with pbar:
-> 1813 for job_id, done, content in self._prepare_split_single(
1814 gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
1815 ):
1816 if done:
1817 result = content
File ~/anaconda3/envs/python3/lib/python3.10/site-packages/datasets/builder.py:1958, in ArrowBasedBuilder._prepare_split_single(self, gen_kwargs, fpath, file_format, max_shard_size, job_id)
1956 if isinstance(e, SchemaInferenceError) and e.__context__ is not None:
1957 e = e.__context__
-> 1958 raise DatasetGenerationError("An error occurred while generating the dataset") from e
1960 yield job_id, True, (total_num_examples, total_num_bytes, writer._features, num_shards, shard_lengths)
DatasetGenerationError: An error occurred while generating the dataset
```
Environment Info:
datasets version: 2.14.5
Python version: 3.10.8
PyArrow version: 12.0.1
Pandas version: 2.0.3
I have also tried to load dataset first and then use cast_column, or save_to_disk and load_from_disk. | 6,364 |
https://github.com/huggingface/datasets/issues/6363 | dataset.transform() hangs indefinitely while finetuning the stable diffusion XL | [
"I think the code hangs on the `accelerator.main_process_first()` context manager exit. To verify this, you can append a print statement to the end of the `accelerator.main_process_first()` block. \r\n\r\n\r\nIf the problem is in `with_transform`, it would help if you could share the error stack trace printed when you interrupt the process (while it hangs)",
"@bhosalems Were you able to fix that ? I face this issue as well",
"@matankley No I am not able to resolve this issue yet.",
"@mariosasko yes the problem seems to be to exit from accelerator.main_process_first(). Is there any known problem?",
"NCCL debug info I get below output, if it helps.\r\n```\r\n11/09/2023 13:36:44 - INFO - __main__ - Distributed environment: MULTI_GPU Backend: nccl\r\nNum processes: 2\r\nProcess index: 1\r\nLocal process index: 1\r\nDevice: cuda:1\r\n\r\nMixed precision type: fp16\r\n\r\nDetected kernel version 5.4.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.\r\n11/09/2023 13:36:44 - INFO - __main__ - Distributed environment: MULTI_GPU Backend: nccl\r\nNum processes: 2\r\nProcess index: 0\r\nLocal process index: 0\r\nDevice: cuda:0\r\n\r\nMixed precision type: fp16\r\n\r\n{'timestep_spacing', 'thresholding', 'variance_type', 'clip_sample_range', 'prediction_type', 'dynamic_thresholding_ratio', 'sample_max_value'} was not found in config. Values will be initialized to default values.\r\n{'norm_num_groups', 'force_upcast'} was not found in config. Values will be initialized to default values.\r\n{'num_attention_heads', 'projection_class_embeddings_input_dim', 'addition_embed_type_num_heads', 'mid_block_only_cross_attention', 'addition_embed_type', 'num_class_embeds', 'upcast_attention', 'cross_attention_norm', 'addition_time_embed_dim', 'time_embedding_dim', 'class_embeddings_concat', 'encoder_hid_dim', 'encoder_hid_dim_type', 'resnet_out_scale_factor', 'attention_type', 'conv_out_kernel', 'only_cross_attention', 'resnet_time_scale_shift', 'resnet_skip_time_act', 'reverse_transformer_layers_per_block', 'conv_in_kernel', 'time_cond_proj_dim', 'use_linear_projection', 'mid_block_type', 'time_embedding_act_fn', 'dropout', 'timestep_post_act', 'dual_cross_attention', 'class_embed_type', 'transformer_layers_per_block', 'time_embedding_type'} was not found in config. Values will be initialized to default values.\r\n{'num_attention_heads', 'projection_class_embeddings_input_dim', 'addition_embed_type_num_heads', 'mid_block_only_cross_attention', 'addition_embed_type', 'num_class_embeds', 'upcast_attention', 'cross_attention_norm', 'addition_time_embed_dim', 'time_embedding_dim', 'class_embeddings_concat', 'encoder_hid_dim', 'encoder_hid_dim_type', 'resnet_out_scale_factor', 'attention_type', 'conv_out_kernel', 'only_cross_attention', 'resnet_time_scale_shift', 'resnet_skip_time_act', 'reverse_transformer_layers_per_block', 'conv_in_kernel', 'time_cond_proj_dim', 'use_linear_projection', 'mid_block_type', 'time_embedding_act_fn', 'dropout', 'timestep_post_act', 'dual_cross_attention', 'class_embed_type', 'transformer_layers_per_block', 'time_embedding_type'} was not found in config. Values will be initialized to default values.\r\ndeepbull5:1311249:1311249 [0] NCCL INFO Bootstrap : Using enp194s0f0:128.205.43.171<0>\r\ndeepbull5:1311249:1311249 [0] NCCL INFO NET/Plugin : No plugin found (libnccl-net.so), using internal implementation\r\ndeepbull5:1311249:1311249 [0] NCCL INFO cudaDriverVersion 11070\r\nNCCL version 2.14.3+cuda11.7\r\ndeepbull5:1311250:1311250 [1] NCCL INFO cudaDriverVersion 11070\r\ndeepbull5:1311249:1311365 [0] NCCL INFO NET/IB : No device found.\r\ndeepbull5:1311249:1311365 [0] NCCL INFO NET/Socket : Using [0]enp194s0f0:128.205.43.171<0>\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Using network Socket\r\ndeepbull5:1311250:1311250 [1] NCCL INFO Bootstrap : Using enp194s0f0:128.205.43.171<0>\r\ndeepbull5:1311250:1311250 [1] NCCL INFO NET/Plugin : No plugin found (libnccl-net.so), using internal implementation\r\ndeepbull5:1311250:1311366 [1] NCCL INFO NET/IB : No device found.\r\ndeepbull5:1311250:1311366 [1] NCCL INFO NET/Socket : Using [0]enp194s0f0:128.205.43.171<0>\r\ndeepbull5:1311250:1311366 [1] NCCL INFO Using network Socket\r\ndeepbull5:1311250:1311366 [1] NCCL INFO Setting affinity for GPU 1 to ff,ffff0000,00ffffff\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Setting affinity for GPU 0 to ff,ffff0000,00ffffff\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Channel 00/04 : 0 1\r\ndeepbull5:1311250:1311366 [1] NCCL INFO Trees [0] -1/-1/-1->1->0 [1] 0/-1/-1->1->-1 [2] -1/-1/-1->1->0 [3] 0/-1/-1->1->-1\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Channel 01/04 : 0 1\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Channel 02/04 : 0 1\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Channel 03/04 : 0 1\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Trees [0] 1/-1/-1->0->-1 [1] -1/-1/-1->0->1 [2] 1/-1/-1->0->-1 [3] -1/-1/-1->0->1\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Channel 00/0 : 0[1000] -> 1[24000] via P2P/IPC\r\ndeepbull5:1311250:1311366 [1] NCCL INFO Channel 00/0 : 1[24000] -> 0[1000] via P2P/IPC\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Channel 01/0 : 0[1000] -> 1[24000] via P2P/IPC\r\ndeepbull5:1311250:1311366 [1] NCCL INFO Channel 01/0 : 1[24000] -> 0[1000] via P2P/IPC\r\ndeepbull5:1311250:1311366 [1] NCCL INFO Channel 02/0 : 1[24000] -> 0[1000] via P2P/IPC\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Channel 02/0 : 0[1000] -> 1[24000] via P2P/IPC\r\ndeepbull5:1311250:1311366 [1] NCCL INFO Channel 03/0 : 1[24000] -> 0[1000] via P2P/IPC\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Channel 03/0 : 0[1000] -> 1[24000] via P2P/IPC\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Connected all rings\r\ndeepbull5:1311249:1311365 [0] NCCL INFO Connected all trees\r\ndeepbull5:1311249:1311365 [0] NCCL INFO threadThresholds 8/8/64 | 16/8/64 | 512 | 512\r\ndeepbull5:1311249:1311365 [0] NCCL INFO 4 coll channels, 4 p2p channels, 2 p2p channels per peer\r\ndeepbull5:1311250:1311366 [1] NCCL INFO Connected all rings\r\ndeepbull5:1311250:1311366 [1] NCCL INFO Connected all trees\r\ndeepbull5:1311250:1311366 [1] NCCL INFO threadThresholds 8/8/64 | 16/8/64 | 512 | 512\r\ndeepbull5:1311250:1311366 [1] NCCL INFO 4 coll channels, 4 p2p channels, 2 p2p channels per peer\r\ndeepbull5:1311249:1311365 [0] NCCL INFO comm 0x88a84ee0 rank 0 nranks 2 cudaDev 0 busId 1000 - Init COMPLETE\r\ndeepbull5:1311250:1311366 [1] NCCL INFO comm 0x89a42f60 rank 1 nranks 2 cudaDev 1 busId 24000 - Init COMPLETE\r\n\r\n```",
"Maybe @muellerzr can help as an `accelerate` maintainer.",
"I don't know what the issue was, but after going through the thread here I loved the issue with https://github.com/huggingface/accelerate/issues/314#issuecomment-1565259831"
] | ### Describe the bug
Multi-GPU fine-tuning the stable diffusion X by following https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/README_sdxl.md hangs indefinitely.
### Steps to reproduce the bug
accelerate launch train_text_to_image_sdxl.py --pretrained_model_name_or_path=$MODEL_NAME --pretrained_vae_model_name_or_path=$VAE_NAME --dataset_name=$DATASET_NAME --enable_xformers_memory_efficient_attention --resolution=512 --center_crop --random_flip --proportion_empty_prompts=0.2 --train_batch_size=1 --gradient_accumulation_steps=4 --gradient_checkpointing --max_train_steps=10000 --use_8bit_adam --learning_rate=1e-06 --lr_scheduler="constant" --lr_warmup_steps=0 --mixed_precision="fp16" --report_to="wandb" --validation_prompt="a cute Sundar Pichai creature" --validation_epochs 5 --checkpointing_steps=5000 --output_dir="sdxl-pokemon-model"
### Expected behavior
It should start the training as it does for the single GPU training. I opened the issue in diffusers **https://github.com/huggingface/diffusers/issues/5534 but it does seem to be an issue with the Pokemon dataset.
I added some debug prints
```
print("==========HERE3=============")
with accelerator.main_process_first():
print(accelerator.is_main_process)
print("===========Here3.1===========")
if args.max_train_samples is not None:
dataset["train"] = dataset["train"].shuffle(seed=args.seed).select(range(args.max_train_samples))
print("===========Here3.2===========")
# Set the training transforms
train_dataset = dataset["train"].with_transform(preprocess_train)
print("==========HERE4=============")
Corresponding Output
Detected kernel version 5.4.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.
10/25/2023 21:18:04 - INFO - main - Distributed environment: MULTI_GPU Backend: nccl
Num processes: 3
Process index: 1
Local process index: 1
Device: cuda:1
Mixed precision type: fp16
10/25/2023 21:18:04 - INFO - main - Distributed environment: MULTI_GPU Backend: nccl
Num processes: 3
Process index: 2
Local process index: 2
Device: cuda:2
Mixed precision type: fp16
10/25/2023 21:18:04 - INFO - main - Distributed environment: MULTI_GPU Backend: nccl
Num processes: 3
Process index: 0
Local process index: 0
Device: cuda:0
Mixed precision type: fp16
You are using a model of type clip_text_model to instantiate a model of type . This is not supported for all configurations of models and can yield errors.
You are using a model of type clip_text_model to instantiate a model of type . This is not supported for all configurations of models and can yield errors.
{‘variance_type’, ‘clip_sample_range’, ‘thresholding’, ‘dynamic_thresholding_ratio’} was not found in config. Values will be initialized to default values.
{‘attention_type’, ‘reverse_transformer_layers_per_block’, ‘dropout’} was not found in config. Values will be initialized to default values.
==========HERE1=============
==========HERE1=============
==========HERE1=============
==========HERE2=============
==========HERE2=============
==========HERE2=============
==========HERE3=============
True
===========Here3.1===========
===========Here3.2===========
==========HERE3=============
==========HERE3=========
```
### Environment info
_libgcc_mutex 0.1 conda_forge conda-forge
_openmp_mutex 4.5 2_kmp_llvm conda-forge
absl-py 2.0.0 pypi_0 pypi
accelerate 0.24.0 pypi_0 pypi
aiohttp 3.8.6 pypi_0 pypi
aiosignal 1.3.1 pypi_0 pypi
appdirs 1.4.4 pyh9f0ad1d_0 conda-forge
async-timeout 4.0.3 pypi_0 pypi
attrs 23.1.0 pypi_0 pypi
bitsandbytes 0.41.1 pypi_0 pypi
blas 1.0 mkl
blessings 1.7 py39h06a4308_1002
brotli-python 1.0.9 py39h6a678d5_7
bzip2 1.0.8 h7b6447c_0
ca-certificates 2023.08.22 h06a4308_0
cachetools 5.3.2 pypi_0 pypi
certifi 2023.7.22 py39h06a4308_0
cffi 1.15.1 py39h5eee18b_3
charset-normalizer 2.0.4 pyhd3eb1b0_0
click 8.1.7 unix_pyh707e725_0 conda-forge
cryptography 41.0.3 py39hdda0065_0
cuda-cudart 11.7.99 0 nvidia
cuda-cupti 11.7.101 0 nvidia
cuda-libraries 11.7.1 0 nvidia
cuda-nvrtc 11.7.99 0 nvidia
cuda-nvtx 11.7.91 0 nvidia
cuda-runtime 11.7.1 0 nvidia
datasets 2.14.6 pypi_0 pypi
diffusers 0.22.0.dev0 pypi_0 pypi
dill 0.3.7 pypi_0 pypi
docker-pycreds 0.4.0 py_0 conda-forge
ffmpeg 4.3 hf484d3e_0 pytorch
filelock 3.12.4 pypi_0 pypi
freetype 2.12.1 h4a9f257_0
frozenlist 1.4.0 pypi_0 pypi
fsspec 2023.10.0 pypi_0 pypi
ftfy 6.1.1 pypi_0 pypi
giflib 5.2.1 h5eee18b_3
gitdb 4.0.11 pyhd8ed1ab_0 conda-forge
gitpython 3.1.40 pyhd8ed1ab_0 conda-forge
gmp 6.2.1 h295c915_3
gnutls 3.6.15 he1e5248_0
google-auth 2.23.3 pypi_0 pypi
google-auth-oauthlib 1.1.0 pypi_0 pypi
gpustat 0.6.0 pyhd3eb1b0_1
grpcio 1.59.0 pypi_0 pypi
huggingface-hub 0.17.3 pypi_0 pypi
idna 3.4 py39h06a4308_0
importlib-metadata 6.8.0 pypi_0 pypi
intel-openmp 2023.1.0 hdb19cb5_46305
jinja2 3.1.2 pypi_0 pypi
jpeg 9e h5eee18b_1
lame 3.100 h7b6447c_0
lcms2 2.12 h3be6417_0
ld_impl_linux-64 2.38 h1181459_1
lerc 3.0 h295c915_0
libcublas 11.10.3.66 0 nvidia
libcufft 10.7.2.124 h4fbf590_0 nvidia
libcufile 1.8.0.34 0 nvidia
libcurand 10.3.4.52 0 nvidia
libcusolver 11.4.0.1 0 nvidia
libcusparse 11.7.4.91 0 nvidia
libdeflate 1.17 h5eee18b_1
libffi 3.4.4 h6a678d5_0
libgcc-ng 13.2.0 h807b86a_2 conda-forge
libgfortran-ng 13.2.0 h69a702a_2 conda-forge
libgfortran5 13.2.0 ha4646dd_2 conda-forge
libiconv 1.16 h7f8727e_2
libidn2 2.3.4 h5eee18b_0
libnpp 11.7.4.75 0 nvidia
libnvjpeg 11.8.0.2 0 nvidia
libpng 1.6.39 h5eee18b_0
libprotobuf 3.20.3 he621ea3_0
libstdcxx-ng 13.2.0 h7e041cc_2 conda-forge
libtasn1 4.19.0 h5eee18b_0
libtiff 4.5.1 h6a678d5_0
libunistring 0.9.10 h27cfd23_0
libwebp 1.3.2 h11a3e52_0
libwebp-base 1.3.2 h5eee18b_0
llvm-openmp 14.0.6 h9e868ea_0
lz4-c 1.9.4 h6a678d5_0
markdown 3.5 pypi_0 pypi
markupsafe 2.1.3 pypi_0 pypi
mkl 2023.1.0 h213fc3f_46343
mkl-service 2.4.0 py39h5eee18b_1
mkl_fft 1.3.8 py39h5eee18b_0
mkl_random 1.2.4 py39hdb19cb5_0
multidict 6.0.4 pypi_0 pypi
multiprocess 0.70.15 pypi_0 pypi
ncurses 6.4 h6a678d5_0
nettle 3.7.3 hbbd107a_1
numpy 1.26.0 py39h5f9d8c6_0
numpy-base 1.26.0 py39hb5e798b_0
nvidia-ml 7.352.0 pyhd3eb1b0_0
oauthlib 3.2.2 pypi_0 pypi
openh264 2.1.1 h4ff587b_0
openjpeg 2.4.0 h3ad879b_0
openssl 3.0.11 h7f8727e_2
packaging 23.2 pypi_0 pypi
pandas 2.1.1 pypi_0 pypi
pathtools 0.1.2 py_1 conda-forge
pillow 10.0.1 py39ha6cbd5a_0
pip 23.3 py39h06a4308_0
protobuf 4.23.4 pypi_0 pypi
psutil 5.9.6 pypi_0 pypi
pyarrow 13.0.0 pypi_0 pypi
pyasn1 0.5.0 pypi_0 pypi
pyasn1-modules 0.3.0 pypi_0 pypi
pycparser 2.21 pyhd3eb1b0_0
pyopenssl 23.2.0 py39h06a4308_0
pysocks 1.7.1 py39h06a4308_0
python 3.9.18 h955ad1f_0
python-dateutil 2.8.2 pypi_0 pypi
python_abi 3.9 2_cp39 conda-forge
pytorch 1.13.1 py3.9_cuda11.7_cudnn8.5.0_0 pytorch
pytorch-cuda 11.7 h778d358_5 pytorch
pytorch-mutex 1.0 cuda pytorch
pytz 2023.3.post1 pypi_0 pypi
pyyaml 6.0.1 pypi_0 pypi
readline 8.2 h5eee18b_0
regex 2023.10.3 pypi_0 pypi
requests 2.31.0 py39h06a4308_0
requests-oauthlib 1.3.1 pypi_0 pypi
rsa 4.9 pypi_0 pypi
safetensors 0.4.0 pypi_0 pypi
scipy 1.11.3 py39h5f9d8c6_0
sentry-sdk 1.32.0 pyhd8ed1ab_0 conda-forge
setproctitle 1.1.10 py39h3811e60_1004 conda-forge
setuptools 68.0.0 py39h06a4308_0
six 1.16.0 pyh6c4a22f_0 conda-forge
smmap 5.0.0 pyhd8ed1ab_0 conda-forge
sqlite 3.41.2 h5eee18b_0
tbb 2021.8.0 hdb19cb5_0
tensorboard 2.15.0 pypi_0 pypi
tensorboard-data-server 0.7.2 pypi_0 pypi
tk 8.6.12 h1ccaba5_0
tokenizers 0.14.1 pypi_0 pypi
torchaudio 0.13.1 py39_cu117 pytorch
torchtriton 2.1.0 py39 pytorch
torchvision 0.14.1 py39_cu117 pytorch
tqdm 4.66.1 pypi_0 pypi
transformers 4.34.1 pypi_0 pypi
typing_extensions 4.7.1 py39h06a4308_0
tzdata 2023.3 pypi_0 pypi
urllib3 1.26.18 py39h06a4308_0
wandb 0.15.12 pyhd8ed1ab_0 conda-forge
wcwidth 0.2.8 pypi_0 pypi
werkzeug 3.0.1 pypi_0 pypi
wheel 0.41.2 py39h06a4308_0
xformers 0.0.22.post7 py39_cu11.7.1_pyt1.13.1 xformers
xxhash 3.4.1 pypi_0 pypi
xz 5.4.2 h5eee18b_0
yaml 0.2.5 h7f98852_2 conda-forge
yarl 1.9.2 pypi_0 pypi
zipp 3.17.0 pypi_0 pypi
zlib 1.2.13 h5eee18b_0
zstd 1.5.5 hc292b87_0 | 6,363 |
https://github.com/huggingface/datasets/issues/6360 | Add support for `Sequence(Audio/Image)` feature in `push_to_hub` | [
"This issue stems from https://github.com/huggingface/datasets/blob/6d2f2a5e0fea3827eccfd1717d8021c15fc4292a/src/datasets/table.py#L2203-L2205\r\n\r\nI'll address it as part of https://github.com/huggingface/datasets/pull/6283.\r\n\r\nIn the meantime, this should work\r\n\r\n```python\r\nimport pyarrow as pa\r\nfrom datasets import Image\r\n\r\ndataset = dataset.with_format(\"arrow\")\r\n\r\ndef embed_images(pa_table):\r\n images_arr = pa.chunked_array(\r\n [\r\n pa.ListArray.from_arrays(chunk.offsets, Image().embed_storage(chunk.values), mask=chunk.is_null())\r\n for chunk in pa_table[\"images\"].chunks\r\n ]\r\n )\r\n return pa_table.set_column(pa_table.schema.get_field_index(\"images\"), \"images\", images_arr)\r\n\r\ndataset = dataset.map(embed_images, batched=True)\r\n\r\ndataset = dataset.with_format(\"python\")\r\n\r\ndataset.push_to_hub(...)\r\n```"
] | ### Feature request
Allow for `Sequence` of `Image` (or `Audio`) to be embedded inside the shards.
### Motivation
Currently, thanks to #3685, when `embed_external_files` is set to True (which is the default) in `push_to_hub`, features of type `Image` and `Audio` are embedded inside the arrow/parquet shards, instead of only storing paths to the files.
I've noticed that this behavior does not extend to `Sequence` of `Image`, when working with a [dataset of timelapse images](https://huggingface.co/datasets/1aurent/Human-Embryo-Timelapse).
### Your contribution
I'll submit a PR if I find a way to add this feature | 6,360 |
https://github.com/huggingface/datasets/issues/6359 | Stuck in "Resolving data files..." | [
"Most likely, the data file inference logic is the problem here.\r\n\r\nYou can run the following code to verify this:\r\n```python\r\nimport time\r\nfrom datasets.data_files import get_data_patterns\r\nstart_time = time.time()\r\nget_data_patterns(\"/path/to/img_dir\")\r\nend_time = time.time()\r\nprint(f\"Elapsed time: {end_time - start_time:.2f}s\")\r\n```\r\n \r\nWe plan to optimize this for the next version (or version after that). In the meantime, specifying the split patterns manually should give better performance:\r\n```python\r\nds = load_dataset(\"imagefolder\", data_files={\"train\": \"path/to/img_dir/train/**\", ...}, split=\"train\")\r\n```",
"Hi, @mariosasko, you are right; data file inference logic is extremely slow.\r\n\r\nI have done a similar test, that is I modify the source code of datasets/load.py to measure the cost of two suspicious operations:\r\n```python\r\ndef get_module(self) -> DatasetModule:\r\n base_path = Path(self.data_dir or \"\").expanduser().resolve().as_posix()\r\n start = time.time()\r\n patterns = sanitize_patterns(self.data_files) if self.data_files is not None else get_data_patterns(base_path)\r\n print(f\"patterns: {time.time() - start}\")\r\n start = time.time()\r\n data_files = DataFilesDict.from_patterns(\r\n patterns,\r\n download_config=self.download_config,\r\n base_path=base_path,\r\n )\r\n print(f\"data_files: {time.time() - start}\")\r\n```\r\nIt gaves:\r\npatterns: 3062.2050700187683\r\ndata_files: 413.9576675891876\r\n\r\nThus, these two operations contribute to almost all of load time. What's going on in them?",
"Furthermore, what's my current workaround about this problem? Should I save it by `save_to_disk()` and load dataset through `load_from_disk`?",
"were you able to solve this issue?, I am facing the same issue"
] | ### Describe the bug
I have an image dataset with 300k images, the size of image is 768 * 768.
When I run `dataset = load_dataset("imagefolder", data_dir="/path/to/img_dir", split='train')` in second time, it takes 50 minutes to finish "Resolving data files" part, what's going on in this part?
From my understand, after Arrow files been created in the first run, the second run should not take time longer than one or two minutes.
### Steps to reproduce the bug
# Run following code two times
dataset = load_dataset("imagefolder", data_dir="/path/to/img_dir", split='train')
### Expected behavior
Fast dataset building
### Environment info
- `datasets` version: 2.14.5
- Platform: Linux-5.15.0-60-generic-x86_64-with-glibc2.35
- Python version: 3.10.11
- Huggingface_hub version: 0.17.3
- PyArrow version: 10.0.1
- Pandas version: 1.5.3 | 6,359 |
https://github.com/huggingface/datasets/issues/6358 | Mounting datasets cache fails due to absolute paths. | [
"You may be able to make it work by tweaking some environment variables, such as [`HF_HOME`](https://huggingface.co/docs/huggingface_hub/main/en/package_reference/environment_variables#hfhome) or [`HF_DATASETS_CACHE`](https://huggingface.co/docs/datasets/cache#cache-directory).",
"> You may be able to make it work by tweaking some environment variables, such as [`HF_HOME`](https://huggingface.co/docs/huggingface_hub/main/en/package_reference/environment_variables#hfhome) or [`HF_DATASETS_CACHE`](https://huggingface.co/docs/datasets/cache#cache-directory).\r\n\r\nI am already doing this. The problem is that, while this seemingly allows flexibility, the absolute paths written into the cache still have the old cache directory. The paths written into the cache should be relative to the cache location to allow this sort of flexibility. Sorry, I omitted this in the reproduction steps, I have now added it.",
"I'm unable to reproduce this with the cache\r\n```bash\r\nexport HF_CACHE=$PWD/hf_cache\r\npython -c \"import datasets; datasets.load_dataset('imdb')\"\r\n```\r\nimported inside a dummy container that is built from\r\n```bash\r\nFROM python:3.9\r\n\r\nWORKDIR /usr/src/app\r\n\r\nRUN pip install datasets\r\n\r\nCOPY ./hf_cache ./hf_cache\r\n\r\nENV HF_HOME=./hf_cache\r\nENV HF_DATASETS_OFFLINE=1\r\n\r\nCMD [\"python\"]\r\n```\r\nWhat do you mean by \"absolute paths written into the cache\"? Paths inside the HF cache paths are based on hash (hashed URL of the downloaded files, etc.)",
"@mariosasko Same problem: the absolute paths written into the cache still have the old cache directory. Like:\r\n\r\n{'bytes': None, 'path': 'E:\\\\work-20240321\\\\datasets\\\\downloads\\\\extracted\\\\9752883596854dc57e01c74cc3f494b2ba63754dadd9e77f9d1932deddbd2273\\\\58f33a03-026f-4adc-b69f-b89d16b9f35a.webp'}\r\n\r\nWhen I move this cached directory to another directory, these datasets cannot be used casue path changes. So, the paths written into the cache should be relative to the cache location to allow this sort of flexibility. ",
"Sorry, the reply on this thread escaped my attention. The problem with @mariosasko's attempted reproduction is the absolute path `./hf_cache` is the same in the host system and the docker container, so naturally the paths would be correct. Modifying the docker image as below should reproduce the error...\r\n\r\n```\r\nFROM python:3.9\r\n\r\nWORKDIR /usr/src/app\r\n\r\nRUN pip install datasets\r\n\r\nCOPY ./hf_cache ./my_cache/\r\n\r\nENV HF_HOME=./my_cache/\r\nENV HF_DATASETS_OFFLINE=1\r\n\r\nCMD [\"python\"]\r\n```\r\n\r\nThe paths written inside the cache will still have `./hf_cache` prefixing all the paths. If they were relative paths (relative to the top level of the cache) this would be avoided."
] | ### Describe the bug
Creating a datasets cache and mounting this into, for example, a docker container, renders the data unreadable due to absolute paths written into the cache.
### Steps to reproduce the bug
1. Create a datasets cache by downloading some data
2. Mount the dataset folder into a docker container or remote system.
3. (Edit) Set `HF_HOME` or `HF_DATASET_CACHE` to point to the mounted cache.
4. Attempt to access the data from within the docker container.
5. An error is thrown saying no file exists at \<absolute path to original cache location\>
### Expected behavior
The data is loaded without error
### Environment info
- `datasets` version: 2.14.4
- Platform: Linux-5.4.0-162-generic-x86_64-with-glibc2.29
- Python version: 3.8.10
- Huggingface_hub version: 0.16.4
- PyArrow version: 13.0.0
- Pandas version: 2.0.3 | 6,358 |
https://github.com/huggingface/datasets/issues/6357 | Allow passing a multiprocessing context to functions that support `num_proc` | [] | ### Feature request
Allow specifying [a multiprocessing context](https://docs.python.org/3/library/multiprocessing.html#contexts-and-start-methods) to functions that support `num_proc` or use multiprocessing pools. For example, the following could be done:
```python
dataset = dataset.map(_func, num_proc=2, mp_context=multiprocess.get_context("spawn"))
```
Or at least the multiprocessing start method ("fork", "spawn", "fork_server" or `None`):
```python
dataset = dataset.map(_func, num_proc=2, mp_start_method="spawn")
```
Another option could be passing the `pool` as an argument.
### Motivation
By default, `multiprocess` (the `multiprocessing`-fork library that this repo uses) uses the "fork" start method for multiprocessing pools (for the default context). It could be changed by using `set_start_method`. However, this conditions the multiprocessing start method from all processing in a Python program that uses the default context, because [you can't call that function more than once](https://docs.python.org/3/library/multiprocessing.html#contexts-and-start-methods:~:text=set_start_method()%20should%20not%20be%20used%20more%20than%20once%20in%20the%20program.). My proposal is to allow using a different multiprocessing context, not to condition the whole Python program.
One reason to change the start method is that "fork" (the default) makes child processes likely deadlock if thread pools were created before (and also this is not supported by POSIX). For example, this happens when using PyTorch because OpenMP threads are used for CPU intra-op parallelism, which is enabled by default (e.g., for context see [`torch.set_num_threads`](https://pytorch.org/docs/stable/generated/torch.set_num_threads.html)). This can also be fixed by setting `torch.set_num_threads(1)` (or similarly by other methods) but this conditions the whole Python program as it can only be set once to guarantee its behavior (similarly to). There are noticeable performance differences when setting this number to 1 even when using GPU(s). Using, e.g., a "spawn" start method would solve this issue.
For more context, see:
* https://discuss.huggingface.co/t/dataset-map-stuck-with-torch-set-num-threads-set-to-2-or-larger/37984
* https://discuss.huggingface.co/t/using-num-proc-1-in-dataset-map-hangs/44310
### Your contribution
I'd be happy to review a PR that makes such a change. And if you really don't have the bandwidth for it, I'd consider creating one. | 6,357 |
https://github.com/huggingface/datasets/issues/6354 | `IterableDataset.from_spark` does not support multiple workers in pytorch `Dataloader` | [
"I am having issues as well with this. \r\n\r\nHowever, the error I am getting is :\r\n`RuntimeError: It appears that you are attempting to reference SparkContext from a broadcast variable, action, or transformation. SparkContext can only be used on the driver, not in code that it run on workers. For more information, see SPARK-5063.`\r\n\r\nAlso did not work with pyspark==3.3.0 and py4j==0.10.9.5"
] | ### Describe the bug
Looks like `IterableDataset.from_spark` does not support multiple workers in pytorch `Dataloader` if I'm not missing anything.
Also, returns not consistent error messages, which probably depend on the nondeterministic order of worker executions
Some exampes I've encountered:
```
File "/local_disk0/.ephemeral_nfs/envs/pythonEnv-68c05436-3512-41c4-88ca-5630012b70d1/lib/python3.10/site-packages/datasets/packaged_modules/spark/spark.py", line 79, in __iter__
yield from self.generate_examples_fn()
File "/local_disk0/.ephemeral_nfs/envs/pythonEnv-68c05436-3512-41c4-88ca-5630012b70d1/lib/python3.10/site-packages/datasets/packaged_modules/spark/spark.py", line 49, in generate_fn
df_with_partition_id = df.select("*", pyspark.sql.functions.spark_partition_id().alias("part_id"))
File "/databricks/spark/python/pyspark/instrumentation_utils.py", line 54, in wrapper
logger.log_failure(
File "/databricks/spark/python/pyspark/databricks/usage_logger.py", line 70, in log_failure
self.logger.recordFunctionCallFailureEvent(
File "/databricks/spark/python/lib/py4j-0.10.9.7-src.zip/py4j/java_gateway.py", line 1322, in __call__
return_value = get_return_value(
File "/databricks/spark/python/pyspark/errors/exceptions/captured.py", line 188, in deco
return f(*a, **kw)
File "/databricks/spark/python/lib/py4j-0.10.9.7-src.zip/py4j/protocol.py", line 342, in get_return_value
return OUTPUT_CONVERTER[type](answer[2:], gateway_client)
KeyError: 'c'
```
```
File "/local_disk0/.ephemeral_nfs/envs/pythonEnv-68c05436-3512-41c4-88ca-5630012b70d1/lib/python3.10/site-packages/datasets/packaged_modules/spark/spark.py", line 79, in __iter__
yield from self.generate_examples_fn()
File "/local_disk0/.ephemeral_nfs/envs/pythonEnv-68c05436-3512-41c4-88ca-5630012b70d1/lib/python3.10/site-packages/datasets/packaged_modules/spark/spark.py", line 49, in generate_fn
df_with_partition_id = df.select("*", pyspark.sql.functions.spark_partition_id().alias("part_id"))
File "/databricks/spark/python/pyspark/sql/utils.py", line 162, in wrapped
return f(*args, **kwargs)
File "/databricks/spark/python/pyspark/sql/functions.py", line 4893, in spark_partition_id
return _invoke_function("spark_partition_id")
File "/databricks/spark/python/pyspark/sql/functions.py", line 98, in _invoke_function
return Column(jf(*args))
File "/databricks/spark/python/lib/py4j-0.10.9.7-src.zip/py4j/java_gateway.py", line 1322, in __call__
return_value = get_return_value(
File "/databricks/spark/python/pyspark/errors/exceptions/captured.py", line 188, in deco
return f(*a, **kw)
File "/databricks/spark/python/lib/py4j-0.10.9.7-src.zip/py4j/protocol.py", line 342, in get_return_value
return OUTPUT_CONVERTER[type](answer[2:], gateway_client)
KeyError: 'm'
```
```
File "/local_disk0/.ephemeral_nfs/envs/pythonEnv-68c05436-3512-41c4-88ca-5630012b70d1/lib/python3.10/site-packages/datasets/packaged_modules/spark/spark.py", line 79, in __iter__
yield from self.generate_examples_fn()
File "/local_disk0/.ephemeral_nfs/envs/pythonEnv-68c05436-3512-41c4-88ca-5630012b70d1/lib/python3.10/site-packages/datasets/packaged_modules/spark/spark.py", line 49, in generate_fn
df_with_partition_id = df.select("*", pyspark.sql.functions.spark_partition_id().alias("part_id"))
File "/databricks/spark/python/pyspark/sql/utils.py", line 162, in wrapped
return f(*args, **kwargs)
File "/databricks/spark/python/pyspark/sql/functions.py", line 4893, in spark_partition_id
return _invoke_function("spark_partition_id")
File "/databricks/spark/python/pyspark/sql/functions.py", line 97, in _invoke_function
jf = _get_jvm_function(name, SparkContext._active_spark_context)
File "/databricks/spark/python/pyspark/sql/functions.py", line 88, in _get_jvm_function
return getattr(sc._jvm.functions, name)
File "/databricks/spark/python/lib/py4j-0.10.9.7-src.zip/py4j/java_gateway.py", line 1725, in __getattr__
raise Py4JError(message)
py4j.protocol.Py4JError: functions does not exist in the JVM
```
### Steps to reproduce the bug
```python
import pandas as pd
import numpy as np
batch_size = 16
pdf = pd.DataFrame({
key: np.random.rand(16*100) for key in ['feature', 'target']
})
test_df = spark.createDataFrame(pdf)
from datasets import IterableDataset
from torch.utils.data import DataLoader
ids = IterableDataset.from_spark(test_df)
for batch in DataLoader(ids, batch_size=16, num_workers=4):
for k, b in batch.items():
print(k, b.shape, sep='\t')
print('\n')
```
### Expected behavior
For `num_workers` equal to 0 or 1 works fine as expected:
```
feature torch.Size([16])
target torch.Size([16])
feature torch.Size([16])
target torch.Size([16])
....
```
Expected to support workers >1.
### Environment info
Databricks 13.3 LTS ML runtime - Spark 3.4.1
pyspark==3.4.1
py4j==0.10.9.7
datasets==2.13.1 and also tested with datasets==2.14.6 | 6,354 |
https://github.com/huggingface/datasets/issues/6353 | load_dataset save_to_disk load_from_disk error | [
"solved.\r\nfsspec version problem",
"I'm using the latest datasets and fsspec , but still got this error!\r\n\r\ndatasets : Version: 2.13.0\r\n\r\nfsspec Version: 2023.10.0\r\n\r\n```\r\nFile \"/home/guoby/app/Anaconda3-2021.05/envs/news/lib/python3.8/site-packages/datasets/load.py\", line 1892, in load_from_disk\r\n return DatasetDict.load_from_disk(dataset_path, keep_in_memory=keep_in_memory, storage_options=storage_options)\r\n File \"/home/guoby/app/Anaconda3-2021.05/envs/news/lib/python3.8/site-packages/datasets/dataset_dict.py\", line 1371, in load_from_disk\r\n dataset_dict[k] = Dataset.load_from_disk(\r\n File \"/home/guoby/app/Anaconda3-2021.05/envs/news/lib/python3.8/site-packages/datasets/arrow_dataset.py\", line 1639, in load_from_disk\r\n fs_token_paths = fsspec.get_fs_token_paths(dataset_path, storage_options=storage_options)\r\n File \"/home/guoby/app/Anaconda3-2021.05/envs/news/lib/python3.8/site-packages/fsspec/core.py\", line 610, in get_fs_token_paths\r\n chain = _un_chain(urlpath0, storage_options or {})\r\n File \"/home/guoby/app/Anaconda3-2021.05/envs/news/lib/python3.8/site-packages/fsspec/core.py\", line 325, in _un_chain\r\n cls = get_filesystem_class(protocol)\r\n File \"/home/guoby/app/Anaconda3-2021.05/envs/news/lib/python3.8/site-packages/fsspec/registry.py\", line 232, in get_filesystem_class\r\n raise ValueError(f\"Protocol not known: {protocol}\")\r\n```",
"These two versions work.\r\n<img width=\"807\" alt=\"截圖 2023-11-22 下午5 55 28\" src=\"https://github.com/huggingface/datasets/assets/77866896/faa8333f-0519-4d69-b243-a8880cd7fc1f\">\r\n",
"datasets==2.10.1 and fsspec==2023.6.0 also works for me.",
"确实"
] | ### Describe the bug
datasets version： 2.10.1
I `load_dataset `and `save_to_disk` sucessfully on windows10( **and I `load_from_disk(/LLM/data/wiki)` succcesfully on windows10**), and I copy the dataset `/LLM/data/wiki`
into a ubuntu system, but when I `load_from_disk(/LLM/data/wiki)` on ubuntu, something weird happens:
```
load_from_disk('/LLM/data/wiki')
File "/usr/local/miniconda3/lib/python3.8/site-packages/datasets/load.py", line 1874, in load_from_disk
return DatasetDict.load_from_disk(dataset_path, keep_in_memory=keep_in_memory, storage_options=storage_options)
File "/usr/local/miniconda3/lib/python3.8/site-packages/datasets/dataset_dict.py", line 1309, in load_from_disk
dataset_dict[k] = Dataset.load_from_disk(
File "/usr/local/miniconda3/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 1543, in load_from_disk
fs_token_paths = fsspec.get_fs_token_paths(dataset_path, storage_options=storage_options)
File "/usr/local/miniconda3/lib/python3.8/site-packages/fsspec/core.py", line 610, in get_fs_token_paths
chain = _un_chain(urlpath0, storage_options or {})
File "/usr/local/miniconda3/lib/python3.8/site-packages/fsspec/core.py", line 325, in _un_chain
cls = get_filesystem_class(protocol)
File "/usr/local/miniconda3/lib/python3.8/site-packages/fsspec/registry.py", line 232, in get_filesystem_class
raise ValueError(f"Protocol not known: {protocol}")
ValueError: Protocol not known: /LLM/data/wiki
```
It seems that something went wrong on the arrow file?
How can I solve this , since currently I can not save_to_disk on ubuntu system
### Steps to reproduce the bug
datasets version： 2.10.1
### Expected behavior
datasets version： 2.10.1
### Environment info
datasets version： 2.10.1 | 6,353 |
https://github.com/huggingface/datasets/issues/6352 | Error loading wikitext data raise NotImplementedError(f"Loading a dataset cached in a {type(self._fs).__name__} is not supported.") | [
"+1 \r\n```\r\nFound cached dataset csv (file:///home/ubuntu/.cache/huggingface/datasets/theSquarePond___csv/theSquarePond--XXXXX-bbf0a8365d693d2c/0.0.0/eea64c71ca8b46dd3f537ed218fc9bf495d5707789152eb2764f5c78fa66d59d)\r\n---------------------------------------------------------------------------\r\nNotImplementedError Traceback (most recent call last)\r\nCell In[14], line 4\r\n 1 get_ipython().system('pip install -U datasets')\r\n 3 # Load dataset from the hub\r\n----> 4 dataset = load_dataset(dataset_name)\r\n\r\nFile ~/anaconda3/envs/python38-env/lib/python3.8/site-packages/datasets/load.py:1810, in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, verification_mode, ignore_verifications, keep_in_memory, save_infos, revision, use_auth_token, task, streaming, num_proc, storage_options, **config_kwargs)\r\n 1806 # Build dataset for splits\r\n 1807 keep_in_memory = (\r\n 1808 keep_in_memory if keep_in_memory is not None else is_small_dataset(builder_instance.info.dataset_size)\r\n 1809 )\r\n-> 1810 ds = builder_instance.as_dataset(split=split, verification_mode=verification_mode, in_memory=keep_in_memory)\r\n 1811 # Rename and cast features to match task schema\r\n 1812 if task is not None:\r\n\r\nFile ~/anaconda3/envs/python38-env/lib/python3.8/site-packages/datasets/builder.py:1128, in DatasetBuilder.as_dataset(self, split, run_post_process, verification_mode, ignore_verifications, in_memory)\r\n 1126 is_local = not is_remote_filesystem(self._fs)\r\n 1127 if not is_local:\r\n-> 1128 raise NotImplementedError(f\"Loading a dataset cached in a {type(self._fs).__name__} is not supported.\")\r\n 1129 if not os.path.exists(self._output_dir):\r\n 1130 raise FileNotFoundError(\r\n 1131 f\"Dataset {self.name}: could not find data in {self._output_dir}. Please make sure to call \"\r\n 1132 \"builder.download_and_prepare(), or use \"\r\n 1133 \"datasets.load_dataset() before trying to access the Dataset object.\"\r\n 1134 )\r\n\r\nNotImplementedError: Loading a dataset cached in a LocalFileSystem is not supported.\r\n```",
"+1\r\n\r\n```\r\nFound cached dataset csv ([file://C:/Users/Shady/.cache/huggingface/datasets/knkarthick___csv/knkarthick--dialogsum-cd36827d3490488d/0.0.0/6954658bab30a358235fa864b05cf819af0e179325c740e4bc853bcc7ec513e1](file:///C:/Users/Shady/.cache/huggingface/datasets/knkarthick___csv/knkarthick--dialogsum-cd36827d3490488d/0.0.0/6954658bab30a358235fa864b05cf819af0e179325c740e4bc853bcc7ec513e1))\r\n---------------------------------------------------------------------------\r\nNotImplementedError Traceback (most recent call last)\r\nCell In[38], line 3\r\n 1 huggingface_dataset_name = \"knkarthick/dialogsum\"\r\n----> 3 dataset = load_dataset(huggingface_dataset_name)\r\n\r\nFile D:\\Desktop\\Workspace\\GenAI\\genai\\lib\\site-packages\\datasets\\load.py:1804, in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, verification_mode, ignore_verifications, keep_in_memory, save_infos, revision, use_auth_token, task, streaming, num_proc, storage_options, **config_kwargs)\r\n 1800 # Build dataset for splits\r\n 1801 keep_in_memory = (\r\n 1802 keep_in_memory if keep_in_memory is not None else is_small_dataset(builder_instance.info.dataset_size)\r\n 1803 )\r\n-> 1804 ds = builder_instance.as_dataset(split=split, verification_mode=verification_mode, in_memory=keep_in_memory)\r\n 1805 # Rename and cast features to match task schema\r\n 1806 if task is not None:\r\n\r\nFile D:\\Desktop\\Workspace\\GenAI\\genai\\lib\\site-packages\\datasets\\builder.py:1108, in DatasetBuilder.as_dataset(self, split, run_post_process, verification_mode, ignore_verifications, in_memory)\r\n 1106 is_local = not is_remote_filesystem(self._fs)\r\n 1107 if not is_local:\r\n-> 1108 raise NotImplementedError(f\"Loading a dataset cached in a {type(self._fs).__name__} is not supported.\")\r\n 1109 if not os.path.exists(self._output_dir):\r\n 1110 raise FileNotFoundError(\r\n 1111 f\"Dataset {self.name}: could not find data in {self._output_dir}. Please make sure to call \"\r\n 1112 \"builder.download_and_prepare(), or use \"\r\n 1113 \"datasets.load_dataset() before trying to access the Dataset object.\"\r\n 1114 )\r\n\r\nNotImplementedError: Loading a dataset cached in a LocalFileSystem is not supported.\r\n```",
"This error stems from a breaking change in `fsspec`. It has been fixed in the latest `datasets` release (`2.14.6`). Updating the installation with `pip install -U datasets` should fix the issue.\r\n",
"> 此错误源于 中的重大更改。此问题已在最新版本 （） 中修复。更新安装应该可以解决此问题。`fsspec``datasets``2.14.6``pip install -U datasets`\r\n\r\nthanks , 太好啦，刚好解决了我的问题，GPT都没解决了，终于被你搞定了",
"https://stackoverflow.com/questions/77433096/notimplementederror-loading-a-dataset-cached-in-a-localfilesystem-is-not-suppor/77433141#77433141",
"Fixed by:\r\n- https://github.com/huggingface/datasets/pull/6334\r\n\r\nThe fix was released in `datasets-2.14.6`.",
"this is fixed in 2.15.0, but broken again in 2.17.0. Can someone verify?",
"I'm on `2.17.1` and can confirm it's broken again. Downgrading to `2.16` helped.",
"> 2.14.6\r\n\r\ni update the version but the error still exist \r\n",
"The issue seems to persist in 2.18.0",
"same problem in 2.18.0",
"Which version of `fsspec` and OS are you using ?",
"> Which version of `fsspec` and OS are you using ?\r\n\r\n`fsspec-2023.10.0` and Windows 10, guess fsspec version too old..."
] | I was trying to load the wiki dataset, but i got this error
traindata = load_dataset('wikitext', 'wikitext-2-raw-v1', split='train')
File "/home/aelkordy/.conda/envs/prune_llm/lib/python3.9/site-packages/datasets/load.py", line 1804, in load_dataset
ds = builder_instance.as_dataset(split=split, verification_mode=verification_mode, in_memory=keep_in_memory)
File "/home/aelkordy/.conda/envs/prune_llm/lib/python3.9/site-packages/datasets/builder.py", line 1108, in as_dataset
raise NotImplementedError(f"Loading a dataset cached in a {type(self._fs).__name__} is not supported.")
NotImplementedError: Loading a dataset cached in a LocalFileSystem is not supported. | 6,352 |
https://github.com/huggingface/datasets/issues/6350 | Different objects are returned from calls that should be returning the same kind of object. | [
"`load_dataset` returns a `DatasetDict` object unless `split` is defined, in which case it returns a `Dataset` (or a list of datasets if `split` is a list). We've discussed dropping `DatasetDict` from the API in https://github.com/huggingface/datasets/issues/5189 to always return the same type in `load_dataset` and support datasets without (explicit) splits. IIRC the main discussion point is deciding what to return when loading a dataset with multiple splits, but `split` is not specified. What would you expect as a return value in that scenario?",
"> `load_dataset` returns a `DatasetDict` object unless `split` is defined, in which case it returns a `Dataset` (or a list of datasets if `split` is a list). We've discussed dropping `DatasetDict` from the API in #5189 to always return the same type in `load_dataset` and support datasets without (explicit) splits. IIRC the main discussion point is deciding what to return when loading a dataset with multiple splits, but `split` is not specified. What would you expect as a return value in that scenario?\r\n\r\nWouldn't a dataset with multiple splits already have keys and their related data arrays?\r\n\r\nLets say the dataset has \"train\" : trainset, \"valid\": validset and \"test\": testset\r\n\r\nSo a dictionary can be returned,, i.e.\r\n\r\n{ \r\n\"train\": trainset,\r\n\"valid\": validset,\r\n\"test\": testset\r\n}\r\n\r\nif a split is provided split=['train[:80%]', 'valid[80%:90%]', 'test[90%:100%]']\r\n\r\nwould also return the same dictionary as above.\r\n\r\nsplit='train[:10%]' should return the same value as split=['train[:10%]']\r\n\r\n{\r\n\"train\": trainset\r\n}\r\n "
] | ### Describe the bug
1. dataset = load_dataset("togethercomputer/RedPajama-Data-1T-Sample", cache_dir=training_args.cache_dir, split='train[:1%]')
2. dataset = load_dataset("togethercomputer/RedPajama-Data-1T-Sample", cache_dir=training_args.cache_dir)
The only difference I would expect these calls to have is the size of the dataset.
But, while 2. returns a dictionary with "train" key in it, 1. returns a dataset WITHOUT any initial "train" keyword.
Both calls are to be used within exactly the same context. They should return identically structured datasets of different size.
### Steps to reproduce the bug
See above.
### Expected behavior
Expect both calls to return the same structured Dataset structure but with different number of elements, i.e. call 1. should have 1% of the data of the call 2.0
### Environment info
Ubuntu 20.04
gcc 9.x.x.
It is really irrelevant. | 6,350 |
https://github.com/huggingface/datasets/issues/6349 | Can't load ds = load_dataset("imdb") | [
"I'm unable to reproduce this error. The server hosting the files may have been down temporarily, so try again.",
"getting the same error",
"I am getting the following error:\r\nEnv: Python3.10\r\ndatasets: 2.10.1\r\nLinux: Amazon Linux2\r\n\r\n`Traceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/home/ec2-user/anaconda3/envs/JupyterSystemEnv/lib/python3.10/site-packages/datasets/load.py\", line 1759, in load_dataset\r\n builder_instance = load_dataset_builder(\r\n File \"/home/ec2-user/anaconda3/envs/JupyterSystemEnv/lib/python3.10/site-packages/datasets/load.py\", line 1496, in load_dataset_builder\r\n dataset_module = dataset_module_factory(\r\n File \"/home/ec2-user/anaconda3/envs/JupyterSystemEnv/lib/python3.10/site-packages/datasets/load.py\", line 1218, in dataset_module_factory\r\n raise e1 from None\r\n File \"/home/ec2-user/anaconda3/envs/JupyterSystemEnv/lib/python3.10/site-packages/datasets/load.py\", line 1202, in dataset_module_factory\r\n ).get_module()\r\n File \"/home/ec2-user/anaconda3/envs/JupyterSystemEnv/lib/python3.10/site-packages/datasets/load.py\", line 767, in get_module\r\n else get_data_patterns_in_dataset_repository(hfh_dataset_info, self.data_dir)\r\n File \"/home/ec2-user/anaconda3/envs/JupyterSystemEnv/lib/python3.10/site-packages/datasets/data_files.py\", line 675, in get_data_patterns_in_dataset_repository\r\n return _get_data_files_patterns(resolver)\r\n File \"/home/ec2-user/anaconda3/envs/JupyterSystemEnv/lib/python3.10/site-packages/datasets/data_files.py\", line 236, in _get_data_files_patterns\r\n data_files = pattern_resolver(pattern)\r\n File \"/home/ec2-user/anaconda3/envs/JupyterSystemEnv/lib/python3.10/site-packages/datasets/data_files.py\", line 486, in _resolve_single_pattern_in_dataset_repository\r\n glob_iter = [PurePath(filepath) for filepath in fs.glob(PurePath(pattern).as_posix()) if fs.isfile(filepath)]\r\n File \"/home/ec2-user/anaconda3/envs/JupyterSystemEnv/lib/python3.10/site-packages/fsspec/spec.py\", line 606, in glob\r\n pattern = glob_translate(path + (\"/\" if ends_with_sep else \"\"))\r\n File \"/home/ec2-user/anaconda3/envs/JupyterSystemEnv/lib/python3.10/site-packages/fsspec/utils.py\", line 734, in glob_translate\r\n raise ValueError(\r\nValueError: Invalid pattern: '**' can only be an entire path component`",
"Resolved by upgrading datasets version to 2.18.0"
] | ### Describe the bug
I did `from datasets import load_dataset, load_metric` and then `ds = load_dataset("imdb")` and it gave me the error:
ExpectedMoreDownloadedFiles: {'http://ai.stanford.edu/~amaas/data/sentiment/aclImdb_v1.tar.gz'}
I tried doing `ds = load_dataset("imdb",download_mode="force_redownload")` as well as reinstalling dataset. I still face this problem.
### Steps to reproduce the bug
1. from datasets import load_dataset, load_metric
2. ds = load_dataset("imdb")
### Expected behavior
It should load and give me this when I run `ds`
DatasetDict({
train: Dataset({
features: ['text', 'label'],
num_rows: 25000
})
test: Dataset({
features: ['text', 'label'],
num_rows: 25000
})
unsupervised: Dataset({
features: ['text', 'label'],
num_rows: 50000
})
})
### Environment info
- `datasets` version: 2.14.6
- Platform: Linux-5.4.0-164-generic-x86_64-with-glibc2.17
- Python version: 3.8.18
- Huggingface_hub version: 0.16.2
- PyArrow version: 13.0.0
- Pandas version: 2.0.2 | 6,349 |
https://github.com/huggingface/datasets/issues/6348 | Parquet stream-conversion fails to embed images/audio files from gated repos | [] | it seems to be an issue with datasets not passing the token to embed_table_storage when generating a dataset
See https://github.com/huggingface/datasets-server/issues/2010 | 6,348 |
https://github.com/huggingface/datasets/issues/6347 | Incorrect example code in 'Create a dataset' docs | [
"This was fixed in https://github.com/huggingface/datasets/pull/6247. You can find the fix in the `main` version of the docs",
"Ah great, thanks :)"
] | ### Describe the bug
On [this](https://huggingface.co/docs/datasets/create_dataset) page, the example code for loading in images and audio is incorrect.
Currently, examples are:
``` python
from datasets import ImageFolder
dataset = load_dataset("imagefolder", data_dir="/path/to/pokemon")
```
and
``` python
from datasets import AudioFolder
dataset = load_dataset("audiofolder", data_dir="/path/to/folder")
```
I'm pretty sure the imports are wrong and should be:
``` python
from datasets import load_dataset
dataset = load_dataset("audiofolder", data_dir="/path/to/folder")
```
I am happy to update this if this is right but just wanted to check before making any changes.
### Steps to reproduce the bug
Go to https://huggingface.co/docs/datasets/create_dataset
### Expected behavior
N/A
### Environment info
N/A | 6,347 |
https://github.com/huggingface/datasets/issues/6345 | support squad structure datasets using a YAML parameter | [] | ### Feature request
Since the squad structure is widely used, I think it could be beneficial to support it using a YAML parameter.
could you implement automatic data loading of squad-like data using squad JSON format, to read it from JSON files and view it in the correct squad structure.
The dataset structure should be like this:
https://huggingface.co/datasets/squad
Columns:id,title,context,question,answers
### Motivation
Dataset repo requires arbitrary Python code execution
### Your contribution
The dataset structure should be like this:
https://huggingface.co/datasets/squad
Columns:id,title,context,question,answers
train and dev sets in squad structure JSON files | 6,345 |
https://github.com/huggingface/datasets/issues/6333 | Support fsspec 2023.10.0 | [
"Hi @albertvillanova @lhoestq \r\n\r\nI believe the pull request that pins the fsspec version (https://github.com/huggingface/datasets/pull/6331) was merged by mistake. Another fix for the issue was merged on the same day an hour apart. See https://github.com/huggingface/datasets/pull/6334\r\n\r\nI'm now having an issue in my project where I can't use newer versions of fsspec.\r\n\r\nCan we remove the pin?\r\n\r\nHave a nice day! :)",
"Hi @tomscholz,\r\n\r\nThanks for pointing this out. I think you are right.\r\n\r\nI am doing some cross-checks and fixing it. ",
"Hi again, @tomscholz.\r\n\r\nAfter a more cautious investigation, I think the pin is OK because there are other reasons for it. Chronologically:\r\n- #6331 \r\n- #6334\r\n- #6336 \r\n- #6337 \r\n\r\nThe reason is that after version 2023.10.0, they changed again the behavior of their `glob` function. See: https://github.com/huggingface/datasets/pull/6337#issuecomment-1774930135\r\nWe are working on our side to support both previous and new glob behavior.\r\n\r\nNote:\r\n- First pin was < 2023.10.0\r\n- Last pin is <= 2023.10.0",
"Fixed by #6334 and #6336."
] | Once root issue is fixed, remove temporary pin of fsspec < 2023.10.0 introduced by:
- #6331
Related to issue:
- #6330
As @ZachNagengast suggested, the issue might be related to:
- https://github.com/fsspec/filesystem_spec/pull/1381 | 6,333 |
https://github.com/huggingface/datasets/issues/6330 | Latest fsspec==2023.10.0 issue with streaming datasets | [
"I also encountered a similar error below.\r\nAppreciate the team could shed some light on this issue.\r\n\r\n```\r\n---------------------------------------------------------------------------\r\nNotImplementedError Traceback (most recent call last)\r\n[/home/ubuntu/work/EveryDream2trainer/prepare_dataset.ipynb](https://vscode-remote+ssh-002dremote-002braspberry-002dg5-002e4x.vscode-resource.vscode-cdn.net/home/ubuntu/work/EveryDream2trainer/prepare_dataset.ipynb) Cell 1 line 4\r\n [1](vscode-notebook-cell://ssh-remote%2Braspberry-g5.4x/home/ubuntu/work/EveryDream2trainer/prepare_dataset.ipynb#W0sdnNjb2RlLXJlbW90ZQ%3D%3D?line=0) from datasets import load_dataset, load_dataset\r\n [3](vscode-notebook-cell://ssh-remote%2Braspberry-g5.4x/home/ubuntu/work/EveryDream2trainer/prepare_dataset.ipynb#W0sdnNjb2RlLXJlbW90ZQ%3D%3D?line=2) # ds = load_dataset(\"parquet\", data_dir=\"/home/ubuntu/work/EveryDream2trainer/datasets/monse_v1/data\")\r\n----> [4](vscode-notebook-cell://ssh-remote%2Braspberry-g5.4x/home/ubuntu/work/EveryDream2trainer/prepare_dataset.ipynb#W0sdnNjb2RlLXJlbW90ZQ%3D%3D?line=3) ds = load_dataset(\"Raspberry-ai/monse-v1\")\r\n\r\nFile [/opt/conda/envs/everydream/lib/python3.10/site-packages/datasets/load.py:1804](https://vscode-remote+ssh-002dremote-002braspberry-002dg5-002e4x.vscode-resource.vscode-cdn.net/opt/conda/envs/everydream/lib/python3.10/site-packages/datasets/load.py:1804), in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, verification_mode, ignore_verifications, keep_in_memory, save_infos, revision, use_auth_token, task, streaming, num_proc, storage_options, **config_kwargs)\r\n 1800 # Build dataset for splits\r\n 1801 keep_in_memory = (\r\n 1802 keep_in_memory if keep_in_memory is not None else is_small_dataset(builder_instance.info.dataset_size)\r\n 1803 )\r\n-> 1804 ds = builder_instance.as_dataset(split=split, verification_mode=verification_mode, in_memory=keep_in_memory)\r\n 1805 # Rename and cast features to match task schema\r\n 1806 if task is not None:\r\n\r\nFile [/opt/conda/envs/everydream/lib/python3.10/site-packages/datasets/builder.py:1108](https://vscode-remote+ssh-002dremote-002braspberry-002dg5-002e4x.vscode-resource.vscode-cdn.net/opt/conda/envs/everydream/lib/python3.10/site-packages/datasets/builder.py:1108), in DatasetBuilder.as_dataset(self, split, run_post_process, verification_mode, ignore_verifications, in_memory)\r\n 1106 is_local = not is_remote_filesystem(self._fs)\r\n 1107 if not is_local:\r\n-> 1108 raise NotImplementedError(f\"Loading a dataset cached in a {type(self._fs).__name__} is not supported.\")\r\n 1109 if not os.path.exists(self._output_dir):\r\n 1110 raise FileNotFoundError(\r\n 1111 f\"Dataset {self.name}: could not find data in {self._output_dir}. Please make sure to call \"\r\n 1112 \"builder.download_and_prepare(), or use \"\r\n 1113 \"datasets.load_dataset() before trying to access the Dataset object.\"\r\n 1114 )\r\n\r\nNotImplementedError: Loading a dataset cached in a LocalFileSystem is not supported.\r\n```\r\n\r\nCode to reproduce the issue:\r\n\r\n```\r\nfrom datasets import load_dataset\r\n\r\nds = load_dataset(\"Raspberry-ai/monse-v1\")\r\n```\r\n\r\n\r\nDependencies:\r\n```\r\nPackage Version\r\n------------------------- ------------\r\nabsl-py 2.0.0\r\naccelerate 0.23.0\r\naiohttp 3.8.4\r\naiosignal 1.3.1\r\nantlr4-python3-runtime 4.9.3\r\nanyio 4.0.0\r\nappdirs 1.4.4\r\nargon2-cffi 23.1.0\r\nargon2-cffi-bindings 21.2.0\r\narrow 1.3.0\r\nasttokens 2.4.0\r\nasync-lru 2.0.4\r\nasync-timeout 4.0.3\r\nattrs 23.1.0\r\nBabel 2.13.0\r\nbackcall 0.2.0\r\nbeautifulsoup4 4.12.2\r\nbitsandbytes 0.41.1\r\nbleach 6.1.0\r\nbraceexpand 0.1.7\r\ncachetools 5.3.1\r\ncertifi 2023.7.22\r\ncffi 1.16.0\r\ncharset-normalizer 3.3.1\r\nclick 8.1.7\r\ncmake 3.27.7\r\ncolorama 0.4.6\r\ncomm 0.1.4\r\ncompel 1.1.6\r\ndatasets 2.11.0\r\ndebugpy 1.8.0\r\ndecorator 5.1.1\r\ndefusedxml 0.7.1\r\ndiffusers 0.18.0\r\ndill 0.3.6\r\ndocker-pycreds 0.4.0\r\ndowg 0.3.1\r\neinops 0.7.0\r\neinops-exts 0.0.4\r\nexceptiongroup 1.1.3\r\nexecuting 2.0.0\r\nfastjsonschema 2.18.1\r\nfilelock 3.12.4\r\nfqdn 1.5.1\r\nfrozenlist 1.4.0\r\nfsspec 2023.10.0\r\nftfy 6.1.1\r\ngitdb 4.0.11\r\nGitPython 3.1.40\r\ngoogle-auth 2.23.3\r\ngoogle-auth-oauthlib 1.1.0\r\ngrpcio 1.59.0\r\nhuggingface-hub 0.18.0\r\nidna 3.4\r\nimportlib-metadata 6.8.0\r\ninflection 0.5.1\r\nipykernel 6.25.2\r\nipython 8.16.1\r\nisoduration 20.11.0\r\njedi 0.19.1\r\nJinja2 3.1.2\r\njoblib 1.3.2\r\njson5 0.9.14\r\njsonpointer 2.4\r\njsonschema 4.19.1\r\njsonschema-specifications 2023.7.1\r\njupyter_client 8.4.0\r\njupyter_core 5.4.0\r\njupyter-events 0.8.0\r\njupyter-lsp 2.2.0\r\njupyter_server 2.8.0\r\njupyter_server_terminals 0.4.4\r\njupyterlab 4.0.7\r\njupyterlab-pygments 0.2.2\r\njupyterlab_server 2.25.0\r\nlightning-utilities 0.9.0\r\nlion-pytorch 0.1.2\r\nlit 17.0.3\r\nMarkdown 3.5\r\nMarkupSafe 2.1.3\r\nmatplotlib-inline 0.1.6\r\nmistune 3.0.2\r\nmore-itertools 10.1.0\r\nmpmath 1.3.0\r\nmultidict 6.0.4\r\nmultiprocess 0.70.14\r\nmypy-extensions 1.0.0\r\nnbclient 0.8.0\r\nnbconvert 7.9.2\r\nnbformat 5.9.2\r\nnest-asyncio 1.5.8\r\nnetworkx 3.2\r\nnltk 3.8.1\r\nnotebook_shim 0.2.3\r\nnumpy 1.23.5\r\noauthlib 3.2.2\r\nomegaconf 2.2.3\r\nopen-clip-torch 2.22.0\r\nopen-flamingo 2.0.0\r\noverrides 7.4.0\r\npackaging 23.2\r\npandas 2.1.1\r\npandocfilters 1.5.0\r\nparso 0.8.3\r\npathtools 0.1.2\r\npexpect 4.8.0\r\npickleshare 0.7.5\r\nPillow 10.1.0\r\npip 23.3.1\r\nplatformdirs 3.11.0\r\nprometheus-client 0.17.1\r\nprompt-toolkit 3.0.39\r\nprotobuf 3.20.1\r\npsutil 5.9.6\r\nptyprocess 0.7.0\r\npure-eval 0.2.2\r\npyarrow 13.0.0\r\npyasn1 0.5.0\r\npyasn1-modules 0.3.0\r\npycparser 2.21\r\npyDeprecate 0.3.2\r\nPygments 2.16.1\r\npynvml 11.4.1\r\npyparsing 3.1.1\r\npyre-extensions 0.0.29\r\npython-dateutil 2.8.2\r\npython-json-logger 2.0.7\r\npytorch-lightning 1.6.5\r\npytz 2023.3.post1\r\nPyYAML 6.0.1\r\npyzmq 25.1.1\r\nreferencing 0.30.2\r\nregex 2023.10.3\r\nrequests 2.31.0\r\nrequests-oauthlib 1.3.1\r\nresponses 0.18.0\r\nrfc3339-validator 0.1.4\r\nrfc3986-validator 0.1.1\r\nrpds-py 0.10.6\r\nrsa 4.9\r\nsafetensors 0.4.0\r\nscipy 1.11.3\r\nSend2Trash 1.8.2\r\nsentencepiece 0.1.98\r\nsentry-sdk 1.32.0\r\nsetproctitle 1.3.3\r\nsetuptools 68.2.2\r\nsix 1.16.0\r\nsmmap 5.0.1\r\nsniffio 1.3.0\r\nsoupsieve 2.5\r\nstack-data 0.6.3\r\nsympy 1.12\r\ntensorboard 2.15.0\r\ntensorboard-data-server 0.7.1\r\nterminado 0.17.1\r\ntimm 0.9.8\r\ntinycss2 1.2.1\r\ntokenizers 0.13.3\r\ntomli 2.0.1\r\ntorch 2.0.1+cu118\r\ntorchmetrics 1.2.0\r\ntorchvision 0.15.2+cu118\r\ntornado 6.3.3\r\ntqdm 4.66.1\r\ntraitlets 5.11.2\r\ntransformers 4.29.2\r\ntriton 2.0.0\r\ntypes-python-dateutil 2.8.19.14\r\ntyping_extensions 4.8.0\r\ntyping-inspect 0.9.0\r\ntzdata 2023.3\r\nuri-template 1.3.0\r\nurllib3 2.0.7\r\nwandb 0.15.12\r\nwcwidth 0.2.8\r\nwebcolors 1.13\r\nwebdataset 0.2.62\r\nwebencodings 0.5.1\r\nwebsocket-client 1.6.4\r\nWerkzeug 3.0.0\r\nwheel 0.41.2\r\nxformers 0.0.20\r\nxxhash 3.4.1\r\nyarl 1.9.2\r\nzipp 3.17.0\r\n```",
"@humpydonkey FWIW setting fsspec down to 2023.9.2 fixed the issue\r\n\r\n`pip install fsspec==2023.9.2`",
"got it, thanks @ZachNagengast ",
"Thanks for reporting and for the investigation, @ZachNagengast! :hugs: \r\n\r\nWe are investigating the root cause of the issue. In the meantime, we are going to pin fsspec < 2023.10.0. ",
"https://stackoverflow.com/questions/77433096/notimplementederror-loading-a-dataset-cached-in-a-localfilesystem-is-not-suppor/77433141#77433141",
"You can also update `datasets`:\r\n\r\n```\r\npip install -U datasets\r\n```\r\n\r\nIt will also update `fsspec` to use the right version",
"This seems to work fine in 2.19.0. Hopefully it will not break again",
"not working for 2.19.1 !"
] | ### Describe the bug
Loading a streaming dataset with this version of fsspec fails with the following error:
`NotImplementedError: Loading a streaming dataset cached in a LocalFileSystem is not supported yet.`
I suspect the issue is with this PR
https://github.com/fsspec/filesystem_spec/pull/1381
### Steps to reproduce the bug
1. Upgrade fsspec to version `2023.10.0`
2. Attempt to load a streaming dataset e.g. `load_dataset("laion/gpt4v-emotion-dataset", split="train", streaming=True)`
3. Observe the following exception:
```
File "/opt/hostedtoolcache/Python/3.11.6/x64/lib/python3.11/site-packages/datasets/load.py", line 2146, in load_dataset
return builder_instance.as_streaming_dataset(split=split)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/hostedtoolcache/Python/3.11.6/x64/lib/python3.11/site-packages/datasets/builder.py", line 1318, in as_streaming_dataset
raise NotImplementedError(
NotImplementedError: Loading a streaming dataset cached in a LocalFileSystem is not supported yet.
```
### Expected behavior
Should stream the dataset as normal.
### Environment info
datasets@main
fsspec==2023.10.0 | 6,330 |
https://github.com/huggingface/datasets/issues/6329 | شبکه های متن به گفتار ابتدا متن داده شده را به بازنمایی میانی | [] | شبکه های متن به گفتار ابتدا متن داده شده را به بازنمایی میانی
| 6,329 |
https://github.com/huggingface/datasets/issues/6328 | شبکه های متن به گفتار ابتدا متن داده شده را به بازنمایی میانی | [
"شبکه های متن به گفتار ابتدا متن داده شده را به بازنمایی میانی"
] | null | 6,328 |
https://github.com/huggingface/datasets/issues/6327 | FileNotFoundError when trying to load the downloaded dataset with `load_dataset(..., streaming=True)` | [
"You can clone the `togethercomputer/RedPajama-Data-1T-Sample` repo and load the dataset with `load_dataset(\"path/to/cloned_repo\")` to use it offline.",
"@mariosasko Thank you for your kind reply! I'll try it as a workaround.\r\nDoes that mean that currently it's not supported to simply load with a short name?",
"It is, but manually downloading repo files to the cache can easily lead to failure (the HF cache is not meant to be modified by a user besides deleting the files 🙂), as in your case. Hence, the clone + `load_dataset(\"path/to/cloned_repo\")` workflow should be used instead."
] | ### Describe the bug
Hi, I'm trying to load the dataset `togethercomputer/RedPajama-Data-1T-Sample` with `load_dataset` in streaming mode, i.e., `streaming=True`, but `FileNotFoundError` occurs.
### Steps to reproduce the bug
I've downloaded the dataset and save it to the cache dir in advance. My hope is loading the files in offline environment and without taking too much hours to prepross the entire data before running into the training process.
So I try the following code to load the files streamingly
```py
dataset = load_dataset('togethercomputer/RedPajama-Data-1T-Sample', streaming=True)
print(next(iter(dataset['train'])))
```
Sadly, it raises the following:
```
FileNotFoundError: [Errno 2] No such file or directory: 'CURRENT_CODE_PATH/arxiv_sample.jsonl'
```
I've noticed that the dataset can be properly found in the begining
```
Using the latest cached version of the module from /root/.cache/huggingface/modules/datasets_modules/datasets/togethercomputer--RedPajama-Data-1T-Sample/6ea3bc8ec2e84ec6d2df1930942e9028ace8c5b9d9143823cf911c50bbd92039 (last modified on Sat Oct 21 20:12:57 2023) since it couldn't be found locally at togethercomputer/RedPajama-Data-1T-Sample., or remotely on the Hugging Face Hub.
```
But it seems that the paths couldn't be properly parsed when loading iteratively.
How should I fix this error. I've tried specifying `data_files` or `data_dir` as `.../arxiv_sample.jsonl` but none of them works.
Thanks.
### Expected behavior
Properly load the dataset.
### Environment info
`datasets==2.14.5` | 6,327 |
https://github.com/huggingface/datasets/issues/6324 | Conversion to Arrow fails due to wrong type heuristic | [
"Unlike Pandas, Arrow is strict with types, so converting the problematic strings to ints (or ints to strings) to ensure all the values have the same type is the only fix. \r\n\r\nJSON support has been requested in Arrow [here](https://github.com/apache/arrow/issues/32538), but I don't expect this to be implemented soon. \r\n\r\nAlso, this type could be represented with the Arrow Union type. However, due to low usage, the Union type has limited support in the Arrow ecosystem (e.g., IIRC Parquet still does not support it). So, we should probably wait a bit more before adding support for it in `datasets`",
"> Unlike Pandas, Arrow is strict with types, so converting the problematic strings to ints (or ints to strings) to ensure all the values have the same type is the only fix.\r\n> \r\n> JSON support has been requested in Arrow [here](https://github.com/apache/arrow/issues/32538), but I don't expect this to be implemented soon.\r\n> \r\n> Also, this type could be represented with the Arrow Union type. However, due to low usage, the Union type has limited support in the Arrow ecosystem (e.g., IIRC Parquet still does not support it). So, we should probably wait a bit more before adding support for it in `datasets`\r\n\r\nOk many thanks, I was able to mitigate the problem by manually checking and converting all problematic fields now."
] | ### Describe the bug
I have a list of dictionaries with valid/JSON-serializable values.
One key is the denominator for a paragraph. In 99.9% of cases its a number, but there are some occurences of '1a', '2b' and so on.
If trying to convert this list to a dataset with `Dataset.from_list()`, I always get
`ArrowInvalid: Could not convert '1' with type str: tried to convert to int64`, presumably because pyarrow tries to convert the keys to integers.
Is there any way to circumvent this and fix dtypes? I didn't find anything in the documentation.
### Steps to reproduce the bug
* create a list of dicts with one key being a string of an integer for the first few thousand occurences and try to convert to dataset.
### Expected behavior
There shouldn't be an error (e.g. some flag to turn off automatic str to numeric conversion).
### Environment info
- `datasets` version: 2.14.5
- Platform: Linux-5.15.0-84-generic-x86_64-with-glibc2.35
- Python version: 3.9.18
- Huggingface_hub version: 0.17.3
- PyArrow version: 13.0.0
- Pandas version: 2.1.1 | 6,324 |
https://github.com/huggingface/datasets/issues/6323 | Loading dataset from large GCS bucket very slow since 2.14 | [] | ### Describe the bug
Since updating to >2.14 we have very slow access to our parquet files on GCS when loading a dataset (>30 min vs 3s). Our GCS bucket has many objects and resolving globs is very slow. I could track down the problem to this change:
https://github.com/huggingface/datasets/blame/bade7af74437347a760830466eb74f7a8ce0d799/src/datasets/data_files.py#L348
The underlying implementation with gcsfs is really slow. Could you go back to the old way if we are simply giving the parquet files and no glob pattern?
Thank you.
### Steps to reproduce the bug
Load a dataset from a GCS bucket that has many files.
### Expected behavior
Used to be fast (3s) in 2.13
### Environment info
datasets==2.14.5 | 6,323 |
https://github.com/huggingface/datasets/issues/6320 | Dataset slice splits can't load training and validation at the same time | [
"The expression \"train+test\" concatenates the splits.\r\n\r\nThe individual splits as separate datasets can be obtained as follows:\r\n```python\r\ntrain_ds, test_ds = load_dataset(\"<dataset_name>\", split=[\"train\", \"test\"])\r\ntrain_10pct_ds, test_10pct_ds = load_dataset(\"<dataset_name>\", split=[\"train[:10%]\", \"test[:%10]\"])\r\n```"
] | ### Describe the bug
According to the [documentation](https://huggingface.co/docs/datasets/v2.14.5/loading#slice-splits) is should be possible to run the following command:
`train_test_ds = datasets.load_dataset("bookcorpus", split="train+test")`
to load the train and test sets from the dataset.
However executing the equivalent code:
`speech_commands_v1 = load_dataset("superb", "ks", split="train+test")`
only yields the following output:
> Dataset({
> features: ['file', 'audio', 'label'],
> num_rows: 54175
> })
Where loading the dataset without the split argument yields:
> DatasetDict({
> train: Dataset({
> features: ['file', 'audio', 'label'],
> num_rows: 51094
> })
> validation: Dataset({
> features: ['file', 'audio', 'label'],
> num_rows: 6798
> })
> test: Dataset({
> features: ['file', 'audio', 'label'],
> num_rows: 3081
> })
> })
Thus, the API seems to be broken in this regard.
This is a bit annoying since I want to be able to use the split argument with `split="train[:10%]+test[:10%]"` to have smaller dataset to work with when validating my model is working correctly.
### Steps to reproduce the bug
`speech_commands_v1 = load_dataset("superb", "ks", split="train+test")`
### Expected behavior
> DatasetDict({
> train: Dataset({
> features: ['file', 'audio', 'label'],
> num_rows: 51094
> })
> test: Dataset({
> features: ['file', 'audio', 'label'],
> num_rows: 3081
> })
> })
### Environment info
```
import datasets
print(datasets.__version__)
```
> 2.14.5
```
import sys
print(sys.version)
```
> 3.9.17 (main, Jul 5 2023, 20:41:20)
> [GCC 11.2.0] | 6,320 |
https://github.com/huggingface/datasets/issues/6319 | Datasets.map is severely broken | [
"Hi! Instead of processing a single example at a time, you should use the batched `map` for the best performance (with `num_proc=1`) - the fast tokenizers can process a batch's samples in parallel in that scenario.\r\n\r\nE.g., the following code in Colab takes an hour to complete:\r\n```python\r\n# !pip install datasets transformers\r\nfrom datasets import load_dataset\r\nfrom transformers import AutoTokenizer\r\ntokenizer = AutoTokenizer.from_pretrained(\"bert-base-cased\")\r\ndataset = dataset.map(lambda ex: tokenizer(ex[\"text\"]), batched=True, remove_columns=[\"text\", \"meta\"])\r\n```",
"Batched is far worse. A single batch of 1000 took hours and that was only 1%\r\n\r\n\r\nOn Thu, Oct 19, 2023, 2:26 PM Mario Šaško ***@***.***> wrote:\r\n\r\n> Hi! You should use the batched map for the best performance (with\r\n> num_proc=1) - the fast tokenizers can process a batch's samples in\r\n> parallel.\r\n>\r\n> E.g., the following code in Colab takes an hour to complete:\r\n>\r\n> # !pip install datasets transformersfrom datasets import load_datasetfrom transformers import AutoTokenizertokenizer = AutoTokenizer.from_pretrained(\"bert-base-cased\")dataset = dataset.map(lambda ex: tokenizer(ex[\"text\"]), batched=True, remove_columns=[\"text\", \"meta\"])\r\n>\r\n> —\r\n> Reply to this email directly, view it on GitHub\r\n> <https://github.com/huggingface/datasets/issues/6319#issuecomment-1771503757>,\r\n> or unsubscribe\r\n> <https://github.com/notifications/unsubscribe-auth/ABDD3ZJHPSRVDEXFNMXR2N3YAFWFZAVCNFSM6AAAAAA6HDKPSCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONZRGUYDGNZVG4>\r\n> .\r\n> You are receiving this because you authored the thread.Message ID:\r\n> ***@***.***>\r\n>\r\n",
"Can you please provide a self-contained reproducer?",
"Which specific version of datasets are you using?\r\n\r\nWhat is the architecture of your colab setup? Ram? Cores? OS?\r\n\r\n\r\nOn Thu, Oct 19, 2023, 2:27 PM pensive introvert ***@***.***>\r\nwrote:\r\n\r\n> Batched is far worse. A single batch of 1000 took hours and that was only\r\n> 1%\r\n>\r\n>\r\n> On Thu, Oct 19, 2023, 2:26 PM Mario Šaško ***@***.***>\r\n> wrote:\r\n>\r\n>> Hi! You should use the batched map for the best performance (with\r\n>> num_proc=1) - the fast tokenizers can process a batch's samples in\r\n>> parallel.\r\n>>\r\n>> E.g., the following code in Colab takes an hour to complete:\r\n>>\r\n>> # !pip install datasets transformersfrom datasets import load_datasetfrom transformers import AutoTokenizertokenizer = AutoTokenizer.from_pretrained(\"bert-base-cased\")dataset = dataset.map(lambda ex: tokenizer(ex[\"text\"]), batched=True, remove_columns=[\"text\", \"meta\"])\r\n>>\r\n>> —\r\n>> Reply to this email directly, view it on GitHub\r\n>> <https://github.com/huggingface/datasets/issues/6319#issuecomment-1771503757>,\r\n>> or unsubscribe\r\n>> <https://github.com/notifications/unsubscribe-auth/ABDD3ZJHPSRVDEXFNMXR2N3YAFWFZAVCNFSM6AAAAAA6HDKPSCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONZRGUYDGNZVG4>\r\n>> .\r\n>> You are receiving this because you authored the thread.Message ID:\r\n>> ***@***.***>\r\n>>\r\n>\r\n",
"from functools import partial\r\nimport transformers\r\nfrom datasets import load_dataset, concatenate_datasets, load_from_disk\r\n\r\nmodel_name_or_path=\"/opt/data/data/daryl149/llama-2-7b-chat-hf\"\r\noutput_dir=\"/opt/data/data/LongLoRA/checkpoints\"\r\ncache_dir=\"/opt/data/data/LongLoRA/cache\"\r\nmodel_max_length=16384\r\n\r\nIGNORE_INDEX = -100\r\nDEFAULT_PAD_TOKEN = \"[PAD]\"\r\nDEFAULT_EOS_TOKEN = \"</s>\"\r\nDEFAULT_BOS_TOKEN = \"<s>\"\r\nDEFAULT_UNK_TOKEN = \"<unk>\"\r\n\r\n\r\ntokenizer = transformers.LlamaTokenizerFast.from_pretrained(\r\n model_name_or_path,\r\n cache_dir=cache_dir,\r\n model_max_length=model_max_length,\r\n padding_side=\"right\",\r\n use_fast=True,\r\n #use_fast=False\r\n)\r\n\r\nspecial_tokens_dict = dict()\r\nif tokenizer.pad_token is None:\r\n special_tokens_dict[\"pad_token\"] = DEFAULT_PAD_TOKEN\r\nif tokenizer.eos_token is None:\r\n special_tokens_dict[\"eos_token\"] = DEFAULT_EOS_TOKEN\r\nif tokenizer.bos_token is None:\r\n special_tokens_dict[\"bos_token\"] = DEFAULT_BOS_TOKEN\r\nif tokenizer.unk_token is None:\r\n special_tokens_dict[\"unk_token\"] = DEFAULT_UNK_TOKEN\r\n\r\ntokenizer.add_special_tokens(special_tokens_dict)\r\n\r\ndef tokenize_fn(tokenizer, example):\r\n context_length = tokenizer.model_max_length\r\n outputs = tokenizer(\r\n tokenizer.eos_token.join(example[\"text\"]),\r\n #truncation=False,\r\n truncation=True,\r\n return_tensors=\"pt\",\r\n #return_tensors=\"np\",\r\n pad_to_multiple_of=context_length,\r\n padding=True,\r\n )\r\n return {\"input_ids\": outputs[\"input_ids\"].view(-1, context_length)}\r\n\r\nfor idx in range(100):\r\n dataset = load_dataset(\"togethercomputer/RedPajama-Data-1T-Sample\",\r\ncache_dir=cache_dir, split=f'train[{idx}%:{idx+1}%]')\r\n dataset = dataset.map(partial(tokenize_fn, tokenizer), batched=False,\r\nnum_proc=16, remove_columns=[\"text\", \"meta\"])\r\n dataset.save_to_disk(training_args.cache_dir + f\"/training_data_{idx}\")\r\n\r\n\r\nOn Thu, Oct 19, 2023 at 2:30 PM Mario Šaško ***@***.***>\r\nwrote:\r\n\r\n> Can you please provide a self-contained reproducer?\r\n>\r\n> —\r\n> Reply to this email directly, view it on GitHub\r\n> <https://github.com/huggingface/datasets/issues/6319#issuecomment-1771509229>,\r\n> or unsubscribe\r\n> <https://github.com/notifications/unsubscribe-auth/ABDD3ZNBZ3BE7Q4EQZZK6MLYAFWURAVCNFSM6AAAAAA6HDKPSCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONZRGUYDSMRSHE>\r\n> .\r\n> You are receiving this because you authored the thread.Message ID:\r\n> ***@***.***>\r\n>\r\n",
"I changed the tokenizer to one without \"Fast suffix, and something changed.\r\nThe fraction, although still slowed a lot at 80% was able to get over the\r\nfinish line of 100%\r\n\r\nI have to do more testng, see if the whole set can be processed\r\n\r\n\r\n\r\nOn Thu, Oct 19, 2023 at 3:03 PM pensive introvert <\r\n***@***.***> wrote:\r\n\r\n> from functools import partial\r\n> import transformers\r\n> from datasets import load_dataset, concatenate_datasets, load_from_disk\r\n>\r\n> model_name_or_path=\"/opt/data/data/daryl149/llama-2-7b-chat-hf\"\r\n> output_dir=\"/opt/data/data/LongLoRA/checkpoints\"\r\n> cache_dir=\"/opt/data/data/LongLoRA/cache\"\r\n> model_max_length=16384\r\n>\r\n> IGNORE_INDEX = -100\r\n> DEFAULT_PAD_TOKEN = \"[PAD]\"\r\n> DEFAULT_EOS_TOKEN = \"</s>\"\r\n> DEFAULT_BOS_TOKEN = \"<s>\"\r\n> DEFAULT_UNK_TOKEN = \"<unk>\"\r\n>\r\n>\r\n> tokenizer = transformers.LlamaTokenizerFast.from_pretrained(\r\n> model_name_or_path,\r\n> cache_dir=cache_dir,\r\n> model_max_length=model_max_length,\r\n> padding_side=\"right\",\r\n> use_fast=True,\r\n> #use_fast=False\r\n> )\r\n>\r\n> special_tokens_dict = dict()\r\n> if tokenizer.pad_token is None:\r\n> special_tokens_dict[\"pad_token\"] = DEFAULT_PAD_TOKEN\r\n> if tokenizer.eos_token is None:\r\n> special_tokens_dict[\"eos_token\"] = DEFAULT_EOS_TOKEN\r\n> if tokenizer.bos_token is None:\r\n> special_tokens_dict[\"bos_token\"] = DEFAULT_BOS_TOKEN\r\n> if tokenizer.unk_token is None:\r\n> special_tokens_dict[\"unk_token\"] = DEFAULT_UNK_TOKEN\r\n>\r\n> tokenizer.add_special_tokens(special_tokens_dict)\r\n>\r\n> def tokenize_fn(tokenizer, example):\r\n> context_length = tokenizer.model_max_length\r\n> outputs = tokenizer(\r\n> tokenizer.eos_token.join(example[\"text\"]),\r\n> #truncation=False,\r\n> truncation=True,\r\n> return_tensors=\"pt\",\r\n> #return_tensors=\"np\",\r\n> pad_to_multiple_of=context_length,\r\n> padding=True,\r\n> )\r\n> return {\"input_ids\": outputs[\"input_ids\"].view(-1, context_length)}\r\n>\r\n> for idx in range(100):\r\n> dataset = load_dataset(\"togethercomputer/RedPajama-Data-1T-Sample\",\r\n> cache_dir=cache_dir, split=f'train[{idx}%:{idx+1}%]')\r\n> dataset = dataset.map(partial(tokenize_fn, tokenizer), batched=False,\r\n> num_proc=16, remove_columns=[\"text\", \"meta\"])\r\n> dataset.save_to_disk(training_args.cache_dir + f\"/training_data_{idx}\")\r\n>\r\n>\r\n> On Thu, Oct 19, 2023 at 2:30 PM Mario Šaško ***@***.***>\r\n> wrote:\r\n>\r\n>> Can you please provide a self-contained reproducer?\r\n>>\r\n>> —\r\n>> Reply to this email directly, view it on GitHub\r\n>> <https://github.com/huggingface/datasets/issues/6319#issuecomment-1771509229>,\r\n>> or unsubscribe\r\n>> <https://github.com/notifications/unsubscribe-auth/ABDD3ZNBZ3BE7Q4EQZZK6MLYAFWURAVCNFSM6AAAAAA6HDKPSCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONZRGUYDSMRSHE>\r\n>> .\r\n>> You are receiving this because you authored the thread.Message ID:\r\n>> ***@***.***>\r\n>>\r\n>\r\n",
"So, using LlamaTokenizerFast was the problem. Changing it to LlamaTokenizer\r\nfixed things,\r\n\r\nOn Thu, Oct 19, 2023 at 4:04 PM pensive introvert <\r\n***@***.***> wrote:\r\n\r\n> I changed the tokenizer to one without \"Fast suffix, and something\r\n> changed. The fraction, although still slowed a lot at 80% was able to get\r\n> over the finish line of 100%\r\n>\r\n> I have to do more testng, see if the whole set can be processed\r\n>\r\n>\r\n>\r\n> On Thu, Oct 19, 2023 at 3:03 PM pensive introvert <\r\n> ***@***.***> wrote:\r\n>\r\n>> from functools import partial\r\n>> import transformers\r\n>> from datasets import load_dataset, concatenate_datasets, load_from_disk\r\n>>\r\n>> model_name_or_path=\"/opt/data/data/daryl149/llama-2-7b-chat-hf\"\r\n>> output_dir=\"/opt/data/data/LongLoRA/checkpoints\"\r\n>> cache_dir=\"/opt/data/data/LongLoRA/cache\"\r\n>> model_max_length=16384\r\n>>\r\n>> IGNORE_INDEX = -100\r\n>> DEFAULT_PAD_TOKEN = \"[PAD]\"\r\n>> DEFAULT_EOS_TOKEN = \"</s>\"\r\n>> DEFAULT_BOS_TOKEN = \"<s>\"\r\n>> DEFAULT_UNK_TOKEN = \"<unk>\"\r\n>>\r\n>>\r\n>> tokenizer = transformers.LlamaTokenizerFast.from_pretrained(\r\n>> model_name_or_path,\r\n>> cache_dir=cache_dir,\r\n>> model_max_length=model_max_length,\r\n>> padding_side=\"right\",\r\n>> use_fast=True,\r\n>> #use_fast=False\r\n>> )\r\n>>\r\n>> special_tokens_dict = dict()\r\n>> if tokenizer.pad_token is None:\r\n>> special_tokens_dict[\"pad_token\"] = DEFAULT_PAD_TOKEN\r\n>> if tokenizer.eos_token is None:\r\n>> special_tokens_dict[\"eos_token\"] = DEFAULT_EOS_TOKEN\r\n>> if tokenizer.bos_token is None:\r\n>> special_tokens_dict[\"bos_token\"] = DEFAULT_BOS_TOKEN\r\n>> if tokenizer.unk_token is None:\r\n>> special_tokens_dict[\"unk_token\"] = DEFAULT_UNK_TOKEN\r\n>>\r\n>> tokenizer.add_special_tokens(special_tokens_dict)\r\n>>\r\n>> def tokenize_fn(tokenizer, example):\r\n>> context_length = tokenizer.model_max_length\r\n>> outputs = tokenizer(\r\n>> tokenizer.eos_token.join(example[\"text\"]),\r\n>> #truncation=False,\r\n>> truncation=True,\r\n>> return_tensors=\"pt\",\r\n>> #return_tensors=\"np\",\r\n>> pad_to_multiple_of=context_length,\r\n>> padding=True,\r\n>> )\r\n>> return {\"input_ids\": outputs[\"input_ids\"].view(-1, context_length)}\r\n>>\r\n>> for idx in range(100):\r\n>> dataset = load_dataset(\"togethercomputer/RedPajama-Data-1T-Sample\",\r\n>> cache_dir=cache_dir, split=f'train[{idx}%:{idx+1}%]')\r\n>> dataset = dataset.map(partial(tokenize_fn, tokenizer), batched=False,\r\n>> num_proc=16, remove_columns=[\"text\", \"meta\"])\r\n>> dataset.save_to_disk(training_args.cache_dir +\r\n>> f\"/training_data_{idx}\")\r\n>>\r\n>>\r\n>> On Thu, Oct 19, 2023 at 2:30 PM Mario Šaško ***@***.***>\r\n>> wrote:\r\n>>\r\n>>> Can you please provide a self-contained reproducer?\r\n>>>\r\n>>> —\r\n>>> Reply to this email directly, view it on GitHub\r\n>>> <https://github.com/huggingface/datasets/issues/6319#issuecomment-1771509229>,\r\n>>> or unsubscribe\r\n>>> <https://github.com/notifications/unsubscribe-auth/ABDD3ZNBZ3BE7Q4EQZZK6MLYAFWURAVCNFSM6AAAAAA6HDKPSCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONZRGUYDSMRSHE>\r\n>>> .\r\n>>> You are receiving this because you authored the thread.Message ID:\r\n>>> ***@***.***>\r\n>>>\r\n>>\r\n",
"Indeed, the tokenizer is super slow. Perhaps @ArthurZucker knows the reason why.\r\n\r\n([This](https://colab.research.google.com/drive/1VgeurX-4Fl2X6aBQTwh_X4kuQKZ6K9L1?usp=sharing) simplified Colab can be used to reproduce the behavior)",
"same issue here\r\nsample to reproduce: https://github.com/philschmid/document-ai-transformers/blob/main/training/donut_sroie.ipynb\r\nwith following map line\r\nhttps://github.com/philschmid/document-ai-transformers/blob/main/training/donut_sroie.ipynb\r\n\r\nIf I directly iterate over the dataset and call the mapping method, it is very fast\r\n```py\r\nfor sample in dataset:\r\n def preprocess_documents_for_donut(sample):\r\n```\r\n\r\nif i removed `.convert('RGB')` It can run to completion without getting stuck. I suspect it has something to do with the Image.\r\n\r\nIf I use batch, it's even slower.",
"@ewfian \r\n\r\n> If I directly iterate over the dataset and call the mapping method, it is very fast\r\n\r\n`Dataset.map` must also convert the images into bytes to write them to an Arrow file (the write itself takes some time, too). \r\n\r\nYou can make the `map` faster by manually converting the images into an \"arrow-compatible\" representation. Otherwise, the Pillow defaults are used when saving an image, which seems particularly slow for the notebook's case.\r\n\r\n```python\r\ndef preprocess_documents_for_donut(sample):\r\n text = json.loads(sample[\"text\"])\r\n d_doc = task_start_token + json2token(text) + eos_token\r\n image = sample[\"image\"].convert('RGB')\r\n # convert image to bytes\r\n buffer = io.BytesIO()\r\n image.save(buffer, format=\"PNG\", compress_level=1)\r\n return {\"image\": {\"bytes\": buffer.getvalue()}, \"text\": d_doc}\r\n\r\nproc_dataset = dataset.map(preprocess_documents_for_donut, writer_batch_size=50)\r\n```",
"The problem I had was to do with map using fork and copying locks from the\r\nparent process in acquired state. I ended up changing the context to use\r\nforkserver instead.\r\n\r\n\r\nOn Wed, Nov 29, 2023, 10:04 PM Mario Šaško ***@***.***> wrote:\r\n\r\n> @ewfian <https://github.com/ewfian>\r\n>\r\n> If I directly iterate over the dataset and call the mapping method, it is\r\n> very fast\r\n>\r\n> Dataset.map must also convert the images into bytes to write them to an\r\n> Arrow file (the write itself takes some time, too).\r\n>\r\n> You can make the map faster by manually converting the images into an\r\n> \"arrow-compatible\" representation. Otherwise, the Pillow defaults are used\r\n> when saving an image, which seems particularly slow for the notebook's case.\r\n>\r\n> def preprocess_documents_for_donut(sample):\r\n> text = json.loads(sample[\"text\"])\r\n> d_doc = task_start_token + json2token(text) + eos_token\r\n> image = sample[\"image\"].convert('RGB')\r\n> # convert image to bytes\r\n> buffer = io.BytesIO()\r\n> image.save(buffer, format=\"PNG\", compress_level=1)\r\n> return {\"image\": {\"bytes\": buffer.getvalue()}, \"text\": d_doc}\r\n> proc_dataset = dataset.map(preprocess_documents_for_donut, writer_batch_size=50)\r\n>\r\n> —\r\n> Reply to this email directly, view it on GitHub\r\n> <https://github.com/huggingface/datasets/issues/6319#issuecomment-1833033973>,\r\n> or unsubscribe\r\n> <https://github.com/notifications/unsubscribe-auth/ABDD3ZKKEKJVWBFH7QHLRJ3YG7ZUJAVCNFSM6AAAAAA6HDKPSCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMZTGAZTGOJXGM>\r\n> .\r\n> You are receiving this because you authored the thread.Message ID:\r\n> ***@***.***>\r\n>\r\n"
] | ### Describe the bug
Regardless of how many cores I used, I have 16 or 32 threads, map slows down to a crawl at around 80% done, lingers maybe until 97% extremely slowly and NEVER finishes the job. It just hangs.
After watching this for 27 hours I control-C out of it. Until the end one process appears to be doing something, but it never ends.
I saw some comments about fast tokenizers using Rust and all and tried different variations. NOTHING works.
### Steps to reproduce the bug
Running it without breaking the dataset into parts results in the same behavior. The loop was an attempt to see if this was a RAM issue.
for idx in range(100):
dataset = load_dataset("togethercomputer/RedPajama-Data-1T-Sample", cache_dir=cache_dir, split=f'train[{idx}%:{idx+1}%]')
dataset = dataset.map(partial(tokenize_fn, tokenizer), batched=False, num_proc=1, remove_columns=["text", "meta"])
dataset.save_to_disk(training_args.cache_dir + f"/training_data_{idx}")
### Expected behavior
I expect map to run at more or less the same speed it starts with and FINISH its processing.
### Environment info
Python 3.8, same with 3.10 makes no difference.
Ubuntu 20.04, | 6,319 |
https://github.com/huggingface/datasets/issues/6317 | sentiment140 dataset unavailable | [
"Thanks for reporting. We are investigating the issue.",
"We have opened an issue in the corresponding Hub dataset: https://huggingface.co/datasets/sentiment140/discussions/3\r\n\r\nLet's continue the discussion there."
] | ### Describe the bug
loading the dataset using load_dataset("sentiment140") returns the following error
ConnectionError: Couldn't reach http://cs.stanford.edu/people/alecmgo/trainingandtestdata.zip (error 403)
### Steps to reproduce the bug
Run the following code (version should not matter).
```
from datasets import load_dataset
data = load_dataset("sentiment140")
```
### Expected behavior
The dataset should be loaded just like any other.
The main issue is that it is no longer hosted by stanford. It is still available from a [Google Drive Link](https://docs.google.com/file/d/0B04GJPshIjmPRnZManQwWEdTZjg/edit).
### Environment info
- `datasets` version: 2.14.5
- Platform: Windows-10-10.0.19045-SP0
- Python version: 3.10.8
- Huggingface_hub version: 0.17.3
- PyArrow version: 13.0.0
- Pandas version: 2.1.1 | 6,317 |
https://github.com/huggingface/datasets/issues/6315 | Hub datasets with CSV metadata raise ArrowInvalid: JSON parse error: Invalid value. in row 0 | [] | When trying to load a Hub dataset that contains a CSV metadata file, it raises an `ArrowInvalid` error:
```
E pyarrow.lib.ArrowInvalid: JSON parse error: Invalid value. in row 0
pyarrow/error.pxi:100: ArrowInvalid
```
See: https://huggingface.co/datasets/lukarape/public_small_papers/discussions/1 | 6,315 |
https://github.com/huggingface/datasets/issues/6311 | cast_column to Sequence with length=4 occur exception raise in datasets/table.py:2146 | [
"Thanks for reporting! We've spotted the bugs with the `array.values` handling and are fixing them in https://github.com/huggingface/datasets/pull/6283 (should be part of the next release).",
"> Thanks for reporting! We've spotted the bugs with the `array.values` handling and are fixing them in #6283 (should be part of the next release).\r\n\r\ni encounter another exception while cast_column to type `Sequence(feature={\"points\": Array2D(shape=(-1, 2), dtype=\"int64\"), \"label\": ClassLabel(num_classes=num_classes, names=names)})`\r\n\r\nwhile my data like this: '{\"points\": [[0.6,0.6], [0.7,0.7], [0.8,0.8]], \"label\": \"A1\"}'\r\n\r\nhere is the backtrace info:\r\n\r\n```\r\n out = func(dataset, *args, **kwargs)\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py\", line 2110, in cast_column\r\n return self.cast(features)\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py\", line 2055, in cast\r\n dataset = dataset.map(\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py\", line 592, in wrapper\r\n out: Union[\"Dataset\", \"DatasetDict\"] = func(self, *args, **kwargs)\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py\", line 557, in wrapper\r\n out: Union[\"Dataset\", \"DatasetDict\"] = func(self, *args, **kwargs)\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py\", line 3097, in map\r\n for rank, done, content in Dataset._map_single(**dataset_kwargs):\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py\", line 3474, in _map_single\r\n batch = apply_function_on_filtered_inputs(\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py\", line 3353, in apply_function_on_filtered_inputs\r\n processed_inputs = function(*fn_args, *additional_args, **fn_kwargs)\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 2329, in table_cast\r\n return cast_table_to_schema(table, schema)\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 2288, in cast_table_to_schema\r\n arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 2288, in <listcomp>\r\n arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 1831, in wrapper\r\n return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 1831, in <listcomp>\r\n return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 2073, in cast_array_to_feature\r\n arrays = [_c(array.field(name), subfeature) for name, subfeature in feature.items()]\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 2073, in <listcomp>\r\n arrays = [_c(array.field(name), subfeature) for name, subfeature in feature.items()]\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 1833, in wrapper\r\n return func(array, *args, **kwargs)\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 2095, in cast_array_to_feature\r\n casted_values = _c(array.values, feature.feature)\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 1833, in wrapper\r\n return func(array, *args, **kwargs)\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 2144, in cast_array_to_feature\r\n return array_cast(array, feature(), allow_number_to_str=allow_number_to_str)\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 1833, in wrapper\r\n return func(array, *args, **kwargs)\r\n File \"/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py\", line 1967, in array_cast\r\n return pa_type.wrap_array(array)\r\n File \"pyarrow/types.pxi\", line 1369, in pyarrow.lib.BaseExtensionType.wrap_array\r\nTypeError: Incompatible storage type for extension<arrow.py_extension_type<Array2DExtensionType>>: expected list<item: list<item: double>>, got list<item: double>\r\n```\r\nand i print(array) in datasets/table.py:1967 indeed get 2D list. is that same issue in #6283 ?\r\n\r\nbesides this, hugging face datasets seems don't naturally support multi-labels which means `Sequence(ClassLabel)` illegal if data is [\"label1\", \"label2\"]. so i have to define a class derived from `ClassLabel`, like this:\r\n\r\n```\r\nclass AisClassLabels(ClassLabel):\r\n def encode_example(self, example_data):\r\n if self.num_classes is None:\r\n raise ValueError(\r\n \"Trying to use ClassLabel feature with undefined number of class. \"\r\n \"Please set ClassLabel.names or num_classes.\"\r\n )\r\n if not isinstance(example_data, list):\r\n example_data = [example_data]\r\n\r\n for i in range(len(example_data)):\r\n if isinstance(example_data[i], str):\r\n example_data[i] = self.str2int(example_data[i])\r\n if not -1 <= example_data[i] < self.num_classes:\r\n raise ValueError(f\"Class label {example_data:d} greater than configured num_classes {self.num_classes}\")\r\n return example_data\r\n```\r\nand it works well in my case. but is there any recommend way to implement multi-labels?",
"`Incompatible storage type for extension<arrow.py_extension_type<Array2DExtensionType>>: expected list<item: list<item: double>>, got list<item: double>`\r\nif i change `Array2D(shape=(-1, 2), dtype=\"int64\")` to `Sequence(Value(\"int64\"))` , every thing goes well. but my data is 2D int list",
"i test Sequence(ClassLabel) is ok if one column is label list. but it is not ok in nested column such as `Sequence(feature= {\"points\": Sequence(Value(\"int32\")), \"label\": Sequence(ClassLabel(num_classes....)))`. in this case i need override ClassLabels. encode_example as i given above."
] | ### Describe the bug
i load a dataset from local csv file which has 187383612 examples, then use `map` to generate new columns for test.
here is my code :
```
import os
from datasets import load_dataset
from datasets.features import Sequence, Value
def add_new_path(example):
example["ais_bbox"] = [100,100,200,200]
example["ais_image_path"] = os.path.join("images", example["image_path"]) if example["image_path"] else ""
return example
ais_dataset = load_dataset("/data/ryan.gao/ais_dataset_cache/raw/1749/")
hf_ds = ais_dataset.map(add_new_path, batched=False, num_proc=32)
ds = hf_ds.cast_column("ais_bbox", Sequence(Value("int32"), length=4))
```
and the `cast_column` raise an exception
```
Casting the dataset: 3%|███▉
...
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 2110, in cast_column
return self.cast(features)
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 2055, in cast
dataset = dataset.map(
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 592, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 557, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 3097, in map
for rank, done, content in Dataset._map_single(**dataset_kwargs):
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 3474, in _map_single
batch = apply_function_on_filtered_inputs(
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 3353, in apply_function_on_filtered_inputs
processed_inputs = function(*fn_args, *additional_args, **fn_kwargs)
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py", line 2329, in table_cast
return cast_table_to_schema(table, schema)
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py", line 2288, in cast_table_to_schema
arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py", line 2288, in <listcomp>
arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py", line 1831, in wrapper
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py", line 1831, in <listcomp>
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
File "/home/protoss.gao/.local/lib/python3.9/site-packages/datasets/table.py", line 2145, in cast_array_to_feature
raise TypeError(f"Couldn't cast array of type\n{array.type}\nto\n{feature}")
TypeError: Couldn't cast array of type
list<item: int64>
to
Sequence(feature=Value(dtype='int32', id=None), length=4, id=None)
```
i check the source code and make debug info:
in datasets/table.py:2092
```
2091 if feature.length > -1:
2092 if feature.length * len(array) == len(array.values):
2093 return pa.FixedSizeListArray.from_arrays(_c(array.values, feature.feature), feature.length)
2094 print(len(array))
2095 print(len(array.values))
```
my feature.length is 4. but feature.length * len(array) == len(array.values) is false.
print(len(array)) is 262
print(len(array.values)) is 4000
then I use "for item in array" to print each item then get 262 * [100,100,200,200]
and use "for item in array.values" to print each item and get 4000 int32 which are 1000 * [100,100,200,200]
i'm wondering the `chunk` in each `array.chunks`, the "chunk.values" may get all the chunks's value rather than single chunk? but i check the pyarrow's doc seems chunk.values is chunk's value not all.
### Steps to reproduce the bug
code provided above.
### Expected behavior
feature.length * len(array) == len(array.values) should be true. and there should not has Exception.
### Environment info
python3.9
x86_64
datasets: 2.14.4
pyarrow: 13.0.0 or 10.0.0 | 6,311 |
https://github.com/huggingface/datasets/issues/6308 | module 'resource' has no attribute 'error' | [
"This (Windows) issue was fixed in `fsspec` in https://github.com/fsspec/filesystem_spec/pull/1275. So, to avoid the error, update the `fsspec` installation with `pip install -U fsspec`.",
"> This (Windows) issue was fixed in `fsspec` in [fsspec/filesystem_spec#1275](https://github.com/fsspec/filesystem_spec/pull/1275). So, to avoid the error, update the `fsspec` installation with `pip install -U fsspec`.\r\n\r\nafter I run `pip install -U fsspec`\r\n\r\nit occurs a new error:\r\n```\r\nERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflict\r\ns.\r\ndatasets 2.14.5 requires fsspec[http]<2023.9.0,>=2023.1.0, but you have fsspec 2023.9.2 which is incompatible.\r\n\r\n```",
"The `fsspec<2023.9.0` upper bound will be removed in the next release. The `ResourceError` fix is also present in version 2023.6.0, so use that version in the meantime (`pip install fsspec==2023.6.0`).",
"> The `fsspec<2023.9.0` upper bound will be removed in the next release. The `ResourceError` fix is also present in version 2023.6.0, so use that version in the meantime (`pip install fsspec==2023.6.0`).\r\n\r\nthanks for reply！"
] | ### Describe the bug
just run import:
`from datasets import load_dataset`
and then:
```
File "C:\ProgramData\anaconda3\envs\py310\lib\site-packages\datasets\__init__.py", line 22, in <module>
from .arrow_dataset import Dataset
File "C:\ProgramData\anaconda3\envs\py310\lib\site-packages\datasets\arrow_dataset.py", line 66, in <module>
from .arrow_reader import ArrowReader
File "C:\ProgramData\anaconda3\envs\py310\lib\site-packages\datasets\arrow_reader.py", line 30, in <module>
from .download.download_config import DownloadConfig
File "C:\ProgramData\anaconda3\envs\py310\lib\site-packages\datasets\download\__init__.py", line 10, in <module>
from .streaming_download_manager import StreamingDownloadManager
File "C:\ProgramData\anaconda3\envs\py310\lib\site-packages\datasets\download\streaming_download_manager.py", line 21, in <module>
from ..filesystems import COMPRESSION_FILESYSTEMS
File "C:\ProgramData\anaconda3\envs\py310\lib\site-packages\datasets\filesystems\__init__.py", line 8, in <module>
import fsspec.asyn
File "C:\ProgramData\anaconda3\envs\py310\lib\site-packages\fsspec\asyn.py", line 157, in <module>
ResourceEror = resource.error
AttributeError: module 'resource' has no attribute 'error'
Process finished with exit code 1
```
and the error codes are:
```
try:
import resource
except ImportError:
resource = None
ResourceError = OSError
else:
ResourceEror = resource.error
```
1. miss spelling : "ResourceEror " should be "ResourceErorr"
2. module 'resource' has no attribute 'error'
### Steps to reproduce the bug
only one step:
`from datasets import load_dataset`
### Expected behavior
slove error: module 'resource' has no attribute 'error'
### Environment info
python=3.10
datasets==2.14.5
| 6,308 |
https://github.com/huggingface/datasets/issues/6306 | pyinstaller : OSError: could not get source code | [
"more information：\r\n``` \r\nFile \"text2vec\\__init__.py\", line 8, in <module>\r\nFile \"<frozen importlib._bootstrap>\", line 1027, in _find_and_load\r\nFile \"<frozen importlib._bootstrap>\", line 1006, in _find_and_load_unlocked\r\nFile \"<frozen importlib._bootstrap>\", line 688, in _load_unlocked\r\nFile \"PyInstaller\\loader\\pyimod02_importers.py\", line 499, in exec_module\r\nFile \"text2vec\\bertmatching_model.py\", line 19, in <module>\r\nFile \"<frozen importlib._bootstrap>\", line 1027, in _find_and_load\r\nFile \"<frozen importlib._bootstrap>\", line 1006, in _find_and_load_unlocked\r\nFile \"<frozen importlib._bootstrap>\", line 688, in _load_unlocked\r\nFile \"PyInstaller\\loader\\pyimod02_importers.py\", line 499, in exec_module\r\nFile \"text2vec\\bertmatching_dataset.py\", line 7, in <module>\r\nFile \"<frozen importlib._bootstrap>\", line 1027, in _find_and_load\r\nFile \"<frozen importlib._bootstrap>\", line 1006, in _find_and_load_unlocked\r\nFile \"<frozen importlib._bootstrap>\", line 688, in _load_unlocked\r\nFile \"PyInstaller\\loader\\pyimod02_importers.py\", line 499, in exec_module\r\nFile \"datasets\\__init__.py\", line 52, in <module>\r\nFile \"<frozen importlib._bootstrap>\", line 1027, in _find_and_load\r\nFile \"<frozen importlib._bootstrap>\", line 1006, in _find_and_load_unlocked\r\nFile \"<frozen importlib._bootstrap>\", line 688, in _load_unlocked\r\nFile \"PyInstaller\\loader\\pyimod02_importers.py\", line 499, in exec_module\r\nFile \"datasets\\inspect.py\", line 30, in <module>\r\nFile \"<frozen importlib._bootstrap>\", line 1027, in _find_and_load\r\nFile \"<frozen importlib._bootstrap>\", line 1006, in _find_and_load_unlocked\r\nFile \"<frozen importlib._bootstrap>\", line 688, in _load_unlocked\r\nFile \"PyInstaller\\loader\\pyimod02_importers.py\", line 499, in exec_module\r\nFile \"datasets\\load.py\", line 58, in <module>\r\nFile \"<frozen importlib._bootstrap>\", line 1027, in _find_and_load\r\nFile \"<frozen importlib._bootstrap>\", line 1006, in _find_and_load_unlocked\r\nFile \"<frozen importlib._bootstrap>\", line 688, in _load_unlocked\r\nFile \"PyInstaller\\loader\\pyimod02_importers.py\", line 499, in exec_module\r\nFile \"datasets\\packaged_modules\\__init__.py\", line 31, in <module>\r\nFile \"inspect.py\", line 1147, in getsource\r\nFile \"inspect.py\", line 1129, in getsourcelines\r\nFile \"inspect.py\", line 958, in findsource\r\nOSError: could not get source code\r\n```\r\n",
"Can you share a reproducer? I haven't been able to reproduce the error myself.",
"> '\r\n\r\nthanks,I solve it.it's about pyinstaller.",
"1",
"> > '\r\n> \r\n> thanks,I solve it.it's about pyinstaller.\r\n\r\nI encountered the same error, how to solve it?"
] | ### Describe the bug
I ran a package with pyinstaller and got the following error：
### Steps to reproduce the bug
```
...
File "datasets\__init__.py", line 52, in <module>
File "<frozen importlib._bootstrap>", line 1027, in _find_and_load
File "<frozen importlib._bootstrap>", line 1006, in _find_and_load_unlocked
File "<frozen importlib._bootstrap>", line 688, in _load_unlocked
File "PyInstaller\loader\pyimod02_importers.py", line 499, in exec_module
File "datasets\inspect.py", line 30, in <module>
File "<frozen importlib._bootstrap>", line 1027, in _find_and_load
File "<frozen importlib._bootstrap>", line 1006, in _find_and_load_unlocked
File "<frozen importlib._bootstrap>", line 688, in _load_unlocked
File "PyInstaller\loader\pyimod02_importers.py", line 499, in exec_module
File "datasets\load.py", line 58, in <module>
File "<frozen importlib._bootstrap>", line 1027, in _find_and_load
File "<frozen importlib._bootstrap>", line 1006, in _find_and_load_unlocked
File "<frozen importlib._bootstrap>", line 688, in _load_unlocked
File "PyInstaller\loader\pyimod02_importers.py", line 499, in exec_module
File "datasets\packaged_modules\__init__.py", line 31, in <module>
File "inspect.py", line 1147, in getsource
File "inspect.py", line 1129, in getsourcelines
File "inspect.py", line 958, in findsource
OSError: could not get source code
```
### Expected behavior
I have looked up the relevant information, but I can't find a suitable reason
### Environment info
```python
python 3.10
datasets 2.14.4
pyinstaller 5.6.2
``` | 6,306 |
https://github.com/huggingface/datasets/issues/6305 | Cannot load dataset with `2.14.5`: `FileNotFound` error | [
"Thanks for reporting, @finiteautomata.\r\n\r\nWe are investigating it. ",
"There is a bug in `datasets`. You can see our proposed fix:\r\n- #6309 "
] | ### Describe the bug
I'm trying to load [piuba-bigdata/articles_and_comments] and I'm stumbling with this error on `2.14.5`. However, this works on `2.10.0`.
### Steps to reproduce the bug
[Colab link](https://colab.research.google.com/drive/1SAftFMQnFE708ikRnJJHIXZV7R5IBOCE#scrollTo=r2R2ipCCDmsg)
```python
Downloading readme: 100%
1.19k/1.19k [00:00<00:00, 30.9kB/s]
---------------------------------------------------------------------------
FileNotFoundError Traceback (most recent call last)
[<ipython-input-2-807c3583d297>](https://localhost:8080/#) in <cell line: 3>()
1 from datasets import load_dataset
2
----> 3 load_dataset("piuba-bigdata/articles_and_comments", split="train")
2 frames
[/usr/local/lib/python3.10/dist-packages/datasets/load.py](https://localhost:8080/#) in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, verification_mode, ignore_verifications, keep_in_memory, save_infos, revision, token, use_auth_token, task, streaming, num_proc, storage_options, **config_kwargs)
2127
2128 # Create a dataset builder
-> 2129 builder_instance = load_dataset_builder(
2130 path=path,
2131 name=name,
[/usr/local/lib/python3.10/dist-packages/datasets/load.py](https://localhost:8080/#) in load_dataset_builder(path, name, data_dir, data_files, cache_dir, features, download_config, download_mode, revision, token, use_auth_token, storage_options, **config_kwargs)
1813 download_config = download_config.copy() if download_config else DownloadConfig()
1814 download_config.storage_options.update(storage_options)
-> 1815 dataset_module = dataset_module_factory(
1816 path,
1817 revision=revision,
[/usr/local/lib/python3.10/dist-packages/datasets/load.py](https://localhost:8080/#) in dataset_module_factory(path, revision, download_config, download_mode, dynamic_modules_path, data_dir, data_files, **download_kwargs)
1506 raise e1 from None
1507 if isinstance(e1, FileNotFoundError):
-> 1508 raise FileNotFoundError(
1509 f"Couldn't find a dataset script at {relative_to_absolute_path(combined_path)} or any data file in the same directory. "
1510 f"Couldn't find '{path}' on the Hugging Face Hub either: {type(e1).__name__}: {e1}"
FileNotFoundError: Couldn't find a dataset script at /content/piuba-bigdata/articles_and_comments/articles_and_comments.py or any data file in the same directory. Couldn't find 'piuba-bigdata/articles_and_comments' on the Hugging Face Hub either: FileNotFoundError: No (supported) data files or dataset script found in piuba-bigdata/articles_and_comments.
```
### Expected behavior
It should load normally.
### Environment info
```
- `datasets` version: 2.14.5
- Platform: Linux-5.15.120+-x86_64-with-glibc2.35
- Python version: 3.10.12
- Huggingface_hub version: 0.18.0
- PyArrow version: 9.0.0
- Pandas version: 1.5.3
``` | 6,305 |
https://github.com/huggingface/datasets/issues/6303 | Parquet uploads off-by-one naming scheme | [
"You can find the reasoning behind this naming scheme [here](https://github.com/huggingface/transformers/pull/16343#discussion_r931182168).\r\n\r\nThis point has been raised several times, so I'd be okay with starting with `00001-` (also to be consistent with the `transformers` sharding), but I'm not sure @lhoestq agrees.",
"We start at 0 in `datasets` for consistency with Apache Spark, Apache Beam, Dask and others.\r\n\r\nAlso note `transformers` isn't a good reference on this topic. I talked with the maintainers when they added shards but it was already released this way. Though we found that there is a backward-compatible way in `transformers` to start at 0, but no request from `transformers` users to changes this AFAIK.",
"not sure it would be a good idea to break the consistency now, IMO",
"Makes sense to start at 0 for plenty of good reasons so I'm on board.\r\n\r\nWhat about the second part `-of-0000X`? With single commit PR #6269 just getting merged, there was a note about issues with 100+ file edits https://github.com/huggingface/datasets/pull/6269#issuecomment-1755428581.\r\n\r\nThat would be my last remaining concern in the context of the `push_to_hub(..., append=True)` work to be done, where appending a single file to the full dataset will require renaming every other existing file in the dataset. If it doesn't seem like a big issue for this work then all the better 👍"
] | ### Describe the bug
I noticed this numbering scheme not matching up in a different project and wanted to raise it as an issue for discussion, what is the actual proper way to have these stored?
<img width="425" alt="image" src="https://github.com/huggingface/datasets/assets/1981179/3ffa2144-7c9a-446f-b521-a5e9db71e7ce">
The `-SSSSS-of-NNNNN` seems to be used widely across the codebase. The section that creates the part in my screenshot is here https://github.com/huggingface/datasets/blob/main/src/datasets/arrow_dataset.py#L5287
There are also some edits to this section in the single commit branch.
### Steps to reproduce the bug
1. Upload a dataset that requires at least two parquet files in it
2. Observe the naming scheme
### Expected behavior
The couple options here are of course **1. keeping it as is**
**2. Starting the index at 1:**
train-00001-of-00002-{hash}.parquet
train-00002-of-00002-{hash}.parquet
**3. My preferred option** (which would solve my specific issue), dropping the total entirely:
train-00000-{hash}.parquet
train-00001-{hash}.parquet
This also solves an issue that will occur with an `append` variable for `push_to_hub` (see https://github.com/huggingface/datasets/issues/6290) where as you add a new parquet file, you need to rename everything in the repo as well.
However, I know there are parts of the repo that use 0 as the starting file or may require the total, so raising the question for discussion.
### Environment info
- `datasets` version: 2.14.6.dev0
- Platform: macOS-14.0-arm64-arm-64bit
- Python version: 3.10.12
- Huggingface_hub version: 0.18.0
- PyArrow version: 12.0.1
- Pandas version: 1.5.3 | 6,303 |
https://github.com/huggingface/datasets/issues/6302 | ArrowWriter/ParquetWriter `write` method does not increase `_num_bytes` and hence datasets not sharding at `max_shard_size` | [
"`writer._num_bytes` is updated every `writer_batch_size`-th call to the `write` method (default `writer_batch_size` is 1000 (examples)). You should be able to see the update by passing a smaller `writer_batch_size` to the `load_dataset_builder`.\r\n\r\nWe could improve this by supporting the string `writer_batch_size` version as we do with `max_shard_size`, and capping `writer_batch_size` to `max_shard_size` in scenarios where the default `writer_batch_size` > `max_shard_size`. ",
"Thanks, reducing `writer_batch_size` solved my problem :)"
] | ### Describe the bug
An example from [1], does not work when limiting shards with `max_shard_size`.
Try the following example with low `max_shard_size`, such as:
```python
builder.download_and_prepare(output_dir, storage_options=storage_options, file_format="parquet", max_shard_size="10MB")
```
The reason for this is that, in line [2] `writer._num_bytes > max_shard_size` is never true, because the `write` method of `ArrowWriter` [3] does not increase `self._num_bytes`.
Such that respective Arrow/Parquet shards are only written to file based on the `writer_batch_size` or `config.DEFAULT_MAX_BATCH_SIZE`, but not based on `max_shard_size`.
[1] https://huggingface.co/docs/datasets/filesystems#download-and-prepare-a-dataset-into-a-cloud-storage
[2] https://github.com/huggingface/datasets/blob/3e8d420808718c9a1453a2e7ee3484ca12c9c70d/src/datasets/builder.py#L1677
[3] https://github.com/huggingface/datasets/blob/3e8d420808718c9a1453a2e7ee3484ca12c9c70d/src/datasets/arrow_writer.py#L459
### Steps to reproduce the bug
Get example from: https://huggingface.co/docs/datasets/filesystems#download-and-prepare-a-dataset-into-a-cloud-storage
Call `builder.download_and_prepare` with low `max_shard_size` such as `10MB`, e.g.:
```python
builder.download_and_prepare(output_dir, storage_options=storage_options, file_format="parquet", max_shard_size="10MB")
```
### Expected behavior
Shards should be written based on `max_shard_size` instead of batch size.
### Environment info
```
>>> import datasets
>>> datasets.__version__
'2.14.6.dev0
``` | 6,302 |
https://github.com/huggingface/datasets/issues/6299 | Support for newer versions of JAX | [] | ### Feature request
Hi,
I like your idea of adapting the datasets library to be usable with JAX. Thank you for that.
However, in your [setup.py](https://github.com/huggingface/datasets/blob/main/setup.py), you enforce old versions of JAX <= 0.3... It is very cumbersome !
What is the rationale for such a limitation ? Can you remove it please ?
Thanks,
### Motivation
This library is unusable with new versions of JAX ?
### Your contribution
Yes. | 6,299 |
https://github.com/huggingface/datasets/issues/6294 | IndexError: Invalid key is out of bounds for size 0 despite having a populated dataset | [
"It looks to be the same issue as the one reported in https://discuss.huggingface.co/t/indexerror-invalid-key-16-is-out-of-bounds-for-size-0.\r\n\r\nCan you check the length of `train_dataset` before the `train_sampler = self._get_train_sampler()` (and after `_remove_unused_columns`) line?"
] | ### Describe the bug
I am encountering an `IndexError` when trying to access data from a DataLoader which wraps around a dataset I've loaded using the `datasets` library. The error suggests that the dataset size is `0`, but when I check the length and print the dataset, it's clear that it has `1166` entries.
### Steps to reproduce the bug
1. Load a dataset with `1166` entries.
2. Create a DataLoader using this dataset.
3. Try iterating over the DataLoader.
code:
```python
def get_train_dataloader(self) -> DataLoader:
if self.train_dataset is None:
raise ValueError("Trainer: training requires a train_dataset.")
train_dataset = self.train_dataset
data_collator = self.data_collator
print(len(train_dataset))
print(train_dataset)
if is_datasets_available() and isinstance(train_dataset, datasets.Dataset):
train_dataset = self._remove_unused_columns(train_dataset, description="training")
else:
data_collator = self._get_collator_with_removed_columns(data_collator, description="training")
train_sampler = self._get_train_sampler()
dl = DataLoader(
train_dataset,
batch_size=self._train_batch_size,
sampler=train_sampler,
collate_fn=data_collator,
drop_last=self.args.dataloader_drop_last,
num_workers=self.args.dataloader_num_workers,
pin_memory=self.args.dataloader_pin_memory,
worker_init_fn=seed_worker,
)
print(dl)
print(len(dl))
for i in dl:
print(i)
break
return dl
```
output :
```
1166
Dataset({
features: ['input_ids', 'special_tokens_mask'],
num_rows: 1166
})
<torch.utils.data.dataloader.DataLoader object ...>
146
```
Error:
```
Traceback (most recent call last):
File "/home/dl/zym/llamaJP/TestUseContinuePretrainLlama.py", line 266, in <module>
train()
File "/home/dl/zym/llamaJP/TestUseContinuePretrainLlama.py", line 260, in train
trainer.train()
File "/root/miniconda3/envs/LLM/lib/python3.10/site-packages/transformers/trainer.py", line 1506, in train
return inner_training_loop(
File "/root/miniconda3/envs/LLM/lib/python3.10/site-packages/transformers/trainer.py", line 1520, in _inner_training_loop
train_dataloader = self.get_train_dataloader()
File "/home/dl/zym/llamaJP/TestUseContinuePretrainLlama.py", line 80, in get_train_dataloader
for i in dl:
File "/root/miniconda3/envs/LLM/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 630, in __next__
data = self._next_data()
File "/root/miniconda3/envs/LLM/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 674, in _next_data
data = self._dataset_fetcher.fetch(index) # may raise StopIteration
File "/root/miniconda3/envs/LLM/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py", line 49, in fetch
data = self.dataset.__getitems__(possibly_batched_index)
File "/root/miniconda3/envs/LLM/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 2807, in __getitems__
batch = self.__getitem__(keys)
File "/root/miniconda3/envs/LLM/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 2803, in __getitem__
return self._getitem(key)
File "/root/miniconda3/envs/LLM/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 2787, in _getitem
pa_subtable = query_table(self._data, key, indices=self._indices if self._indices is not None else None)
File "/root/miniconda3/envs/LLM/lib/python3.10/site-packages/datasets/formatting/formatting.py", line 583, in query_table
_check_valid_index_key(key, size)
File "/root/miniconda3/envs/LLM/lib/python3.10/site-packages/datasets/formatting/formatting.py", line 536, in _check_valid_index_key
_check_valid_index_key(int(max(key)), size=size)
File "/root/miniconda3/envs/LLM/lib/python3.10/site-packages/datasets/formatting/formatting.py", line 526, in _check_valid_index_key
raise IndexError(f"Invalid key: {key} is out of bounds for size {size}")
IndexError: Invalid key: 1116 is out of bounds for size 0
```
### Expected behavior
I expect to be able to iterate over the DataLoader without encountering an IndexError since the dataset is populated.
### Environment info
- `datasets` library version: [2.14.5]
- Platform: [Linux]
- Python version: 3.10
- Other libraries involved: HuggingFace Transformers | 6,294 |
https://github.com/huggingface/datasets/issues/6293 | Choose columns to stream parquet data in streaming mode | [] | Currently passing columns= to load_dataset in streaming mode fails
```
Tried to load parquet data with columns '['link']' with mismatching features '{'caption': Value(dtype='string', id=None), 'image': {'bytes': Value(dtype='binary', id=None), 'path': Value(dtype='null', id=None)}, 'link': Value(dtype='string', id=None), 'message_id': Value(dtype='string', id=None), 'timestamp': Value(dtype='string', id=None)}'
```
similar to https://github.com/huggingface/datasets/issues/6039
reported at https://huggingface.co/datasets/laion/dalle-3-dataset/discussions/3#65259a09617407d4520f4ad9 | 6,293 |
https://github.com/huggingface/datasets/issues/6292 | how to load the image of dtype float32 or float64 | [
"Hi! Can you provide a code that reproduces the issue?\r\n\r\nAlso, which version of `datasets` are you using? You can check this by running `python -c \"import datasets; print(datasets.__version__)\"` inside the env. We added support for \"float images\" in `datasets 2.9`."
] | _FEATURES = datasets.Features(
{
"image": datasets.Image(),
"text": datasets.Value("string"),
},
)
The datasets builder seems only support the unit8 data. How to load the float dtype data? | 6,292 |
https://github.com/huggingface/datasets/issues/6291 | Casting type from Array2D int to Array2D float crashes | [
"Thanks for reporting! I've opened a PR with a fix"
] | ### Describe the bug
I am on a school project and the initial type for feature annotations are `Array2D(shape=(None, 4))`. I am trying to cast this type to a `float64` and pyarrow gives me this error :
```
Traceback (most recent call last):
File "/home/alan/dev/ClassezDesImagesAvecDesAlgorithmesDeDeeplearning/src/sdd/data/dataset.py", line 141, in <module>
dataset = StanfordDogsDataset(size, 5).original(True).demo()
File "<attrs generated init __main__.StanfordDogsDataset>", line 4, in __init__
File "/home/alan/dev/ClassezDesImagesAvecDesAlgorithmesDeDeeplearning/src/sdd/data/dataset.py", line 33, in __attrs_post_init__
self.dataset = self.dataset.cast_column(
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/fingerprint.py", line 511, in wrapper
out = func(dataset, *args, **kwargs)
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 2110, in cast_column
return self.cast(features)
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 2055, in cast
dataset = dataset.map(
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 592, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 557, in wrapper
out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 3097, in map
for rank, done, content in Dataset._map_single(**dataset_kwargs):
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 3474, in _map_single
batch = apply_function_on_filtered_inputs(
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 3353, in apply_function_on_filtered_inputs
processed_inputs = function(*fn_args, *additional_args, **fn_kwargs)
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/table.py", line 2328, in table_cast
return cast_table_to_schema(table, schema)
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/table.py", line 2287, in cast_table_to_schema
arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/table.py", line 2287, in <listcomp>
arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/table.py", line 1831, in wrapper
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/table.py", line 1831, in <listcomp>
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/table.py", line 2143, in cast_array_to_feature
return array_cast(array, feature(), allow_number_to_str=allow_number_to_str)
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/table.py", line 1833, in wrapper
return func(array, *args, **kwargs)
File "/home/alan/.cache/pypoetry/virtualenvs/sdd-2XWLAjSi-py3.10/lib/python3.10/site-packages/datasets/table.py", line 1967, in array_cast
return pa_type.wrap_array(array)
File "pyarrow/types.pxi", line 1369, in pyarrow.lib.BaseExtensionType.wrap_array
TypeError: Incompatible storage type for extension<arrow.py_extension_type<Array2DExtensionType>>: expected list<item: list<item: double>>, got list<item: list<item: int32>>
```
### Steps to reproduce the bug
```python
dataset = datasets.load_dataset("Alanox/stanford-dogs", split="full")
dataset = dataset.cast_column("annotations", Array2D((None, 4), "float64"))
```
### Expected behavior
It should simply cast the column feature type to a `float64` without error
### Environment info
datasets == 2.14.5 | 6,291 |
https://github.com/huggingface/datasets/issues/6290 | Incremental dataset (e.g. `.push_to_hub(..., append=True)`) | [
"Yea I think waiting for #6269 would be best, or branching from it. For reference, this [PR](https://github.com/LAION-AI/Discord-Scrapers/pull/2) is progressing pretty well which will do similar using the hf hub for our LAION dataset bot https://github.com/LAION-AI/Discord-Scrapers/pull/2. ",
"Is there any update on this?"
] | ### Feature request
Have the possibility to do `ds.push_to_hub(..., append=True)`.
### Motivation
Requested in this [comment](https://huggingface.co/datasets/laion/dalle-3-dataset/discussions/3#65252597c4edc168202a5eaa) and
this [comment](https://huggingface.co/datasets/laion/dalle-3-dataset/discussions/4#6524f675c9607bdffb208d8f). Discussed internally on [slack](https://huggingface.slack.com/archives/C02EMARJ65P/p1696950642610639?thread_ts=1690554266.830949&cid=C02EMARJ65P).
### Your contribution
What I suggest to do for parquet datasets is to use `CommitOperationCopy` + `CommitOperationDelete` from `huggingface_hub`:
1. list files
2. copy files from parquet-0001-of-0004 to parquet-0001-of-0005
3. delete files like parquet-0001-of-0004
4. generate + add last parquet file parquet-0005-of-0005
=> make a single commit with all commit operations at once
I think it should be quite straightforward to implement. Happy to review a PR (maybe conflicting with the ongoing "1 commit push_to_hub" PR https://github.com/huggingface/datasets/pull/6269) | 6,290 |
https://github.com/huggingface/datasets/issues/6288 | Dataset.from_pandas with a DataFrame of PIL.Images | [
"A duplicate of https://github.com/huggingface/datasets/issues/4796.\r\n\r\nWe could get this for free by implementing the `Image` feature as an extension type, as shown in [this](https://colab.research.google.com/drive/1Uzm_tXVpGTwbzleDConWcNjacwO1yxE4?usp=sharing) Colab (example with UUIDs).\r\n",
"+1 to this\r\nCalling this line with a df that contains a PIL image (as they are returned from load_dataset)\r\n`ds = Dataset.from_pandas(df)`\r\nResults in this error:\r\n`ArrowInvalid: ('Could not convert <PIL.PngImagePlugin.PngImageFile image mode=RGB size=1024x1024 at 0x2B41F2D70> with type PngImageFile: did not recognize Python value type when inferring an Arrow data type', 'Conversion failed for column image with type object')`"
] | Currently type inference doesn't know what to do with a Pandas Series of PIL.Image objects, though it would be nice to get a Dataset with the Image type this way | 6,288 |
https://github.com/huggingface/datasets/issues/6287 | map() not recognizing "text" | [
"There is no \"text\" column in the `amazon_reviews_multi`, hence the `KeyError`. You can get the column names by running `dataset.column_names`."
] | ### Describe the bug
The [map() documentation](https://huggingface.co/docs/datasets/v2.14.5/en/package_reference/main_classes#datasets.Dataset.map) reads:
`
ds = ds.map(lambda x: tokenizer(x['text'], truncation=True, padding=True), batched=True)`
I have been trying to reproduce it in my code as:
`tokenizedDataset = dataset.map(lambda x: tokenizer(x['text']), batched=True)`
But it doesn't work as it throws the error:
> KeyError: 'text'
Can you please guide me on how to fix it?
### Steps to reproduce the bug
1. `from datasets import load_dataset
dataset = load_dataset("amazon_reviews_multi")`
2. Then this code: `from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("bert-base-cased")`
3. The line I quoted above (which I have been trying)
### Expected behavior
As mentioned in the documentation, it should run without any error and map the tokenization on the whole dataset.
### Environment info
Python 3.10.2 | 6,287 |
https://github.com/huggingface/datasets/issues/6285 | TypeError: expected str, bytes or os.PathLike object, not dict | [
"You should be able to load the images by modifying the `load_dataset` call like this:\r\n```python\r\ndataset = load_dataset(\"imagefolder\", data_dir=\"/content/datasets/PotholeDetectionYOLOv8-1\")\r\n```\r\n\r\nThe `imagefolder` builder expects the image files to be in `path/label/image_file` (e.g. .`.../train/dog/image_1.jpg`), so the solution for the labels in your case is to create metadata files (one for each split; as explained [here](https://huggingface.co/docs/datasets/image_dataset#imagefolder)) that map the images to their labels.",
"> You should be able to load the images by modifying the `load_dataset` call like this:\r\n> \r\n> ```python\r\n> dataset = load_dataset(\"imagefolder\", data_dir=\"/content/datasets/PotholeDetectionYOLOv8-1\")\r\n> ```\r\n> \r\n> The `imagefolder` builder expects the image files to be in `path/label/image_file` (e.g. .`.../train/dog/image_1.jpg`), so the solution for the labels in your case is to create metadata files (one for each split; as explained [here](https://huggingface.co/docs/datasets/image_dataset#imagefolder)) that map the images to their labels.\r\n\r\nI tried like this but only uploads images and not labels, Andyrasika/potholes-dataset",
"As explained in my previous comment, you need to define metadata files to load the labels or update the paths to be in the format `train/label/image` (`train- image /n -labels` is not supported by the loader).",
"I downloaded my file after annotating using roboflow . It gives train-\r\nimages, labels , test- images, labels , valid- images, labels . I hope it\r\ngives you an idea of the dataset . Please advise on this dataset\r\n\r\nOn Tue, Oct 10, 2023 at 18:12 Mario Šaško ***@***.***> wrote:\r\n\r\n> As explained in my previous comment, you need to define metadata files to\r\n> load the labels or update the paths to be in the format train/label/image\r\n> (train- image /n -labels is not supported by the loader).\r\n>\r\n> —\r\n> Reply to this email directly, view it on GitHub\r\n> <https://github.com/huggingface/datasets/issues/6285#issuecomment-1755335215>,\r\n> or unsubscribe\r\n> <https://github.com/notifications/unsubscribe-auth/AE4LJNN56FWWTSBYTSTUWHLX6U7CVAVCNFSM6AAAAAA5YHCSTGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONJVGMZTKMRRGU>\r\n> .\r\n> You are receiving this because you authored the thread.Message ID:\r\n> ***@***.***>\r\n>\r\n"
] | ### Describe the bug
my dataset is in form : train- image /n -labels
and tried the code:
```
from datasets import load_dataset
data_files = {
"train": "/content/datasets/PotholeDetectionYOLOv8-1/train/",
"validation": "/content/datasets/PotholeDetectionYOLOv8-1/valid/",
"test": "/content/datasets/PotholeDetectionYOLOv8-1/test/"
}
dataset = load_dataset("imagefolder", data_dir=data_files)
dataset
```
got error:
```
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
[<ipython-input-29-2ef1926f73d9>](https://localhost:8080/#) in <cell line: 8>()
6 "test": "/content/datasets/PotholeDetectionYOLOv8-1/test/"
7 }
----> 8 dataset = load_dataset("imagefolder", data_dir=data_files)
9 dataset
6 frames
[/usr/lib/python3.10/pathlib.py](https://localhost:8080/#) in _parse_args(cls, args)
576 parts += a._parts
577 else:
--> 578 a = os.fspath(a)
579 if isinstance(a, str):
580 # Force-cast str subclasses to str (issue #21127)
TypeError: expected str, bytes or os.PathLike object, not dict
```
### Steps to reproduce the bug
as share above
### Expected behavior
load images and labels , but my dataset only uploads images
- https://huggingface.co/datasets/Andyrasika/potholes-dataset
### Environment info
colab pro | 6,285 |
https://github.com/huggingface/datasets/issues/6284 | Add Belebele multiple-choice machine reading comprehension (MRC) dataset | [
"This dataset is already available on the Hub: https://huggingface.co/datasets/facebook/belebele.\r\n"
] | ### Feature request
Belebele is a multiple-choice machine reading comprehension (MRC) dataset spanning 122 language variants. This dataset enables the evaluation of mono- and multi-lingual models in high-, medium-, and low-resource languages. Each question has four multiple-choice answers and is linked to a short passage from the [FLORES-200](https://github.com/facebookresearch/flores/tree/main/flores200) dataset. The human annotation procedure was carefully curated to create questions that discriminate between different levels of generalizable language comprehension and is reinforced by extensive quality checks. While all questions directly relate to the passage, the English dataset on its own proves difficult enough to challenge state-of-the-art language models. Being fully parallel, this dataset enables direct comparison of model performance across all languages. Belebele opens up new avenues for evaluating and analyzing the multilingual abilities of language models and NLP systems.
Please refer to paper for more details, [The Belebele Benchmark: a Parallel Reading Comprehension Dataset in 122 Language Variants](https://arxiv.org/abs/2308.16884).
## Composition
- 900 questions per language variant
- 488 distinct passages, there are 1-2 associated questions for each.
- For each question, there is 4 multiple-choice answers, exactly 1 of which is correct.
- 122 language/language variants (including English).
- 900 x 122 = 109,800 total questions.
### Motivation
official repo https://github.com/facebookresearch/belebele
### Your contribution
- | 6,284 |
https://github.com/huggingface/datasets/issues/6280 | Couldn't cast array of type fixed_size_list to Sequence(Value(float64)) | [
"Thanks for reporting! I've opened a PR with a fix.",
"Thanks for the quick response @mariosasko! I just installed your branch via `poetry add 'git+https://github.com/huggingface/datasets#fix-array_values'` and I can confirm it works on the example provided.\r\n\r\nFollow up question for you, should `None`s be supported in these types of features as they are in others?\r\n\r\nFor example, the following script:\r\n\r\n```\r\nfrom datasets import Features, Value, Sequence, ClassLabel, Dataset\r\n\r\ndataset_features = Features({\r\n 'text': Value('string'),\r\n 'embedding': Sequence(Value('double'), length=2),\r\n 'categories': Sequence(ClassLabel(names=sorted([\r\n 'one',\r\n 'two',\r\n 'three'\r\n ]))),\r\n})\r\n\r\ndataset = Dataset.from_dict(\r\n {\r\n 'text': ['A'] * 10000,\r\n \"embedding\": [None] * 10000, # THIS LINE CHANGED\r\n 'categories': [[0]] * 10000,\r\n },\r\n features=dataset_features\r\n)\r\n\r\ndef test_mapper(r):\r\n r['text'] = list(map(lambda t: t + ' b', r['text']))\r\n return r\r\n\r\n\r\ndataset = dataset.map(test_mapper, batched=True, batch_size=10, features=dataset_features, num_proc=2)\r\n```\r\n\r\nfails with\r\n\r\n```\r\nTraceback (most recent call last):\r\n File \"/home/jmif/.virtualenvs/llm-training/lib/python3.10/site-packages/multiprocess/pool.py\", line 125, in worker\r\n result = (True, func(*args, **kwds))\r\n File \"/home/jmif/.virtualenvs/llm-training/lib/python3.10/site-packages/datasets/utils/py_utils.py\", line 1354, in _write_generator_to_queue\r\n for i, result in enumerate(func(**kwargs)):\r\n File \"/home/jmif/.virtualenvs/llm-training/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 3493, in _map_single\r\n writer.write_batch(batch)\r\n File \"/home/jmif/.virtualenvs/llm-training/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 549, in write_batch\r\n array = cast_array_to_feature(col_values, col_type) if col_type is not None else col_values\r\n File \"/home/jmif/.virtualenvs/llm-training/lib/python3.10/site-packages/datasets/table.py\", line 1831, in wrapper\r\n return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])\r\n File \"/home/jmif/.virtualenvs/llm-training/lib/python3.10/site-packages/datasets/table.py\", line 1831, in <listcomp>\r\n return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])\r\n File \"/home/jmif/.virtualenvs/llm-training/lib/python3.10/site-packages/datasets/table.py\", line 2160, in cast_array_to_feature\r\n raise TypeError(f\"Couldn't cast array of type\\n{array.type}\\nto\\n{feature}\")\r\nTypeError: Couldn't cast array of type\r\nfixed_size_list<item: double>[2]\r\nto\r\nSequence(feature=Value(dtype='float64', id=None), length=2, id=None)\r\n```\r\n\r\nIdeally we can have empty embedding columns as well!",
"This part of PyArrow is buggy and inconsistent regarding features implemented across the types, so the only option is to operate on the Arrow buffer level to fix issues such as the above one.",
"Ok - can you take the POC I did [here](https://github.com/huggingface/datasets/commit/15443098e9ce053943172f7ec6fce3769d7dff6e)? Happy to turn this into an actual PR but would appreciate feedback on the implementation before I take another pass!"
] | ### Describe the bug
I have a dataset with an embedding column, when I try to map that dataset I get the following exception:
```
Traceback (most recent call last):
File "/Users/jmif/.virtualenvs/llm-training/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 3189, in map
for rank, done, content in iflatmap_unordered(
File "/Users/jmif/.virtualenvs/llm-training/lib/python3.10/site-packages/datasets/utils/py_utils.py", line 1387, in iflatmap_unordered
[async_result.get(timeout=0.05) for async_result in async_results]
File "/Users/jmif/.virtualenvs/llm-training/lib/python3.10/site-packages/datasets/utils/py_utils.py", line 1387, in <listcomp>
[async_result.get(timeout=0.05) for async_result in async_results]
File "/Users/jmif/.virtualenvs/llm-training/lib/python3.10/site-packages/multiprocess/pool.py", line 774, in get
raise self._value
TypeError: Couldn't cast array of type
fixed_size_list<item: float>[2]
to
Sequence(feature=Value(dtype='float32', id=None), length=2, id=None)
```
### Steps to reproduce the bug
Here's a simple repro script:
```
from datasets import Features, Value, Sequence, ClassLabel, Dataset
dataset_features = Features({
'text': Value('string'),
'embedding': Sequence(Value('double'), length=2),
'categories': Sequence(ClassLabel(names=sorted([
'one',
'two',
'three'
]))),
})
dataset = Dataset.from_dict(
{
'text': ['A'] * 10000,
'embedding': [[0.0, 0.1]] * 10000,
'categories': [[0]] * 10000,
},
features=dataset_features
)
def test_mapper(r):
r['text'] = list(map(lambda t: t + ' b', r['text']))
return r
dataset = dataset.map(test_mapper, batched=True, batch_size=10, features=dataset_features, num_proc=2)
```
Removing the embedding column fixes the issue!
### Expected behavior
The mapping completes successfully.
### Environment info
- `datasets` version: 2.14.4
- Platform: macOS-14.0-arm64-arm-64bit
- Python version: 3.10.12
- Huggingface_hub version: 0.17.1
- PyArrow version: 13.0.0
- Pandas version: 2.0.3 | 6,280 |
https://github.com/huggingface/datasets/issues/6279 | Batched IterableDataset | [
"This is exactly what I was looking for. It would also be very useful for me :-)",
"This issue is really smashing the selling point of HF datasets... The only workaround I've found so far is to create a customized IterableDataloader which improves the loading speed to some extent.\r\n\r\nFor example I've a HF dataset `dt_train` with `len(dt_train) == 1M`. Using plain DataLoader is extremely slow:\r\n```\r\n%%time\r\ndl_train = DataLoader(dt_train, batch_size=128, shuffle = True)\r\nfor batch in dl_train:\r\n pass\r\n``` \r\n\r\n```\r\nCPU times: user 24min 35s, sys: 704 ms, total: 24min 36s\r\nWall time: 24min 37s\r\n```\r\nAnd DataLoader works even worse with HF's iterable_dataset:\r\n```\r\n%%time\r\ndt_train_ = dt_train.with_format(None).to_iterable_dataset(num_shards=64).shuffle(buffer_size=10_000)\r\ndl_train = DataLoader(dt_train_, batch_size=128)\r\nfor batch in dl_train:\r\n pass\r\n```\r\n```\r\nCPU times: user 1h 6min 2s, sys: 4.28 s, total: 1h 6min 6s\r\nWall time: 1h 7min 53s\r\n```\r\nWorkaround by running a customized wrapper:\r\n```\r\n%%time\r\nfrom torch.utils.data import DataLoader, IterableDataset\r\n\r\nclass Dataset2Iterable(IterableDataset):\r\n \"\"\"\r\n Wrapper to use a HF dataset as pytorch IterableDataset to speed up data loading.\r\n \"\"\"\r\n def __init__(self, dataset, batch_size=1, shuffle=True):\r\n super(Dataset2Iterable).__init__()\r\n self.dataset = dataset\r\n self.batch_size = batch_size\r\n self.shuffle = shuffle\r\n\r\n def __iter__(self):\r\n if self.shuffle: self.dataset.shuffle()\r\n return self.dataset.iter(batch_size=self.batch_size)\r\n\r\ndl_train = DataLoader(Dataset2Iterable(dt_train, batch_size = 128), batch_size=1, num_workers=0)\r\nfor n in range(2):\r\n for batch in dl_train:\r\n pass\r\n```\r\nThe speed still is slower than using tensorflow's loader but improved a lot than previous code:\r\n```\r\nCPU times: user 4min 18s, sys: 0 ns, total: 4min 18s\r\nWall time: 4min 20s\r\n```\r\nNote that the way I implemented `Dataset2Iterable` will only work with `num_workers=0`.",
"I can confirm that @zhh210's solution works with `num_workers=0`. However, for my use case, this was still slower than tokenizing on the fly through a collator and leveraging multiple workers in the dataloder.\r\n\r\n@lhoestq I think this is an important use case (e.g., streaming from a large dataset, online or stored on disk). What do you think might be the best solution to move forward?",
"I guess it can be implemented using a batched`.map()` under the hood that returns a single item containing the input batch.\r\n\r\nIn the meantime you can use this:\r\n\r\n```python\r\ndef batch(unbatched: dict[str, list]) -> dict[str, list]:\r\n return {k: [v] for k, v in unbatched}\r\n\r\nbatched_dataset = dataset.map(batch, batched=True, batch_size=batch_size)\r\n```\r\n\r\nThough it would be great to have a `.batch()` method indeed, I'd be happy to help with anyone wants to open a PR",
"If no one else is planning to work on this, I can take it on. I'll wait until next week, and if no one has started a PR by then, I'll go ahead and open one."
] | ### Feature request
Hi,
could you add an implementation of a batched `IterableDataset`. It already support an option to do batch iteration via `.iter(batch_size=...)` but this cannot be used in combination with a torch `DataLoader` since it just returns an iterator.
### Motivation
The current implementation loads each element of a batch individually which can be very slow in cases of a big batch_size. I did some experiments [here](https://discuss.huggingface.co/t/slow-dataloader-with-big-batch-size/57224) and using a batched iteration would speed up data loading significantly.
### Your contribution
N/A | 6,279 |
https://github.com/huggingface/datasets/issues/6277 | FileNotFoundError: Couldn't find a module script at /content/paws-x/paws-x.py. Module 'paws-x' doesn't exist on the Hugging Face Hub either. | [
"`evaluate.load(\"paws-x\", \"es\")` throws the error because there is no such metric in the `evaluate` lib.\r\n\r\nSo, this is unrelated to our lib."
] | ### Describe the bug
I'm encountering a "FileNotFoundError" while attempting to use the "paws-x" dataset to retrain the DistilRoBERTa-base model. The error message is as follows:
FileNotFoundError: Couldn't find a module script at /content/paws-x/paws-x.py. Module 'paws-x' doesn't exist on the Hugging Face Hub either.
### Steps to reproduce the bug
https://colab.research.google.com/drive/11xUUFxloClpmqLvDy_Xxfmo3oUzjY5nx#scrollTo=kUn74FigzhHm
### Expected behavior
The the trained model
### Environment info
colab, "paws-x" dataset , DistilRoBERTa-base model | 6,277 |
https://github.com/huggingface/datasets/issues/6276 | I'm trying to fine tune the openai/whisper model from huggingface using jupyter notebook and i keep getting this error | [
"Since you are using Windows, maybe moving the `map` call inside `if __name__ == \"__main__\"` can fix the issue:\r\n```python\r\nif __name__ == \"__main__\":\r\n common_voice = common_voice.map(prepare_dataset, remove_columns=common_voice.column_names[\"train\"], num_proc=4)\r\n```\r\n\r\nOtherwise, the only solution is to set `num_proc=1`.",
"> Since you are using Windows, maybe moving the `map` call inside `if __name__ == \"__main__\"` can fix the issue:\r\n> \r\n> ```python\r\n> if __name__ == \"__main__\":\r\n> common_voice = common_voice.map(prepare_dataset, remove_columns=common_voice.column_names[\"train\"], num_proc=4)\r\n> ```\r\n> \r\n> Otherwise, the only solution is to set `num_proc=1`.\r\n\r\nThank you very much for the response, i eventually tried setting `num_proc=1` and now the jupyter notebook kernel keers dying after running the command, what do you think the issue could be, could it be that my system is not capable of running the command \"i'm using a Lenovo Thinkpad T440 with no GPU\"",
"Firstly, you didn't define feature_extractor variable. Secondly, it is large nlp model. Hence you should use proper gpu, otherwise your machine's cpu will be overclock and you can do nothing."
] | ### Describe the bug
I'm trying to fine tune the openai/whisper model from huggingface using jupyter notebook and i keep getting this error, i'm following the steps in this blog post
https://huggingface.co/blog/fine-tune-whisper
I tried google collab and it works but because I'm on the free version the training doesn't complete
the error comes in jupyter notebook when i run this line
`common_voice = common_voice.map(prepare_dataset, remove_columns=common_voice.column_names["train"], num_proc=4)`
here is the error message
```
Map (num_proc=4): 0% 0/2506 [00:52<?, ? examples/s]
The above exception was the direct cause of the following exception:
NameError Traceback (most recent call last) Cell In[19], line 1 ----> 1 common_voice = common_voice.map(prepare_dataset, remove_columns=common_voice.column_names["train"], num_proc=4)
File ~\anaconda\Lib\site-packages\datasets\dataset_dict.py:853, in DatasetDict.map(self, function, with_indices, with_rank, input_columns, batched, batch_size, drop_last_batch, remove_columns, keep_in_memory, load_from_cache_file, cache_file_names, writer_batch_size, features, disable_nullable, fn_kwargs, num_proc, desc) 850 if cache_file_names is None: 851 cache_file_names = {k: None for k in self} 852 return DatasetDict( --> 853 { 854 k: dataset.map( 855 function=function, 856 with_indices=with_indices, 857 with_rank=with_rank, 858 input_columns=input_columns, 859 batched=batched, 860 batch_size=batch_size, 861 drop_last_batch=drop_last_batch, 862 remove_columns=remove_columns, 863 keep_in_memory=keep_in_memory, 864 load_from_cache_file=load_from_cache_file, 865 cache_file_name=cache_file_names[k], 866 writer_batch_size=writer_batch_size, 867 features=features, 868 disable_nullable=disable_nullable, 869 fn_kwargs=fn_kwargs, 870 num_proc=num_proc, 871 desc=desc, 872 ) 873 for k, dataset in self.items() 874 } 875 )
File ~\anaconda\Lib\site-packages\datasets\dataset_dict.py:854, in <dictcomp>(.0) 850 if cache_file_names is None: 851 cache_file_names = {k: None for k in self} 852 return DatasetDict( 853 { --> 854 k: dataset.map( 855 function=function, 856 with_indices=with_indices, 857 with_rank=with_rank, 858 input_columns=input_columns, 859 batched=batched, 860 batch_size=batch_size, 861 drop_last_batch=drop_last_batch, 862 remove_columns=remove_columns, 863 keep_in_memory=keep_in_memory, 864 load_from_cache_file=load_from_cache_file, 865 cache_file_name=cache_file_names[k], 866 writer_batch_size=writer_batch_size, 867 features=features, 868 disable_nullable=disable_nullable, 869 fn_kwargs=fn_kwargs, 870 num_proc=num_proc, 871 desc=desc, 872 ) 873 for k, dataset in self.items() 874 } 875 )
File ~\anaconda\Lib\site-packages\datasets\arrow_dataset.py:592, in transmit_tasks.<locals>.wrapper(*args, **kwargs) 590 self: "Dataset" = kwargs.pop("self") 591 # apply actual function --> 592 out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs) 593 datasets: List["Dataset"] = list(out.values()) if isinstance(out, dict) else [out] 594 for dataset in datasets: 595 # Remove task templates if a column mapping of the template is no longer valid
File ~\anaconda\Lib\site-packages\datasets\arrow_dataset.py:557, in transmit_format.<locals>.wrapper(*args, **kwargs) 550 self_format = { 551 "type": self._format_type, 552 "format_kwargs": self._format_kwargs, 553 "columns": self._format_columns, 554 "output_all_columns": self._output_all_columns, 555 } 556 # apply actual function --> 557 out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs) 558 datasets: List["Dataset"] = list(out.values()) if isinstance(out, dict) else [out] 559 # re-apply format to the output
File ~\anaconda\Lib\site-packages\datasets\arrow_dataset.py:3189, in Dataset.map(self, function, with_indices, with_rank, input_columns, batched, batch_size, drop_last_batch, remove_columns, keep_in_memory, load_from_cache_file, cache_file_name, writer_batch_size, features, disable_nullable, fn_kwargs, num_proc, suffix_template, new_fingerprint, desc) 3182 logger.info(f"Spawning {num_proc} processes") 3183 with logging.tqdm( 3184 disable=not logging.is_progress_bar_enabled(), 3185 unit=" examples", 3186 total=pbar_total, 3187 desc=(desc or "Map") + f" (num_proc={num_proc})", 3188 ) as pbar: -> 3189 for rank, done, content in iflatmap_unordered( 3190 pool, Dataset._map_single, kwargs_iterable=kwargs_per_job 3191 ): 3192 if done: 3193 shards_done += 1
File ~\anaconda\Lib\site-packages\datasets\utils\py_utils.py:1394, in iflatmap_unordered(pool, func, kwargs_iterable) 1391 finally: 1392 if not pool_changed: 1393 # we get the result in case there's an error to raise -> 1394 [async_result.get(timeout=0.05) for async_result in async_results]
File ~\anaconda\Lib\site-packages\datasets\utils\py_utils.py:1394, in <listcomp>(.0) 1391 finally: 1392 if not pool_changed: 1393 # we get the result in case there's an error to raise -> 1394 [async_result.get(timeout=0.05) for async_result in async_results]
File ~\anaconda\Lib\site-packages\multiprocess\pool.py:774, in ApplyResult.get(self, timeout) 772 return self._value 773 else: --> 774 raise self._value
NameError: name 'feature_extractor' is not defined
```
### Steps to reproduce the bug
1. follow the steps in this blog post
https://huggingface.co/blog/fine-tune-whisper
2. run this line of code
`common_voice = common_voice.map(prepare_dataset, remove_columns=common_voice.column_names["train"], num_proc=4)`
3. I'm using jupyter notebook from anaconda
### Expected behavior
No error message
### Environment info
datasets version: 2.8.0
Python version: 3.11
Windows 10 | 6,276 |
https://github.com/huggingface/datasets/issues/6275 | Would like to Contribute a dataset | [
"Hi! The process of contributing a dataset is explained here: https://huggingface.co/docs/datasets/upload_dataset. Also, check https://huggingface.co/docs/datasets/image_dataset for a more detailed explanation of how to share an image dataset."
] | I have a dataset of 2500 images that can be used for color-blind machine-learning algorithms. Since , there was no dataset available online , I made this dataset myself and would like to contribute this now to community | 6,275 |
https://github.com/huggingface/datasets/issues/6274 | FileNotFoundError for dataset with multiple builder config | [
"Please tell me if the above info is not enough for solving the problem. I will then make my dataset public temporarily so that you can really reproduce the bug. "
] | ### Describe the bug
When there is only one config and only the dataset name is entered when using datasets.load_dataset(), it works fine. But if I create a second builder_config for my dataset and enter the config name when using datasets.load_dataset(), the following error will happen.
FileNotFoundError: [Errno 2] No such file or directory: 'C:/Users/chenx/.cache/huggingface/datasets/my_dataset/0_shot_multiple_choice/1.0.0/97c3854a012cfd6b045e3be4c864739902af2d818bb9235b047baa94c302e9a2.incomplete/my_dataset-test-00000-00000-of-NNNNN.arrow'
The "XXX.incomplete folder" in the cache folder of my dataset will disappear before "generating test split", which does not happen when config name is not entered and the config name is "default"
C:\Users\chenx\.cache\huggingface\datasets\my_dataset\0_shot_multiple_choice\1.0.0
The folder that is supposed to remain under the above directory will disappear, and the data generator will not have a place to generate data into.
### Steps to reproduce the bug
test = load_dataset('my_dataset', '0_shot_multiple_choice')
### Expected behavior
FileNotFoundError: [Errno 2] No such file or directory: 'C:/Users/chenx/.cache/huggingface/datasets/my_dataset/0_shot_multiple_choice/1.0.0/97c3854a012cfd6b045e3be4c864739902af2d818bb9235b047baa94c302e9a2.incomplete/my_dataset-test-00000-00000-of-NNNNN.arrow'
### Environment info
datasets 2.14.5
python 3.8.18 | 6,274 |
https://github.com/huggingface/datasets/issues/6273 | Broken Link to PubMed Abstracts dataset . | [
"This has already been reported in the HF Course repo (https://github.com/huggingface/course/issues/623).",
"@lhoestq @albertvillanova @lewtun I don't think we are allowed to host these data files on the Hub (due to DMCA), which means the only option is to use a different dataset in the course (and to re-record the video 🙂), no?",
"Keeping the video is maybe fine, we can add a note on youtube to suggest to load a dataset with a different name. Maybe C4 ? And update the code snippets on the website ?",
"Maybe you want to try it with the PUBMED dataset that I reproduced based on the The [PubMed Abstract GitHub Site](http://github.com/thoppe/The-Pile-PubMed) and uploaded on the HuggingFace:\r\n\r\n```\r\nfrom datasets import load_dataset\r\npubmed_dataset = load_dataset(\"hwang2006/PUBMED_title_abstracts_2020_baseline\")\r\npubmed_dataset\r\n\r\n#Downloading data: 100%\r\n#7.98G/7.98G [11:47<00:00, 9.68MB/s]\r\n#Generating train split: 17722096/0 [00:36<00:00, 505376.37 examples/s]\r\n\r\n#DatasetDict({\r\n# train: Dataset({\r\n# features: ['meta', 'text'],\r\n# num_rows: 17722096\r\n# })\r\n#})\r\n```",
"孔令涛说感谢感谢"
] | ### Describe the bug
The link provided for the dataset is broken,
data_files =
[https://the-eye.eu/public/AI/pile_preliminary_components/PUBMED_title_abstracts_2019_baseline.jsonl.zst](url)
The
### Steps to reproduce the bug
Steps to reproduce:
1) Head over to [https://huggingface.co/learn/nlp-course/chapter5/4?fw=pt#big-data-datasets-to-the-rescue](url)
2) In the Section "What is the Pile?", you can see a code snippet that contains the broken link.
### Expected behavior
The link should Redirect to the "PubMed Abstracts dataset" as expected .
### Environment info
. | 6,273 |
https://github.com/huggingface/datasets/issues/6272 | Duplicate `data_files` when named `<split>/<split>.parquet` | [
"Also reported in https://github.com/huggingface/datasets/issues/6259",
"I think it's best to drop duplicates with a `set` (as a temporary fix) and improve the patterns when/if https://github.com/fsspec/filesystem_spec/pull/1382 gets merged. @lhoestq Do you have some other ideas?",
"Alternatively we could just use this no ?\r\n\r\n```python\r\nif config.FSSPEC_VERSION < version.parse(\"2023.9.0\"):\r\n KEYWORDS_IN_PATH_NAME_BASE_PATTERNS = [\r\n \"{keyword}[{sep}/]**\",\r\n \"**[{sep}]{keyword}[{sep}/]**\",\r\n \"**/{keyword}[{sep}/]**\",\r\n ]\r\nelse:\r\n KEYWORDS_IN_PATH_NAME_BASE_PATTERNS = [\r\n \"{keyword}[{sep}/]**\",\r\n \"**/*[{sep}]{keyword}[{sep}/]**\",\r\n \"**/*/{keyword}[{sep}/]**\",\r\n ]\r\n```\r\n\r\nThis way no need to implement sets, which would require a bit of work since we've always considered a list of pattern to be resolved as the concatenated list of resolved files for each pattern (including duplicates)\r\n",
"Arf `\"**/*/{keyword}[{sep}/]**\"` does return `data/keyword.txt` in latest `fsspec` but not in `glob.glob`\r\n\r\nEDIT: actually forgot to set `recursive=True`",
"Actually `glob.glob` does return it with `recursive=True` ! my bad",
"Pff just tested and my idea sucks, pattern 1 and 3 obviously give duplicates ",
"> I think it's best to drop duplicates with a set (as a temporary fix)\r\n\r\nI started https://github.com/huggingface/datasets/pull/6278 to use DataFilesSet objects instead of DataFilesList"
] | e.g. with `u23429/stock_1_minute_ticker`
```ipython
In [1]: from datasets import *
In [2]: b = load_dataset_builder("u23429/stock_1_minute_ticker")
Downloading readme: 100%|██████████████████████████| 627/627 [00:00<00:00, 246kB/s]
In [3]: b.config.data_files
Out[3]:
{NamedSplit('train'): ['hf://datasets/u23429/stock_1_minute_ticker@65c973cf4ec061f01a363b40da4c1bb128ba4166/train/train.parquet',
'hf://datasets/u23429/stock_1_minute_ticker@65c973cf4ec061f01a363b40da4c1bb128ba4166/train/train.parquet'],
NamedSplit('validation'): ['hf://datasets/u23429/stock_1_minute_ticker@65c973cf4ec061f01a363b40da4c1bb128ba4166/validation/validation.parquet',
'hf://datasets/u23429/stock_1_minute_ticker@65c973cf4ec061f01a363b40da4c1bb128ba4166/validation/validation.parquet'],
NamedSplit('test'): ['hf://datasets/u23429/stock_1_minute_ticker@65c973cf4ec061f01a363b40da4c1bb128ba4166/test/test.parquet',
'hf://datasets/u23429/stock_1_minute_ticker@65c973cf4ec061f01a363b40da4c1bb128ba4166/test/test.parquet']}
```
This bug issue is present in the current `datasets` 2.14.5 and also on `main` even after https://github.com/huggingface/datasets/pull/6244 cc @mariosasko | 6,272 |
https://github.com/huggingface/datasets/issues/6271 | Overwriting Split overwrites data but not metadata, corrupting dataset | [] | ### Describe the bug
I want to be able to overwrite/update/delete splits in my dataset. Currently the only way to do is to manually go into the dataset and delete the split. If I try to overwrite programmatically I end up in an error state and (somewhat) corrupting the dataset. Read below.
**Current Behavior**
When I push to an existing split I get this error:
`ValueError: Split complexRoofLocation_01Apr2023_to_31May2023test already present`
This seems to suggest that the library doesn't support overwriting splits.
**Potential Bug**
What’s strange is that datasets, despite the operation erroring out with the ValueError above, does, in fact, overwrite the split:
`Pushing dataset shards to the dataset hub: 100% [.....................] 1/1 [00:00<00:00, 55.04it/s]`
Even though you got an error message and your code fails, your dataset is now changed. That seems like a bug. Either don't change the dataset, or don't throw the error and allow the script to proceed.
Additional Bug
While it overwrites the split, it doesn’t overwrite the split’s information. Because of this when you pull down the dataset you may end up getting a `NonMatchingSplitsSizesError` if the size of the dataset during the overwrite is different. For example, my original split had 5 rows, but on my overwrite, I only had 4. Then when I try to download the dataset, I get a `NonMatchingSplitsSizesError` because the dataset's data.json states there’s 5 but only 4 exist in the split.
Expected Behavior
This corrupts the dataset rendering it unusable (until you take manual intervention). Either the library should let the overwrite happen (which it does but should also update the metadata) or it shouldn’t do anything.
### Steps to reproduce the bug
[Colab Notebook](https://colab.research.google.com/drive/1bqVkD06Ngs9MQNdSk_ygCG6y1UqXA4pC?usp=sharing)
### Expected behavior
The split should be overwritten and I should be able to use the new version of the dataset without issue.
### Environment info
- `datasets` version: 2.14.5
- Platform: Linux-5.15.120+-x86_64-with-glibc2.35
- Python version: 3.10.12
- Huggingface_hub version: 0.17.3
- PyArrow version: 9.0.0
- Pandas version: 1.5.3
| 6,271 |
https://github.com/huggingface/datasets/issues/6270 | Dataset.from_generator raises with sharded gen_args | [
"`gen_kwargs` should be a `dict`, as stated in the docstring, but you are passing a `list`.\r\n\r\nSo, to fix the error, replace the list of dicts with a dict of lists (and slightly modify the generator function):\r\n```python\r\nfrom pathlib import Path\r\nimport datasets\r\n\r\ndef process_yaml(files):\r\n for f in files:\r\n # process\r\n yield dict(...)\r\n\r\n\r\nif __name__ == '__main__':\r\n import sys\r\n dir = Path(sys.argv[0]).parent\r\n ds = datasets.Dataset.from_generator(process_yaml, gen_kwargs={'files': [f for f in dir.glob('*.yml')]})\r\n ds.to_json('training.jsonl')\r\n```",
"That runs, and because my dataset is small, it's what I did to get past the problem.\r\nHowever, it does not produce a sharded dataset. From the doc string I expect there ought to be a way to call from_generator such that num_shards in the resulting data set is equal to the number of items in the list.\r\nThe part of the doc string that your suggestion is not responsive to is:\r\n` You can define a sharded dataset by passing the list of shards in *g\r\nen_kwargs*.\r\n`\r\n\r\nWhat your suggestion does is calls the generator once, with the list argument, and produces a single shard dataset.\r\n",
"The sharding mentioned here refers to using this function with `num_proc` (multiprocessing splits the `kwargs` into shards and passes them to the generator function)\r\n\r\n> That runs, and because my dataset is small, it's what I did to get past the problem.\r\n\r\n`from_generator` generates a memory-mapped dataset (can be larger than RAM), so the dataset size should not be an issue unless the generator function's implementation does not properly free the memory.\r\n",
"It sounds like you are saying that num_proc affects the form of gen_kwargs.\r\nAre you saying that for non-zero num_proc gen_kwargs should be a list whose length is the same as num_proc?\r\nOr are you saying that for non-zero num_proc, gen_kwargs should be a dict whose elements are lists the length of num_proc?\r\n",
"I ran some tests. So, it looks like with num_proc greater than 1, gen_kwargs is expected to be a dict of lists. It calls the generator also with a dict of lists, but the lists are split.\r\nI.E. if my original has `gen_kwargs=dict(a=[0,1,2])`, then my generator might get called with `gen_kwalrgs=dict([0])`.\r\nThat all makes sense, but I definitely think there is room for improvement in the doc string here.\r\nIn order to suggest improvements to the doc string, I need to look at how the gen_kwargs are split, and figure out if:\r\n* num_proc needs to exactly equal the length of the lists\r\n* num_proc needs to evenly divide the length of the lists\r\n* Or there's no required relationship.\r\nI'll look into that and then propose an improved doc string if no one else gets to it first.",
"Okay, that was fun; I took a dive through the dataset code and feel like I have a much better understanding.\r\nHere is my understanding of the behavior:\r\n* max_proc is an upper limit on the number of shards that `from_generator` produces\r\n* If `max_proc` is greater than 1, then all lists in *gen_kwargs* must be the same length\r\n* If the lists in *gen_kwargs* are shorter than *num_proc* elements, *num_proc* will be reduced and a warning produced. Put another way, `min(list_length, num_shards)` shards will be produced\r\n* The members of the lists in *gen_kwargs* will be partitioned among the created jobs.\r\nTo validate the above, take a look at\r\n`_number_of_shards_in_gen_kwargs` and `_distribute_shards` and `_split_gen_kwargs` in utils/sharding.py.\r\nI've also chased down starting at *from_generator* all the way through to GeneratorBuilder and the calls to the functions in sharding.py.\r\nTomorrow I'll take a look at the contributing guidelines and see what's involved in putting together a PR to improve the doc string."
] | ### Describe the bug
According to the docs of Datasets.from_generator:
```
gen_kwargs(`dict`, *optional*):
Keyword arguments to be passed to the `generator` callable.
You can define a sharded dataset by passing the list of shards in `gen_kwargs`.
```
So I'd expect that if gen_kwargs was a list, then my generator would be called once for each element in the list with the dict in the list for that element.
It doesn't work that way though.
### Steps to reproduce the bug
```python
#!/usr/bin/python
from pathlib import Path
import datasets
def process_yaml(file):
yield dict(example=42)
if __name__ == '__main__':
import sys
dir = Path(sys.argv[0]).parent
ds = datasets.Dataset.from_generator(process_yaml, gen_kwargs=[{'file':f} for f in dir.glob('*.yml')],
)
ds.to_json('training.jsonl')
```
```
Generating train split: 0 examples [00:00, ? examples/s]
Traceback (most recent call last):
File "/tmp/dataset_bug.py", line 13, in <module>
ds = datasets.Dataset.from_generator(process_yaml, gen_kwargs=[{'file':f} for f in dir.glob('*.yml')],
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/hartmans/ai/venv/lib/python3.11/site-packages/datasets/arrow_dataset.py", line 1072, in from_generator
).read()
^^^^^^
File "/home/hartmans/ai/venv/lib/python3.11/site-packages/datasets/io/generator.py", line 47, in read
self.builder.download_and_prepare(
File "/home/hartmans/ai/venv/lib/python3.11/site-packages/datasets/builder.py", line 954, in download_and_prepare
self._download_and_prepare(
File "/home/hartmans/ai/venv/lib/python3.11/site-packages/datasets/builder.py", line 1717, in _download_and_prepare
super()._download_and_prepare(
File "/home/hartmans/ai/venv/lib/python3.11/site-packages/datasets/builder.py", line 1049, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/home/hartmans/ai/venv/lib/python3.11/site-packages/datasets/builder.py", line 1555, in _prepare_split
for job_id, done, content in self._prepare_split_single(
File "/home/hartmans/ai/venv/lib/python3.11/site-packages/datasets/builder.py", line 1656, in _prepare_split_single
generator = self._generate_examples(**gen_kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: datasets.packaged_modules.generator.generator.Generator._generate_examples() argument after ** must be a ```
mapping, not list
### Expected behavior
I would expect that process_yaml would be called once for each yaml file in the directory where the script is run.
I also tried with the list being in gen_kwargs, but in that case process_yaml gets called with a list.
### Environment info
- `datasets` version: 2.14.6.dev0 (git commit 0cc77d7f45c7369; also tested with 2.14.0)
- Platform: Linux-6.1.0-10-amd64-x86_64-with-glibc2.36
- Python version: 3.11.2
- Huggingface_hub version: 0.16.4
- PyArrow version: 12.0.1
- Pandas version: 2.0.3
| 6,270 |
https://github.com/huggingface/datasets/issues/6267 | Multi label class encoding | [
"You can use a `Sequence(ClassLabel(...))` feature type to represent a list of labels, and `cast_column`/`cast` to perform the \"string to label\" conversion (`class_encode_column` does support nested fields), e.g., in your case:\r\n```python\r\nfrom datasets import Dataset, Sequence, ClassLabel\r\ndata = {\r\n 'text': ['one', 'two', 'three', 'four'],\r\n 'labels': [['a', 'b'], ['b'], ['b', 'c'], ['a', 'd']]\r\n}\r\n\r\ndataset = Dataset.from_dict(data)\r\ndataset = dataset.cast_column('labels', Sequence(ClassLabel(names=[\"a\", \"b\", \"c\", \"d\"])))\r\n```",
"Great! Can you elaborate on \"class_encode_column does support nested fields\"? Do you mean that there is a way to `class_encode_column` on a Sequence?",
"Yes, exactly! This would be a nice contribution, though.",
"Sorry, I'm still not following. Are you saying that there currently exists a way to call `class_encode_column` on a `Sequence(ClassLabel)` type? Or that the underlying data structures support it and a contribution of a method to do that would be welcome?",
"`class_encode_column ` currently does not support `Sequence(ClassLabel)`. Implementing support for this would be a nice contribution.\r\n\r\nIn the meantime, this limitation can be circumvented by fetching (unique) labels and calling `.cast_column(col, Sequence(ClassLabel(names=labels)))`.",
"Ok makes sense, can you take a look at the POC implementation I did [here](https://github.com/huggingface/datasets/commit/15443098e9ce053943172f7ec6fce3769d7dff6e)? Happy to take another pass / submit as a PR but would be helpful if I got a thumbs up that this was directionally correct with respect to implementation / architecture. ",
"There is no need to introduce a new type (`MultiLabel`) for this feature. Also, I think we can keep the logic inside a single method instead of separating the two cases.\r\n\r\nMaybe https://github.com/huggingface/datasets/pull/4277 can help with the implementation. We extended `align_labels_with_mapping` to support `Sequence(ClassLabel(...))` in that PR (initially, it only worked with `ClassLabel(...)`)"
] | ### Feature request
I have a multi label dataset and I'd like to be able to class encode the column and store the mapping directly in the features just as I can with a single label column. `class_encode_column` currently does not support multi labels.
Here's an example of what I'd like to encode:
```
data = {
'text': ['one', 'two', 'three', 'four'],
'labels': [['a', 'b'], ['b'], ['b', 'c'], ['a', 'd']]
}
dataset = Dataset.from_dict(data)
dataset = dataset.class_encode_column('labels')
```
I did some digging into the code base to evaluate the feasibility of this (note I'm very new to this code base) and from what I noticed the `ClassLabel` feature is still stored as an underlying raw data type of int so I thought a `MultiLabel` feature could similarly be stored as a Sequence of ints, thus not requiring significant serialization / conversion work to / from arrow.
I did a POC of this [here](https://github.com/huggingface/datasets/commit/15443098e9ce053943172f7ec6fce3769d7dff6e) and included a simple test case (please excuse all the commented out tests, going for speed of POC here and didn't want to fight IDE to debug a single test). In the test I just assert that `num_classes` is the same to show that things are properly serializing, but if you break after loading from disk you'll see the dataset correct and the dataset feature is as expected.
After digging more I did notice a few issues
- After loading from disk I noticed type of the `labels` class is `Sequence` not `MultiLabel` (though the added `feature` attribute came through). This doesn't happen for `ClassLabel` but I couldn't find the encode / decode code paths that handle this.
- I subclass `Sequence` in `MultiLabel` to leverage existing serialization, but this does miss the custom encode logic that `ClassLabel` has. I'm not sure of the best way to approach this as I haven't fully understood the encode / decode flow for datasets. I suspect my simple implementation will need some improvement as it'll require a significant amount of repeated logic to mimic `ClassLabel` behavior.
### Motivation
See above - would like to support multi label class encodings.
### Your contribution
This would be a big help for us and we're open to contributing but I'll likely need some guidance on how to implement to fit the encode / decode flow. Some suggestions on tests / would be great too, I'm guessing in addition to the class encode tests (that I'll need to expand) we'll need encode / decode tests. | 6,267 |
https://github.com/huggingface/datasets/issues/6263 | CI is broken: ImportError: cannot import name 'context' from 'tensorflow.python' | [] | Python 3.10 CI is broken for `test_py310`.
See: https://github.com/huggingface/datasets/actions/runs/6322990957/job/17169678812?pr=6262
```
FAILED tests/test_py_utils.py::TempSeedTest::test_tensorflow - ImportError: cannot import name 'context' from 'tensorflow.python' (/opt/hostedtoolcache/Python/3.10.13/x64/lib/python3.10/site-packages/tensorflow/python/__init__.py)
```
```
_________________________ TempSeedTest.test_tensorflow _________________________
[gw1] linux -- Python 3.10.13 /opt/hostedtoolcache/Python/3.10.13/x64/bin/python
self = <tests.test_py_utils.TempSeedTest testMethod=test_tensorflow>
@require_tf
def test_tensorflow(self):
import tensorflow as tf
from tensorflow.keras import layers
model = layers.Dense(2)
def gen_random_output():
x = tf.random.uniform((1, 3))
return model(x).numpy()
> with temp_seed(42, set_tensorflow=True):
tests/test_py_utils.py:155:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
/opt/hostedtoolcache/Python/3.10.13/x64/lib/python3.10/contextlib.py:135: in __enter__
return next(self.gen)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
seed = 42, set_pytorch = False, set_tensorflow = True
@contextmanager
def temp_seed(seed: int, set_pytorch=False, set_tensorflow=False):
"""Temporarily set the random seed. This works for python numpy, pytorch and tensorflow."""
np_state = np.random.get_state()
np.random.seed(seed)
if set_pytorch and config.TORCH_AVAILABLE:
import torch
torch_state = torch.random.get_rng_state()
torch.random.manual_seed(seed)
if torch.cuda.is_available():
torch_cuda_states = torch.cuda.get_rng_state_all()
torch.cuda.manual_seed_all(seed)
if set_tensorflow and config.TF_AVAILABLE:
import tensorflow as tf
> from tensorflow.python import context as tfpycontext
E ImportError: cannot import name 'context' from 'tensorflow.python' (/opt/hostedtoolcache/Python/3.10.13/x64/lib/python3.10/site-packages/tensorflow/python/__init__.py)
/opt/hostedtoolcache/Python/3.10.13/x64/lib/python3.10/site-packages/datasets/utils/py_utils.py:257: ImportError
``` | 6,263 |
https://github.com/huggingface/datasets/issues/6261 | Can't load a dataset | [
"I believe is due to the fact that doesn't work with .tgz files.",
"`JourneyDB/JourneyDB` is a gated dataset, so this error means you are not authenticated to access it, either by using an invalid token or by not agreeing to the terms in the dialog on the dataset page.\r\n\r\n> I believe is due to the fact that doesn't work with .tgz files.\r\n\r\nIndeed, the dataset's data files structure is not supported natively by `datasets`. To load it, one option is to clone the repo (or download it with `huggingface_hub.snapshot_download`) and use `Dataset.from_generator` to process the files.",
"> JourneyDB/JourneyDB is a gated dataset, so this error means you are not authenticated to access it, either by using an invalid token or by not agreeing to the terms in the dialog on the dataset page.´\r\n\r\nI did authentication with:\r\n\r\n```\r\nfrom huggingface_hub import notebook_login\r\nnotebook_login()\r\n```\r\n\r\nIsn't that the correct way to do it?\r\n\r\n> Indeed, the dataset's data files structure is not supported natively by datasets. To load it, one option is to clone the repo (or download it with huggingface_hub.snapshot_download) and use Dataset.from_generator to process the files.\r\n\r\nGreat suggestion I will give it a try.",
"Have you accepted the terms in the dialog [here](https://huggingface.co/datasets/JourneyDB/JourneyDB)?\r\n\r\nIIRC Kaggle preinstalls an outdated `datasets` version, so it's also a good idea to update it before importing `datasets` (and do the same for `huggingface_hub`)",
"Sorry for the late reply. Yes, I did. Thanks for the tip!"
] | ### Describe the bug
Can't seem to load the JourneyDB dataset.
It throws the following error:
```
---------------------------------------------------------------------------
FileNotFoundError Traceback (most recent call last)
Cell In[15], line 2
1 # If the dataset is gated/private, make sure you have run huggingface-cli login
----> 2 dataset = load_dataset("JourneyDB/JourneyDB", data_files="data", use_auth_token=True)
File /opt/conda/lib/python3.10/site-packages/datasets/load.py:1664, in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, ignore_verifications, keep_in_memory, save_infos, revision, use_auth_token, task, streaming, **config_kwargs)
1661 ignore_verifications = ignore_verifications or save_infos
1663 # Create a dataset builder
-> 1664 builder_instance = load_dataset_builder(
1665 path=path,
1666 name=name,
1667 data_dir=data_dir,
1668 data_files=data_files,
1669 cache_dir=cache_dir,
1670 features=features,
1671 download_config=download_config,
1672 download_mode=download_mode,
1673 revision=revision,
1674 use_auth_token=use_auth_token,
1675 **config_kwargs,
1676 )
1678 # Return iterable dataset in case of streaming
1679 if streaming:
File /opt/conda/lib/python3.10/site-packages/datasets/load.py:1490, in load_dataset_builder(path, name, data_dir, data_files, cache_dir, features, download_config, download_mode, revision, use_auth_token, **config_kwargs)
1488 download_config = download_config.copy() if download_config else DownloadConfig()
1489 download_config.use_auth_token = use_auth_token
-> 1490 dataset_module = dataset_module_factory(
1491 path,
1492 revision=revision,
1493 download_config=download_config,
1494 download_mode=download_mode,
1495 data_dir=data_dir,
1496 data_files=data_files,
1497 )
1499 # Get dataset builder class from the processing script
1500 builder_cls = import_main_class(dataset_module.module_path)
File /opt/conda/lib/python3.10/site-packages/datasets/load.py:1238, in dataset_module_factory(path, revision, download_config, download_mode, force_local_path, dynamic_modules_path, data_dir, data_files, **download_kwargs)
1236 raise ConnectionError(f"Couln't reach the Hugging Face Hub for dataset '{path}': {e1}") from None
1237 if isinstance(e1, FileNotFoundError):
-> 1238 raise FileNotFoundError(
1239 f"Couldn't find a dataset script at {relative_to_absolute_path(combined_path)} or any data file in the same directory. "
1240 f"Couldn't find '{path}' on the Hugging Face Hub either: {type(e1).__name__}: {e1}"
1241 ) from None
1242 raise e1 from None
1243 else:
FileNotFoundError: Couldn't find a dataset script at /kaggle/working/JourneyDB/JourneyDB/JourneyDB.py or any data file in the same directory. Couldn't find 'JourneyDB/JourneyDB' on the Hugging Face Hub either: FileNotFoundError: Unable to find data in dataset repository JourneyDB/JourneyDB with any supported extension ['csv', 'tsv', 'json', 'jsonl', 'parquet', 'txt', 'blp', 'bmp', 'dib', 'bufr', 'cur', 'pcx', 'dcx', 'dds', 'ps', 'eps', 'fit', 'fits', 'fli', 'flc', 'ftc', 'ftu', 'gbr', 'gif', 'grib', 'h5', 'hdf', 'png', 'apng', 'jp2', 'j2k', 'jpc', 'jpf', 'jpx', 'j2c', 'icns', 'ico', 'im', 'iim', 'tif', 'tiff', 'jfif', 'jpe', 'jpg', 'jpeg', 'mpg', 'mpeg', 'msp', 'pcd', 'pxr', 'pbm', 'pgm', 'ppm', 'pnm', 'psd', 'bw', 'rgb', 'rgba', 'sgi', 'ras', 'tga', 'icb', 'vda', 'vst', 'webp', 'wmf', 'emf', 'xbm', 'xpm', 'zip']
```
### Steps to reproduce the bug
1)
```
from huggingface_hub import notebook_login
notebook_login()
```
2)
```
!pip install -q datasets
from datasets import load_dataset
```
3)
`dataset = load_dataset("JourneyDB/JourneyDB", data_files="data", use_auth_token=True)`
### Expected behavior
Load the dataset
### Environment info
Notebook | 6,261 |
https://github.com/huggingface/datasets/issues/6260 | REUSE_DATASET_IF_EXISTS don't work | [
"Hi! Unfortunately, the current behavior is to delete the downloaded data when this error happens. So, I've opened a PR that removes the problematic import to avoid losing data due to `apache_beam` not being installed (we host the preprocessed version of `natual_questions` on the HF GCS, so requiring `apache_beam` in that case doesn't make sense)",
"Thanks for your reply. I met another question that I set `export HF_DATASETS_CACHE=/data/lxy/.cache` , but each time I run load_datasets, the datasets module still looking for NQ in the wrong default cache dir '/home/lxy/.cache' 。How to avoid this incorrect behavior. I am sure HF_DATASETS_CACHE was set correctly since I use echo & to check it.\r\n\r\nby the way I delete the file in '/home/lxy/.cache' since I found there has some kb size file seems useless.",
"You need to set this variable before the `datasets` import. Then, you can use `import datasets; datasets.config.HF_DATASETS_CACHE` to verify the cache location."
] | ### Describe the bug
I use the following code to download natural_question dataset. Even though I have completely download it, the next time I run this code, the new download procedure will start and cover the original /data/lxy/NQ
config=datasets.DownloadConfig(resume_download=True,max_retries=100,cache_dir=r'/data/lxy/NQ',download_desc='NQ')
data=datasets.load_dataset('natural_questions',cache_dir=r'/data/lxy/NQ',download_config=config,download_mode=DownloadMode.REUSE_DATASET_IF_EXISTS)
---
Since I don't have apache_beam installed, it throw a exception. After I pip install apache_beam ,the download restart..
### Steps to reproduce the bug
run this two line code
config=datasets.DownloadConfig(resume_download=True,max_retries=100,cache_dir=r'/data/lxy/NQ',download_desc='NQ')
data=datasets.load_dataset('natural_questions',cache_dir=r'/data/lxy/NQ',download_config=config,download_mode=DownloadMode.REUSE_DATASET_IF_EXISTS)
### Expected behavior
Download behavior can be correctly follow DownloadMode
### Environment info
- `datasets` version: 2.14.4
- Platform: Linux-3.10.0-1160.88.1.el7.x86_64-x86_64-with-glibc2.17
- Python version: 3.9.17
- Huggingface_hub version: 0.16.4
- PyArrow version: 11.0.0
- Pandas version: 2.0.3 | 6,260 |
https://github.com/huggingface/datasets/issues/6259 | Duplicated Rows When Loading Parquet Files from Root Directory with Subdirectories | [
"Thanks for reporting this issue! We should be able to avoid this by making our `glob` patterns more precise. In the meantime, you can load the dataset by directly assigning splits to the data files: \r\n```python\r\nfrom datasets import load_dataset\r\nds = load_dataset(\"parquet\", data_files={\"train\": \"testing123/train/output_train.parquet\", \"validation\": \"testing123/val/output_val.parquet\"})\r\n```"
] | ### Describe the bug
When parquet files are saved in "train" and "val" subdirectories under a root directory, and datasets are then loaded using `load_dataset("parquet", data_dir="root_directory")`, the resulting dataset has duplicated rows for both the training and validation sets.
### Steps to reproduce the bug
1. Create a root directory, e.g., "testing123".
2. Under "testing123", create two subdirectories: "train" and "val".
3. Create and save a parquet file with 3 unique rows in the "train" subdirectory.
4. Create and save a parquet file with 4 unique rows in the "val" subdirectory.
5. Load the datasets from the root directory using `load_dataset("parquet", data_dir="testing123")`
6. Iterate through the datasets and print the rows
Here's a collab reproducing these steps:
https://colab.research.google.com/drive/11NEdImnQ3OqJlwKSHRMhr7jCBesNdLY4?usp=sharing
### Expected behavior
- Training set should contain 3 unique rows.
- Validation set should contain 4 unique rows.
### Environment info
- `datasets` version: 2.14.5
- Platform: Linux-5.15.120+-x86_64-with-glibc2.35
- Python version: 3.10.12
- Huggingface_hub version: 0.17.2
- PyArrow version: 9.0.0
- Pandas version: 1.5.3 | 6,259 |
https://github.com/huggingface/datasets/issues/6257 | HfHubHTTPError - exceeded our hourly quotas for action: commit | [
"how is your dataset structured? (file types, how many commits and files are you trying to push, etc)",
"I succeeded in uploading it after several attempts with an hour gap between each attempt (inconvenient but worked). The final dataset is [here](https://huggingface.co/datasets/yuvalkirstain/pickapic_v2), code and context to the dataset can be found [here](https://github.com/yuvalkirstain/PickScore/).\r\nI can close the issue if this behavior is intended, as most users probably do not need to upload large-scale datasets.",
"We could fix this by creating a single commit for all the (Parquet) shards in `push_to_hub` instead of one commit per shard, as we currently do. \r\n\r\n@Wauplin Any updates on the 2-step commit process suggested by you that we need to implement this?",
"> Any updates on the 2-step commit process suggested by you that we need to implement this?\r\n\r\nRe-prioritizing this, sorry. Will let you know but probably can be done this week."
] | ### Describe the bug
I try to upload a very large dataset of images, and get the following error:
```
File /fsx-multigen/yuvalkirstain/miniconda/envs/pickapic/lib/python3.10/site-packages/huggingface_hub/hf_api.py:2712, in HfApi.create_commit(self, repo_id, operations, commit_message, commit_description, token, repo_type, revision, create_pr, num_threads, parent_commit, run_as_future)
2710 try:
2711 commit_resp = get_session().post(url=commit_url, headers=headers, data=data, params=params)
-> 2712 hf_raise_for_status(commit_resp, endpoint_name="commit")
2713 except RepositoryNotFoundError as e:
2714 e.append_to_message(_CREATE_COMMIT_NO_REPO_ERROR_MESSAGE)
File /fsx-multigen/yuvalkirstain/miniconda/envs/pickapic/lib/python3.10/site-packages/huggingface_hub/utils/_errors.py:301, in hf_raise_for_status(response, endpoint_name)
297 raise BadRequestError(message, response=response) from e
299 # Convert `HTTPError` into a `HfHubHTTPError` to display request information
300 # as well (request id and/or server error message)
--> 301 raise HfHubHTTPError(str(e), response=response) from e
HfHubHTTPError: 429 Client Error: Too Many Requests for url: https://huggingface.co/api/datasets/yuvalkirstain/pickapic_v2/commit/main (Request ID: Root=1-65112399-12d63f7d7f28bfa40a36a0fd)
You have exceeded our hourly quotas for action: commit. We invite you to retry later.
```
this makes it much less convenient to host large datasets on HF hub.
### Steps to reproduce the bug
Upload a very large dataset of images
### Expected behavior
the upload to work well
### Environment info
- `datasets` version: 2.13.1
- Platform: Linux-5.15.0-1033-aws-x86_64-with-glibc2.31
- Python version: 3.10.11
- Huggingface_hub version: 0.15.1
- PyArrow version: 12.0.1
- Pandas version: 1.5.3 | 6,257 |
https://github.com/huggingface/datasets/issues/6256 | load_dataset() function's cache_dir does not seems to work | [
"Can you share the error message?\r\n\r\nAlso, it would help if you could check whether `huggingface_hub`'s download behaves the same:\r\n```python\r\nfrom huggingface_hub import snapshot_download\r\nsnapshot_download(\"trec\", repo_type=\"dataset\", cache_dir='/path/to/my/dir)\r\n```\r\n\r\nIn the next major release, we aim to switch to `huggingface_hub` for file download/caching, but we could align the `cache_dir`'s `umask` behavior earlier than this if their solution works for your use case."
] | ### Describe the bug
datasets version: 2.14.5
when trying to run the following command
trec = load_dataset('trec', split='train[:1000]', cache_dir='/path/to/my/dir')
I keep getting error saying the command does not have permission to the default cache directory on my macbook pro machine.
It seems the cache_dir parameter cannot change the dataset saving directory from the default
what ever explained in the https://huggingface.co/docs/datasets/cache does not seem to work
### Steps to reproduce the bug
datasets version: 2.14.5
when trying to run the following command
trec = load_dataset('trec', split='train[:1000]', cache_dir='/path/to/my/dir')
I keep getting error saying the command does not have permission to the default cache directory on my macbook pro machine.
It seems the cache_dir parameter cannot change the dataset saving directory from the default
what ever explained in the https://huggingface.co/docs/datasets/cache does not seem to work
### Expected behavior
the dataset should be saved to the cache_dir points to
### Environment info
datasets version: 2.14.5
macos X: Ventura 13.4.1 (c) | 6,256 |
https://github.com/huggingface/datasets/issues/6254 | Dataset.from_generator() cost much more time in vscode debugging mode then running mode | [
"Answered on the forum: https://discuss.huggingface.co/t/dataset-from-generator-cost-much-more-time-in-vscode-debugging-mode-then-running-mode/56005/2"
] | ### Describe the bug
Hey there,
I’m using Dataset.from_generator() to convert a torch_dataset to the Huggingface Dataset.
However, when I debug my code on vscode, I find that it runs really slow on Dataset.from_generator() which may even 20 times longer then run the script on terminal.
### Steps to reproduce the bug
I write a simple test code :
```python
import os
from functools import partial
from typing import Callable
import torch
import time
from torch.utils.data import Dataset as TorchDataset
from datasets import load_from_disk, Dataset as HFDataset
import torch
from torch.utils.data import Dataset
class SimpleDataset(Dataset):
def __init__(self, data):
self.data = data
self.keys = list(data[0].keys())
def __len__(self):
return len(self.data)
def __getitem__(self, index):
sample = self.data[index]
return {key: sample[key] for key in self.keys}
def TorchDataset2HuggingfaceDataset(torch_dataset: TorchDataset, cache_dir: str = None
) -> HFDataset:
"""
convert torch dataset to huggingface dataset
"""
generator : Callable[[], TorchDataset] = lambda: (sample for sample in torch_dataset)
return HFDataset.from_generator(generator, cache_dir=cache_dir)
if __name__ == '__main__':
data = [
{'id': 1, 'name': 'Alice'},
{'id': 2, 'name': 'Bob'},
{'id': 3, 'name': 'Charlie'}
]
torch_dataset = SimpleDataset(data)
start_time = time.time()
huggingface_dataset = TorchDataset2HuggingfaceDataset(torch_dataset)
end_time = time.time()
print("time: ", end_time - start_time)
print(huggingface_dataset)
```
### Expected behavior
this test on my machine report that the running time on terminal is 0.086,
however the running time in debugging mode on vscode is 0.25, which I think is much longer than expected.
I’d like to know is the anything wrong in the code or just because of debugging?
I have traced the code and I find is this func which I get stuck.
```python
def create_config_id(
self,
config_kwargs: dict,
custom_features: Optional[Features] = None,
) -> str:
...
# stuck in this line
suffix = Hasher.hash(config_kwargs_to_add_to_suffix)
```
### Environment info
- `datasets` version: 2.12.0
- Platform: Linux-5.11.0-27-generic-x86_64-with-glibc2.31
- Python version: 3.11.3
- Huggingface_hub version: 0.17.2
- PyArrow version: 11.0.0
- Pandas version: 2.0.1 | 6,254 |
https://github.com/huggingface/datasets/issues/6252 | exif_transpose not done to Image (PIL problem) | [
"Indeed, it makes sense to do this by default. \r\n\r\nIn the meantime, you can use `.with_transform` to transpose the images when accessing them:\r\n\r\n```python\r\nimport PIL.ImageOps\r\n\r\ndef exif_transpose_transform(batch):\r\n batch[\"image\"] = [PIL.ImageOps.exif_transpose(image) for image in batch[\"image\"]]\r\n return batch\r\n\r\ndataset = dataset.with_transform(exif_transpose_transform)\r\n```",
"This operation sets some `Image` attributes to `None` (`.format`, `.filename`, etc.), causing our tests to fail, so I think we should wait for Datasets 3.0 to make this change. In version 3.0, storing image paths will be replaced by embedding image bytes, so there will be fewer instances where we use the `.filename` attribute."
] | ### Feature request
I noticed that some of my images loaded using PIL have some metadata related to exif that can rotate them when loading.
Since the dataset.features.Image uses PIL for loading, the loaded image may be rotated (width and height will be inverted) thus for tasks as object detection and layoutLM this can create some inconsistencies (between input bboxes and input images).
For now there is no option in datasets.features.Image to specify that. We need to do the following when preparing examples (when preparing images for training, test or inference):
```
from PIL import Image, ImageOps
pil = ImageOps.exif_transpose(pil)
```
reference: https://stackoverflow.com/a/63950647/5720150
Is it possible to add this by default to the datasets.feature.Image ? or to add the option to do the ImageOps.exif_transpose?
Thank you
### Motivation
Prevent having inverted data related to exif metadata that may affect object detection tasks
### Your contribution
Changing in datasets.featrues.Image I can help with that. | 6,252 |
https://github.com/huggingface/datasets/issues/6246 | Add new column to dataset | [
"I think it's an issue with the code.\r\n\r\nSpecifically:\r\n```python\r\ndataset = dataset['train'].add_column(\"/workspace/data\", new_column)\r\n```\r\n\r\nNow `dataset` is the train set with a new column. \r\nTo fix this, you can do:\r\n\r\n```python\r\ndataset['train'] = dataset['train'].add_column(\"/workspace/data\", new_column)\r\n```",
"> I think it's an issue with the code.\r\n> \r\n> Specifically:\r\n> \r\n> ```python\r\n> dataset = dataset['train'].add_column(\"/workspace/data\", new_column)\r\n> ```\r\n> \r\n> Now `dataset` is the train set with a new column. To fix this, you can do:\r\n> \r\n> ```python\r\n> dataset['train'] = dataset['train'].add_column(\"/workspace/data\", new_column)\r\n> ```\r\n\r\nThanks for your response, but i can not access mask images, please let me know why the problem still persists. Here is the notebook for reference: https://colab.research.google.com/drive/10lZ_zLtU4itYVmIVTvIEVbjfOtCZaAZy?usp=sharing ",
"I think there is a slight misunderstanding.\r\n```python\r\nnew_column = [\"mask\"] * len(dataset[\"train\"])\r\ndataset['train'] = dataset['train'].add_column(\"/workspace/data\", new_column)\r\n```\r\n\r\nadds a column with the string `mask` to your dataset.\r\nIf you're trying to load the images `\"mask_{idx}.png\"` in your dataset, you could try:\r\n\r\n```\r\nfrom datasets import Image\r\n\r\ndataset['train'] = dataset['train'].map(lambda u, idx: {'mask': f\"/workspace/data/mask_{idx}.png\", with_indices=True).cast_column(\"mask\", Image())\r\n```\r\n\r\nWhat this does is that it adds a column to your dataset name `mask` with the path to the mask, then it cast the column as an `Image` feature.\r\n\r\nThis [link](https://huggingface.co/docs/datasets/v2.5.1/en/image_load) explains how to load images.\r\n\r\nHope this helps!",
"> I think there is a slight misunderstanding.\r\n> \r\n> ```python\r\n> new_column = [\"mask\"] * len(dataset[\"train\"])\r\n> dataset['train'] = dataset['train'].add_column(\"/workspace/data\", new_column)\r\n> ```\r\n> \r\n> adds a column with the string `mask` to your dataset. If you're trying to load the images `\"mask_{idx}.png\"` in your dataset, you could try:\r\n> \r\n> ```\r\n> from datasets import Image\r\n> \r\n> dataset['train'] = dataset['train'].map(lambda u, idx: {'mask': f\"/workspace/data/mask_{idx}.png\", with_indices=True).cast_column(\"mask\", Image())\r\n> ```\r\n> \r\n> What this does is that it adds a column to your dataset name `mask` with the path to the mask, then it cast the column as an `Image` feature.\r\n> \r\n> This [link](https://huggingface.co/docs/datasets/v2.5.1/en/image_load) explains how to load images.\r\n> \r\n> Hope this helps!\r\n\r\nThank you very much, this is really helpful...\r\ni made some changes for it to work:\r\n```\r\ndataset['train'] = dataset['train'].map(lambda u, idx: {'mask': f\"/content/data/mask_{idx}.png\"}, with_indices=True).cast_column(\"mask\", Image())\r\n```\r\nThanks Again @Dref360 "
] | ### Describe the bug
```
---------------------------------------------------------------------------
KeyError Traceback (most recent call last)
[<ipython-input-9-bd197b36b6a0>](https://localhost:8080/#) in <cell line: 1>()
----> 1 dataset['train']['/workspace/data']
3 frames
[/usr/local/lib/python3.10/dist-packages/datasets/formatting/formatting.py](https://localhost:8080/#) in _check_valid_column_key(key, columns)
518 def _check_valid_column_key(key: str, columns: List[str]) -> None:
519 if key not in columns:
--> 520 raise KeyError(f"Column {key} not in the dataset. Current columns in the dataset: {columns}")
521
522
KeyError: "Column train not in the dataset. Current columns in the dataset: ['image', '/workspace/data']"
```
### Steps to reproduce the bug
please find the notebook for reference: https://colab.research.google.com/drive/10lZ_zLtU4itYVmIVTvIEVbjfOtCZaAZy?usp=sharing
### Expected behavior
add column to the dataset
### Environment info
colab pro | 6,246 |
https://github.com/huggingface/datasets/issues/6242 | Data alteration when loading dataset with unspecified inner sequence length | [
"While this issue may seem specific, it led to a silent problem in my workflow that took days to diagnose. If this feature is not intended to be supported, an error should be raised when encountering this configuration to prevent such issues.",
"Thanks for reporting! This is a MRE:\r\n\r\n```python\r\nimport pyarrow as pa\r\nfrom datasets.table import cast_array_to_feature\r\nfrom datasets import Sequence, Value\r\ndata = [\r\n [[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]],\r\n [[7.0, 8.0, 9.0], [10.0, 11.0, 12.0]],\r\n]\r\narr = pa.array(data, pa.list_(pa.list_(pa.float32(), 3)))\r\ncast_array_to_feature(arr, Sequence(Sequence(Value(\"float32\"))))\r\n```\r\n\r\nI've opened a PR with a fix."
] | ### Describe the bug
When a dataset saved with a specified inner sequence length is loaded without specifying that length, the original data is altered and becomes inconsistent.
### Steps to reproduce the bug
```python
from datasets import Dataset, Features, Value, Sequence, load_dataset
# Repository ID
repo_id = "my_repo_id"
# Define features with a specific length of 3 for each inner sequence
specified_features = Features({"key": Sequence(Sequence(Value("float32"), length=3))})
# Create a dataset with the specified features
data = [
[[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]],
[[7.0, 8.0, 9.0], [10.0, 11.0, 12.0]],
]
dataset = Dataset.from_dict({"key": data}, features=specified_features)
# Push the dataset to the hub
dataset.push_to_hub(repo_id)
# Define features without specifying the length
unspecified_features = Features({"key": Sequence(Sequence(Value("float32")))})
# Load the dataset from the hub with this new feature definition
dataset = load_dataset(f"qgallouedec/{repo_id}", split="train", features=unspecified_features)
# The obtained data is altered
print(dataset.to_dict()) # {'key': [[[1.0], [2.0]], [[3.0], [4.0]]]}
```
### Expected behavior
```python
print(dataset.to_dict()) # {'key': [[[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]], [[7.0, 8.0, 9.0], [10.0, 11.0, 12.0]]]}
```
### Environment info
- `datasets` version: 2.14.4
- Platform: Linux-6.2.0-32-generic-x86_64-with-glibc2.35
- Python version: 3.9.12
- Huggingface_hub version: 0.15.1
- PyArrow version: 12.0.1
- Pandas version: 2.0.3 | 6,242 |
https://github.com/huggingface/datasets/issues/6240 | Dataloader stuck on multiple GPUs | [
"What type of dataset are you using in this script? `torch.utils.data.Dataset` or `datasets.Dataset`? Please share the `datasets` package version if it's the latter. Otherwise, it's better to move this issue to the `accelerate` repo.",
"Very sorry, I thought I had a repo in `accelerate!`\r\nI will close this issue and repo the issue in the appropriate place."
] | ### Describe the bug
I am trying to get CLIP to fine-tuning with my code.
When I tried to run it on multiple GPUs using accelerate, I encountered the following phenomenon.
- Validation dataloader stuck in 2nd epoch only on multi-GPU
Specifically, when the "for inputs in valid_loader:" process is finished, it does not proceed to the next step. train_loader process is completed. Also, both train and valid are working correctly in the first epoch.
The accelerate command at that time is as follows.
`accelerate launch --multi_gpu --num_processes=2 {script_name.py} {--arg1} {--arg2} ...`
- This will not happen when single GPU is used.
`CUDA_VISIBLE_DEVICES="0" accelerate launch {script_name.py} --arg1 --arg2 ...`
- Setting num_workers=0 in dataloader did not change the result.
### Steps to reproduce the bug
1. The codes for fine-tuning the regular CLIP were updated for accelerate.
2. Run the code with the accelerate command as `accelerate launch --multi_gpu --num_processes=2 {script_name.py} {--arg1} {--arg2} ...` and the above problem will occur.
3. CUDA_VISIBLE_DEVICES="0" accelerate launch {script_name.py} --arg1 --arg2 ...` , it works fine.
### Expected behavior
It Should end normally as if it was run on a single GPU.
### Environment info
Since `datasets-cli env` did not work, the environment is described below.
- OS: Ubuntu 22.04 with Docker
- Docker: 24.0.5, build ced0996
- Python: 3.10.12
- torch==2.0.1
- accelerate==0.21.0
- transformers==4.33.1 | 6,240 |
https://github.com/huggingface/datasets/issues/6239 | Load local audio data doesn't work | [
"I think this is the same issue as https://github.com/huggingface/datasets/issues/4776. Maybe installing `ffmpeg` can fix it:\r\n```python\r\nadd-apt-repository -y ppa:savoury1/ffmpeg4\r\napt-get -qq install -y ffmpeg\r\n```\r\n\r\nHowever, the best solution is to use a newer version of `datasets`. In the recent releases, we've replaced `torchaudio` with `soundfile`, which is easier to install and faster.",
"@mariosasko \r\nThanks for your help"
] | ### Describe the bug
I get a RuntimeError from the following code:
```python
audio_dataset = Dataset.from_dict({"audio": ["/kaggle/input/bengaliai-speech/train_mp3s/000005f3362c.mp3"]}).cast_column("audio", Audio())
audio_dataset[0]
```
### Traceback
<details>
```python
RuntimeError Traceback (most recent call last)
Cell In[33], line 1
----> 1 train_dataset[0]
File /opt/conda/lib/python3.10/site-packages/datasets/arrow_dataset.py:1764, in Dataset.__getitem__(self, key)
1762 def __getitem__(self, key): # noqa: F811
1763 """Can be used to index columns (by string names) or rows (by integer index or iterable of indices or bools)."""
-> 1764 return self._getitem(
1765 key,
1766 )
File /opt/conda/lib/python3.10/site-packages/datasets/arrow_dataset.py:1749, in Dataset._getitem(self, key, decoded, **kwargs)
1747 formatter = get_formatter(format_type, features=self.features, decoded=decoded, **format_kwargs)
1748 pa_subtable = query_table(self._data, key, indices=self._indices if self._indices is not None else None)
-> 1749 formatted_output = format_table(
1750 pa_subtable, key, formatter=formatter, format_columns=format_columns, output_all_columns=output_all_columns
1751 )
1752 return formatted_output
File /opt/conda/lib/python3.10/site-packages/datasets/formatting/formatting.py:532, in format_table(table, key, formatter, format_columns, output_all_columns)
530 python_formatter = PythonFormatter(features=None)
531 if format_columns is None:
--> 532 return formatter(pa_table, query_type=query_type)
533 elif query_type == "column":
534 if key in format_columns:
File /opt/conda/lib/python3.10/site-packages/datasets/formatting/formatting.py:281, in Formatter.__call__(self, pa_table, query_type)
279 def __call__(self, pa_table: pa.Table, query_type: str) -> Union[RowFormat, ColumnFormat, BatchFormat]:
280 if query_type == "row":
--> 281 return self.format_row(pa_table)
282 elif query_type == "column":
283 return self.format_column(pa_table)
File /opt/conda/lib/python3.10/site-packages/datasets/formatting/formatting.py:312, in PythonFormatter.format_row(self, pa_table)
310 row = self.python_arrow_extractor().extract_row(pa_table)
311 if self.decoded:
--> 312 row = self.python_features_decoder.decode_row(row)
313 return row
File /opt/conda/lib/python3.10/site-packages/datasets/formatting/formatting.py:221, in PythonFeaturesDecoder.decode_row(self, row)
220 def decode_row(self, row: dict) -> dict:
--> 221 return self.features.decode_example(row) if self.features else row
File /opt/conda/lib/python3.10/site-packages/datasets/features/features.py:1386, in Features.decode_example(self, example)
1376 def decode_example(self, example: dict):
1377 """Decode example with custom feature decoding.
1378
1379 Args:
(...)
1383 :obj:`dict[str, Any]`
1384 """
-> 1386 return {
1387 column_name: decode_nested_example(feature, value)
1388 if self._column_requires_decoding[column_name]
1389 else value
1390 for column_name, (feature, value) in zip_dict(
1391 {key: value for key, value in self.items() if key in example}, example
1392 )
1393 }
File /opt/conda/lib/python3.10/site-packages/datasets/features/features.py:1387, in <dictcomp>(.0)
1376 def decode_example(self, example: dict):
1377 """Decode example with custom feature decoding.
1378
1379 Args:
(...)
1383 :obj:`dict[str, Any]`
1384 """
1386 return {
-> 1387 column_name: decode_nested_example(feature, value)
1388 if self._column_requires_decoding[column_name]
1389 else value
1390 for column_name, (feature, value) in zip_dict(
1391 {key: value for key, value in self.items() if key in example}, example
1392 )
1393 }
File /opt/conda/lib/python3.10/site-packages/datasets/features/features.py:1087, in decode_nested_example(schema, obj)
1085 # Object with special decoding:
1086 elif isinstance(schema, (Audio, Image)):
-> 1087 return schema.decode_example(obj) if obj is not None else None
1088 return obj
File /opt/conda/lib/python3.10/site-packages/datasets/features/audio.py:103, in Audio.decode_example(self, value)
101 raise ValueError(f"An audio sample should have one of 'path' or 'bytes' but both are None in {value}.")
102 elif path is not None and path.endswith("mp3"):
--> 103 array, sampling_rate = self._decode_mp3(file if file else path)
104 elif path is not None and path.endswith("opus"):
105 if file:
File /opt/conda/lib/python3.10/site-packages/datasets/features/audio.py:241, in Audio._decode_mp3(self, path_or_file)
238 except RuntimeError as err:
239 raise ImportError("To support decoding 'mp3' audio files, please install 'sox'.") from err
--> 241 array, sampling_rate = torchaudio.load(path_or_file, format="mp3")
242 if self.sampling_rate and self.sampling_rate != sampling_rate:
243 if not hasattr(self, "_resampler") or self._resampler.orig_freq != sampling_rate:
File /opt/conda/lib/python3.10/site-packages/torchaudio/backend/sox_io_backend.py:256, in load(filepath, frame_offset, num_frames, normalize, channels_first, format)
254 if ret is not None:
255 return ret
--> 256 return _fallback_load(filepath, frame_offset, num_frames, normalize, channels_first, format)
File /opt/conda/lib/python3.10/site-packages/torchaudio/backend/sox_io_backend.py:30, in _fail_load(filepath, frame_offset, num_frames, normalize, channels_first, format)
22 def _fail_load(
23 filepath: str,
24 frame_offset: int = 0,
(...)
28 format: Optional[str] = None,
29 ) -> Tuple[torch.Tensor, int]:
---> 30 raise RuntimeError("Failed to load audio from {}".format(filepath))
RuntimeError: Failed to load audio from /kaggle/input/bengaliai-speech/train_mp3s/000005f3362c.mp3
```
</details>
### Steps to reproduce the bug
1. - Create a custom dataset using Local files of type mp3.
3. - Try to read the first audio item.
### Expected behavior
Expected output
```python
audio_dataset[0]["audio"]
{'array': array([ 0. , 0.00024414, -0.00024414, ..., -0.00024414,
0. , 0. ], dtype=float32),
'path': 'path/to/audio_1',
'sampling_rate': 16000}
```
### Environment info
N/A | 6,239 |
https://github.com/huggingface/datasets/issues/6238 | `dataset.filter` ALWAYS removes the first item from the dataset when using batched=True | [
"`filter` treats the function's output as a (selection) mask - `True` keeps the sample, and `False` drops it. In your case, `bool(0)` evaluates to `False`, so dropping the first sample is the correct behavior.",
"Oh gosh! 🤦 I totally misunderstood the API! My apologies!"
] | ### Describe the bug
If you call batched=True when calling `filter`, the first item is _always_ filtered out, regardless of the filter condition.
### Steps to reproduce the bug
Here's a minimal example:
```python
def filter_batch_always_true(batch, indices):
print("First index being passed into this filter function: ", indices[0])
return indices # Keep all indices
data = {"value": list(range(10))}
dataset = Dataset.from_dict(data)
filtered_dataset = dataset.filter(filter_batch_always_true, with_indices=True, batched=True)
print("Length of original dataset: ", len(dataset))
print("Length of filtered_dataset: ", len(filtered_dataset))
print("Is equal to original? ", len(filtered_dataset) == len(dataset))
print("First item of filtered dataset: ", filtered_dataset[0])
print("Last item of filtered dataset: ", filtered_dataset[-1])
```
prints:
```
First index being passed into this filter function: 0
Length of original dataset: 10
Length of filtered_dataset: 9
Is equal to original? False
First item of filtered dataset: {'value': 1}
Last item of filtered dataset: {'value': 9}
```
### Expected behavior
Filter should respect the filter condition.
### Environment info
- `datasets` version: 2.14.4
- Platform: macOS-13.5-arm64-arm-64bit
- Python version: 3.9.18
- Huggingface_hub version: 0.17.1
- PyArrow version: 10.0.1
- Pandas version: 2.0.2
| 6,238 |
https://github.com/huggingface/datasets/issues/6237 | Tokenization with multiple workers is too slow | [
"[This](https://huggingface.co/docs/datasets/nlp_process#map) is the most performant way to tokenize a dataset (`batched=True, num_proc=None, return_tensors=\"np\"`) \r\n\r\nIf`tokenizer.is_fast` returns `True`, `num_proc` must be `None/1` to benefit from the fast tokenizers' parallelism (the fast tokenizers are implemented in Rust, and Rust multi-threading doesn't work well with Python multi-processing)"
] | I am trying to tokenize a few million documents with multiple workers but the tokenization process is taking forever.
Code snippet:
```
raw_datasets.map(
encode_function,
batched=False,
num_proc=args.preprocessing_num_workers,
load_from_cache_file=not args.overwrite_cache,
remove_columns=[name for name in raw_datasets["train"].column_names if name not in ["input_ids", "labels", "attention_mask"]],
desc="Tokenizing data",
)
```
Details:
```
transformers==4.28.0.dev0
datasets==4.28.0.dev0
preprocessing_num_workers==48
```
tokenizer == decapoda-research/llama-7b-hf
| 6,237 |
https://github.com/huggingface/datasets/issues/6236 | Support buffer shuffle for to_tf_dataset | [
"cc @Rocketknight1 ",
"Hey! You can implement this yourself, just:\r\n\r\n1) Create the dataset with `to_tf_dataset()` with `shuffle=False`\r\n2) Add an `unbatch()` at the end (or use batch_size=1)\r\n3) Add a `shuffle()` to the resulting dataset with your desired buffer size\r\n4) Add a `batch()` at the end again to re-batch your dataset.\r\n\r\nNote that the way we construct datasets in `to_tf_dataset()`, we don't actually shuffle the entire dataset in-memory, using `tf.data.Dataset.shuffle()`! Instead, we shuffle an index array and then load from the dataset with that. This means that shuffling with `tf.data.Dataset.shuffle()` will probably be slower and use more memory than our approach - I don't think adding the option for smaller shuffle buffers will actually save you memory on this!",
"Thanks for your reply! @Rocketknight1 \r\n\"We don't actually shuffle the entire dataset in-memory, using tf.data.Dataset.shuffle()! Instead, we shuffle an index array and then load from the dataset with that.\"\r\nIn such case, there will be random access to dataset data during shuffling. When the dataset is large, the performance can be X10 times slow. I have tried many ways with to_tf_dataset() trying to achieve comparable performance with tf.data.Dataset().shuffle(buffer_size).batch(). But the performance with to_tf_dataset() is still slow. \r\n"
] | ### Feature request
I'm using to_tf_dataset to convert a large dataset to tf.data.Dataset and use Keras fit to train model.
Currently, to_tf_dataset only supports full size shuffle, which can be very slow on large dataset.
tf.data.Dataset support buffer shuffle by default.
shuffle(
buffer_size, seed=None, reshuffle_each_iteration=None, name=None
)
### Motivation
I'm very frustrated to find the loading with shuffling large dataset is very slow. It seems impossible to shuffle before training Keras with big dataset.
### Your contribution
NA | 6,236 |
https://github.com/huggingface/datasets/issues/6235 | Support multiprocessing for download/extract nestedly | [] | ### Feature request
Current multiprocessing for download/extract is not done nestedly. For example, when processing SlimPajama, there is only 3 processes (for train/test/val), while there are many files inside these 3 folders
```
Downloading data files #0: 0%| | 0/1 [00:00<?, ?obj/s]
Downloading data files #1: 0%| | 0/1 [00:00<?, ?obj/s]
Downloading data files #2: 0%| | 0/1 [00:00<?, ?obj/s]
Extracting data files #0: 0%| | 0/1 [00:00<?, ?obj/s]
Extracting data files #1: 0%| | 0/1 [00:00<?, ?obj/s][A
Extracting data files #2: 0%| | 0/1 [00:00<?, ?obj/s][A[A
```
### Motivation
speedup dataset loading
### Your contribution
I can help test the feature | 6,235 |
https://github.com/huggingface/datasets/issues/6229 | Apply inference on all images in the dataset | [
"From what I see, `MMSegInferencer` supports NumPy arrays, so replace the line `image_path = example['image']` with `image_path = np.array(example['image'])` to fix the issue (`example[\"image\"]` is a `PIL.Image` object). ",
"> From what I see, `MMSegInferencer` supports NumPy arrays, so replace the line `image_path = example['image']` with `image_path = np.array(example['image'])` to fix the issue (`example[\"image\"]` is a `PIL.Image` object).\r\n\r\nThanks @mariosasko for your reply...\r\ni tried :\r\n```\r\n# Define a function to apply the code to each image in the dataset\r\ndef process_image(image_path):\r\n print(\"Processing image:\", image_path)\r\n result = inferencer(image_path)['predictions']\r\n mask = np.where(result == 12, 255, 0).astype('uint8')\r\n return Image.fromarray(mask)\r\n\r\n# Process and save masks for each image in the dataset\r\nfor idx, example in enumerate(dataset['train']):\r\n image_path = np.array(example['image'])\r\n mask_image = process_image(image_path)\r\n mask_image.save(f\"mask_{idx}.png\")\r\n```\r\nand got\r\n```\r\nProcessing image: [[[202 165 87]\r\n [203 166 88]\r\n [207 168 91]\r\n ...\r\n [243 205 122]\r\n [244 202 120]\r\n [242 200 118]]\r\n\r\n [[202 165 87]\r\n [203 166 88]\r\n [207 168 91]\r\n ...\r\n [244 206 123]\r\n [245 203 121]\r\n [243 201 119]]\r\n\r\n [[203 164 87]\r\n [204 165 88]\r\n [207 168 91]\r\n ...\r\n [245 207 126]\r\n [246 204 122]\r\n [245 203 121]]\r\n\r\n ...\r\n\r\n [[154 123 56]\r\n [155 124 57]\r\n [158 125 56]\r\n ...\r\n [ 3 3 1]\r\n [ 3 3 1]\r\n [ 3 3 1]]\r\n\r\n [[154 123 56]\r\n [154 123 56]\r\n [155 124 57]\r\n ...\r\n [ 2 2 0]\r\n [ 2 2 0]\r\n [ 2 2 0]]\r\n\r\n [[152 121 54]\r\n [152 121 54]\r\n [153 122 55]\r\n ...\r\n [ 2 2 0]\r\n [ 2 2 0]\r\n [ 2 2 0]]]\r\nInference ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ \r\nProcessing image: [[[ 39 44 40]\r\n [ 39 44 40]\r\n [ 39 43 44]\r\n ...\r\n [187 185 164]\r\n [208 204 175]\r\n [203 198 166]]\r\n\r\n [[ 42 47 43]\r\n [ 40 45 41]\r\n [ 40 44 45]\r\n ...\r\n [188 186 165]\r\n [202 198 169]\r\n [201 196 164]]\r\n\r\n [[ 41 46 42]\r\n [ 39 44 40]\r\n [ 40 44 45]\r\n ...\r\n [187 184 165]\r\n [197 193 166]\r\n [201 196 166]]\r\n\r\n ...\r\n\r\n [[ 29 27 30]\r\n [ 28 26 29]\r\n [ 25 23 26]\r\n ...\r\n [ 48 33 28]\r\n [ 44 31 25]\r\n [ 39 26 20]]\r\n\r\n [[ 34 29 33]\r\n [ 32 27 31]\r\n [ 29 24 28]\r\n ...\r\n [ 30 17 11]\r\n [ 36 23 15]\r\n [ 41 28 20]]\r\n\r\n [[ 35 30 34]\r\n [ 33 28 32]\r\n [ 28 23 27]\r\n ...\r\n [ 28 15 9]\r\n [ 41 28 20]\r\n [ 46 33 25]]]\r\nInference ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ \r\nProcessing image: [[[ 65 53 55]\r\n [ 65 53 55]\r\n [ 51 39 41]\r\n ...\r\n [133 127 111]\r\n [150 141 124]\r\n [133 124 107]]\r\n\r\n [[ 58 45 52]\r\n [ 61 48 55]\r\n [ 51 38 45]\r\n ...\r\n [148 141 123]\r\n [178 169 152]\r\n [144 135 118]]\r\n\r\n [[ 79 66 83]\r\n [ 73 60 77]\r\n [ 65 51 66]\r\n ...\r\n [140 131 114]\r\n [142 133 116]\r\n [147 136 118]]\r\n\r\n ...\r\n\r\n [[132 122 133]\r\n [ 95 85 94]\r\n [ 61 51 60]\r\n ...\r\n [ 39 28 42]\r\n [ 46 36 45]\r\n [ 25 16 21]]\r\n\r\n [[150 143 151]\r\n [114 107 115]\r\n [ 64 54 63]\r\n ...\r\n [ 47 35 47]\r\n [ 38 27 35]\r\n [140 129 133]]\r\n\r\n [[145 138 146]\r\n [115 108 116]\r\n [ 69 59 67]\r\n ...\r\n [ 31 19 31]\r\n [128 117 123]\r\n [196 185 189]]]\r\nInference ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ \r\nProcessing image: [[[159 151 140]\r\n [171 163 152]\r\n [161 148 142]\r\n ...\r\n [198 184 171]\r\n [189 175 162]\r\n [183 169 156]]\r\n\r\n [[128 118 106]\r\n [138 128 116]\r\n [138 125 116]\r\n ...\r\n [200 186 173]\r\n [190 176 163]\r\n [187 173 160]]\r\n\r\n [[165 153 137]\r\n [170 158 142]\r\n [174 162 148]\r\n ...\r\n [200 187 171]\r\n [188 175 159]\r\n [182 169 153]]\r\n```\r\nHowever , when trying to add to:\r\n```\r\nfrom datasets import load_dataset\r\ndataset = load_dataset('Andyrasika/cat_kingdom')\r\ndataset\r\n```\r\ni did \r\n```\r\nnew_column = [\"mask\"] * len(dataset[\"train\"])\r\nnew_column\r\ndataset = dataset.add_column(\"/workspace/data\", new_column)\r\n\r\nprint(dataset)\r\n```\r\ngot error:\r\n```\r\n---------------------------------------------------------------------------\r\nAttributeError Traceback (most recent call last)\r\nCell In[11], line 3\r\n 1 new_column = [\"mask\"] * len(dataset[\"train\"])\r\n 2 new_column\r\n----> 3 dataset = dataset.add_column(\"/workspace/data\", new_column)\r\n 5 print(dataset)\r\n\r\nAttributeError: 'DatasetDict' object has no attribute 'add_column'\r\n```",
"https://github.com/huggingface/datasets/issues/6246 resolved the `add_column` error, so I'm closing this issue :) "
] | ### Describe the bug
```
---------------------------------------------------------------------------
NotImplementedError Traceback (most recent call last)
Cell In[14], line 11
9 for idx, example in enumerate(dataset['train']):
10 image_path = example['image']
---> 11 mask_image = process_image(image_path)
12 mask_image.save(f"mask_{idx}.png")
Cell In[14], line 4, in process_image(image_path)
2 def process_image(image_path):
3 print("Processing image:", image_path)
----> 4 result = inferencer(image_path)['predictions']
5 mask = np.where(result == 12, 255, 0).astype('uint8')
6 return Image.fromarray(mask)
File /usr/local/lib/python3.10/dist-packages/mmseg/apis/mmseg_inferencer.py:183, in MMSegInferencer.__call__(self, inputs, return_datasamples, batch_size, show, wait_time, out_dir, img_out_dir, pred_out_dir, **kwargs)
180 pred_out_dir = ''
181 img_out_dir = ''
--> 183 return super().__call__(
184 inputs=inputs,
185 return_datasamples=return_datasamples,
186 batch_size=batch_size,
187 show=show,
188 wait_time=wait_time,
189 img_out_dir=img_out_dir,
190 pred_out_dir=pred_out_dir,
191 **kwargs)
File /usr/local/lib/python3.10/dist-packages/mmengine/infer/infer.py:221, in BaseInferencer.__call__(self, inputs, return_datasamples, batch_size, **kwargs)
218 inputs = self.preprocess(
219 ori_inputs, batch_size=batch_size, **preprocess_kwargs)
220 preds = []
--> 221 for data in (track(inputs, description='Inference')
222 if self.show_progress else inputs):
223 preds.extend(self.forward(data, **forward_kwargs))
224 visualization = self.visualize(
225 ori_inputs, preds,
226 **visualize_kwargs) # type: ignore # noqa: E501
File /usr/local/lib/python3.10/dist-packages/rich/progress.py:168, in track(sequence, description, total, auto_refresh, console, transient, get_time, refresh_per_second, style, complete_style, finished_style, pulse_style, update_period, disable, show_speed)
157 progress = Progress(
158 *columns,
159 auto_refresh=auto_refresh,
(...)
164 disable=disable,
165 )
167 with progress:
--> 168 yield from progress.track(
169 sequence, total=total, description=description, update_period=update_period
170 )
File /usr/local/lib/python3.10/dist-packages/rich/progress.py:1210, in Progress.track(self, sequence, total, task_id, description, update_period)
1208 if self.live.auto_refresh:
1209 with _TrackThread(self, task_id, update_period) as track_thread:
-> 1210 for value in sequence:
1211 yield value
1212 track_thread.completed += 1
File /usr/local/lib/python3.10/dist-packages/mmengine/infer/infer.py:291, in BaseInferencer.preprocess(self, inputs, batch_size, **kwargs)
266 """Process the inputs into a model-feedable format.
267
268 Customize your preprocess by overriding this method. Preprocess should
(...)
287 Any: Data processed by the ``pipeline`` and ``collate_fn``.
288 """
289 chunked_data = self._get_chunk_data(
290 map(self.pipeline, inputs), batch_size)
--> 291 yield from map(self.collate_fn, chunked_data)
File /usr/local/lib/python3.10/dist-packages/mmengine/infer/infer.py:588, in BaseInferencer._get_chunk_data(self, inputs, chunk_size)
586 chunk_data = []
587 for _ in range(chunk_size):
--> 588 processed_data = next(inputs_iter)
589 chunk_data.append(processed_data)
590 yield chunk_data
File /usr/local/lib/python3.10/dist-packages/mmcv/transforms/base.py:12, in BaseTransform.__call__(self, results)
9 def __call__(self,
10 results: Dict) -> Optional[Union[Dict, Tuple[List, List]]]:
---> 12 return self.transform(results)
File /usr/local/lib/python3.10/dist-packages/mmcv/transforms/wrappers.py:88, in Compose.transform(self, results)
79 """Call function to apply transforms sequentially.
80
81 Args:
(...)
85 dict or None: Transformed results.
86 """
87 for t in self.transforms:
---> 88 results = t(results) # type: ignore
89 if results is None:
90 return None
File /usr/local/lib/python3.10/dist-packages/mmcv/transforms/base.py:12, in BaseTransform.__call__(self, results)
9 def __call__(self,
10 results: Dict) -> Optional[Union[Dict, Tuple[List, List]]]:
---> 12 return self.transform(results)
File /usr/local/lib/python3.10/dist-packages/mmseg/datasets/transforms/loading.py:496, in InferencerLoader.transform(self, single_input)
494 inputs = single_input
495 else:
--> 496 raise NotImplementedError
498 if 'img' in inputs:
499 return self.from_ndarray(inputs)
NotImplementedError:
````
### Steps to reproduce the bug
```
from datasets import load_dataset
dataset = load_dataset('Andyrasika/cat_kingdom')
dataset
from mmseg.apis import MMSegInferencer
checkpoint_name = 'segformer_mit-b5_8xb2-160k_ade20k-640x640'
inferencer = MMSegInferencer(model=checkpoint_name)
# Define a function to apply the code to each image in the dataset
def process_image(image_path):
print("Processing image:", image_path)
result = inferencer(image_path)['predictions']
mask = np.where(result == 12, 255, 0).astype('uint8')
return Image.fromarray(mask)
# Process and save masks for each image in the dataset
for idx, example in enumerate(dataset['train']):
image_path = example['image']
mask_image = process_image(image_path)
mask_image.save(f"mask_{idx}.png")
```
### Expected behavior
create a separate column with masks in the dataset and further shows as a separate column in hub
### Environment info
jupyter notebook RTX 3090 | 6,229 |
https://github.com/huggingface/datasets/issues/6225 | Conversion from RGB to BGR in Object Detection tutorial | [
"Good catch!"
] | The [tutorial](https://huggingface.co/docs/datasets/main/en/object_detection) mentions the necessity of conversion the input image from BGR to RGB
> albumentations expects the image to be in BGR format, not RGB, so you’ll have to convert the image before applying the transform.
[Link to tutorial](https://github.com/huggingface/datasets/blob/0a068dbf3b446417ffd89d32857608394ec699e6/docs/source/object_detection.mdx#L77)
However, relevant albumentations' tutorials [on channels conversion](https://albumentations.ai/docs/examples/example/#read-the-image-from-the-disk-and-convert-it-from-the-bgr-color-space-to-the-rgb-color-space) and [on boxes](https://albumentations.ai/docs/examples/example_bboxes/) imply that it's not really true no more.
I suggest removing this outdated conversion from the tutorial. | 6,225 |
https://github.com/huggingface/datasets/issues/6221 | Support saving datasets with custom formatting | [
"Not a fan of pickling this sort of stuff either.\r\nNote that users can also share the code in their dataset documentation."
] | Requested in https://discuss.huggingface.co/t/using-set-transform-on-a-dataset-leads-to-an-exception/53036.
I am not sure if supporting this is the best idea for the following reasons:
>For this to work, we would have to pickle a custom transform, which means the transform and the objects it references need to be serializable. Also, deserializing these bytes would make `load_from_disk` unsafe, so I'm not sure this is a good idea.
@lhoestq WDYT?
| 6,221 |
https://github.com/huggingface/datasets/issues/6217 | `Dataset.to_dict()` ignore `decode=True` with Image feature | [
"We need to implement the `Image` type as a PyArrow extension type (to allow us to override the Python conversion) for this to work as expected. For now, it's best to use your approach indeed."
] | ### Describe the bug
`Dataset.to_dict` seems to ignore the decoding instruction passed in features.
### Steps to reproduce the bug
```python
import datasets
import numpy as np
from PIL import Image
img = np.random.randint(0, 256, (5, 5, 3), dtype=np.uint8)
img = Image.fromarray(img)
features = datasets.Features({"image": datasets.Image(decode=True)})
dataset = datasets.Dataset.from_dict({"image": [img]}, features=features)
print({key: dataset[key] for key in dataset.column_names})
# {'image': [<PIL.PngImagePlugin.PngImageFile image mode=RGB size=5x5 at 0x7EFBC80E15B0>]}
print(dataset.to_dict())
# {'image': [{'bytes': b'\x89PNG\r\n\x1a\n\x00\x00\x00\rIHDR\x00\x00\x00\x05\x00\x00\x00\x05\x08\x02\x00\x00\x00\x02\r\xb1\xb2\x00\x00\x00[IDATx\x9c\x01P\x00\xaf\xff\x01\x13\x1b<7\xe7\xe0\xdc^6\xed\x04\xc7M\xd2\x9f\x00X\x1b\xb0?\x1ba\x15\xc5 o\xd0\x80\xbe\x19/\x01\xec\x95\x1f\x9f\xffj\xfa1\xa7\xc4X\xea\xbe\xa4g\x00\xc4\x15\xdeC\xc7 \xbbaqe\xc8\xb9\xa9q\xe7\x00,?M\xc0)\xdaD`}\xb1\xdci\x1e\xafC\xa9]%.@\xa6\xf0\xb3\x00\x00\x00\x00IEND\xaeB`\x82', 'path': None}]}
```
### Expected behavior
I would expect `{key: dataset[key] for key in dataset.column_names}` and `dataset.to_dict()` to be equivalent. If the previous behavior is expected, then it should be stated [in the doc](https://huggingface.co/docs/datasets/v2.14.4/en/package_reference/main_classes#datasets.Dataset.to_dict).
### Environment info
- `datasets` version: 2.14.4
- Platform: Linux-6.2.0-31-generic-x86_64-with-glibc2.35
- Python version: 3.9.12
- Huggingface_hub version: 0.15.1
- PyArrow version: 12.0.1
- Pandas version: 2.0.3
- Pillow 9.5.0
- numpy 1.25.2 | 6,217 |
https://github.com/huggingface/datasets/issues/6214 | Unpin fsspec < 2023.9.0 | [] | Once root issue is fixed, remove temporary pin of fsspec < 2023.9.0 introduced by:
- #6210
Related to issue:
- #6209
After investigation, I think the root issue is related to the new glob behavior with double asterisk `**` they have introduced in:
- https://github.com/fsspec/filesystem_spec/pull/1329 | 6,214 |
https://github.com/huggingface/datasets/issues/6212 | Tilde (~) is not supported for data_files | [
"Hi @exs-avianello, is it really needed? Note you can alternatively use `pathlib.Path` among others as it follows:\r\n\r\n```python\r\nimport datasets\r\nfrom pathlib import Path\r\n\r\n# save a parquet file at ~/path/to/data.parquet\r\n\r\ndata_files = Path.home() / \"path/to/data.parquet\"\r\ndataset = datasets.load_dataset(\"parquet\", data_files=data_files)\r\n```",
"Hi @alvarobartt ! \r\n\r\nThis is definitely just a \"nice to have\" and I am personally more than happy to just use absolute paths client-side. I just wanted to flag it up in case it can help improve the package even more 🙌 It might not be immediately obvious from the stack trace that the error is triggered by the `~` in the path"
] | ### Describe the bug
Attempting to `load_dataset` from a path starting with `~` (as a shorthand for the user's home directory) seems not to be fully working - at least as far as the `parquet` dataset builder is concerned.
(the same file can be loaded correctly if providing its absolute path instead)
I think that this is very similar to https://github.com/huggingface/datasets/issues/5757, but for `data_files` rather than `data_dir`
### Steps to reproduce the bug
```python
import datasets
# save a parquet file at ~/path/to/data.parquet
data_files = "~/path/to/data.parquet"
dataset = datasets.load_dataset("parquet", data_files=data_files)
```
```
Downloading data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 12671.61it/s]
Extracting data files: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 22671.91it/s]
Generating train split: 0 examples [00:00, ? examples/s]
Traceback (most recent call last):
File ".venv/lib/python3.11/site-packages/datasets/builder.py", line 1949, in _prepare_split_single
num_examples, num_bytes = writer.finalize()
^^^^^^^^^^^^^^^^^
File ".venv/lib/python3.11/site-packages/datasets/arrow_writer.py", line 598, in finalize
raise SchemaInferenceError("Please pass `features` or at least one example when writing data")
datasets.arrow_writer.SchemaInferenceError: Please pass `features` or at least one example when writing data
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File ".venv/lib/python3.11/site-packages/datasets/load.py", line 2133, in load_dataset
builder_instance.download_and_prepare(
File ".venv/lib/python3.11/site-packages/datasets/builder.py", line 954, in download_and_prepare
self._download_and_prepare(
File ".venv/lib/python3.11/site-packages/datasets/builder.py", line 1049, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File ".venv/lib/python3.11/site-packages/datasets/builder.py", line 1813, in _prepare_split
for job_id, done, content in self._prepare_split_single(
File ".venv/lib/python3.11/site-packages/datasets/builder.py", line 1958, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.builder.DatasetGenerationError: An error occurred while generating the dataset
```
### Expected behavior
Can use `~` shorthand in paths when loading local (parquet) datasets.
### Environment info
`datasets 2.14.3`
| 6,212 |
https://github.com/huggingface/datasets/issues/6209 | CI is broken with AssertionError: 3 failed, 12 errors | [] | Our CI is broken: 3 failed, 12 errors
See: https://github.com/huggingface/datasets/actions/runs/6069947111/job/16465138041
```
=========================== short test summary info ============================
FAILED tests/test_load.py::ModuleFactoryTest::test_LocalDatasetModuleFactoryWithoutScript_with_data_dir - AssertionError: assert ({NamedSplit('train'): ['/tmp/pytest-of-runner/pytest-0/popen-gw1/test_LocalDatasetModuleFactory2/data_dir2/subdir1/train.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_LocalDatasetModuleFactory2/data_dir2/subdir1/train.txt'], NamedSplit('test'): ['/tmp/pytest-of-runner/pytest-0/popen-gw1/test_LocalDatasetModuleFactory2/data_dir2/subdir1/test.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_LocalDatasetModuleFactory2/data_dir2/subdir1/test.txt']} is not None and 2 == 1)
+ where 2 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw1/test_LocalDatasetModuleFactory2/data_dir2/subdir1/train.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_LocalDatasetModuleFactory2/data_dir2/subdir1/train.txt'])
FAILED tests/test_load.py::test_load_dataset_arrow[False] - AssertionError: assert 20 == 10
+ where 20 = Dataset({\n features: ['col_1'],\n num_rows: 20\n}).num_rows
FAILED tests/test_load.py::test_load_dataset_arrow[True] - assert 20 == 10
ERROR tests/packaged_modules/test_audiofolder.py::test_data_files_with_metadata_and_multiple_splits[jsonl-False] - AssertionError: assert 6 == 3
+ where 6 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_2/audiofolder_data_dir_with_metadata/train/audio_file.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_2/audiofolder_data_dir_with_metadata/train/audio_file2.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_2/audiofolder_data_dir_with_metadata/train/metadata.jsonl', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_2/audiofolder_data_dir_with_metadata/train/audio_file.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_2/audiofolder_data_dir_with_metadata/train/audio_file2.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_2/audiofolder_data_dir_with_metadata/train/metadata.jsonl'])
ERROR tests/packaged_modules/test_audiofolder.py::test_data_files_with_metadata_and_multiple_splits[jsonl-True] - AssertionError: assert 6 == 3
+ where 6 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_3/audiofolder_data_dir_with_metadata/train/audio_file.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_3/audiofolder_data_dir_with_metadata/train/audio_file2.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_3/audiofolder_data_dir_with_metadata/train/metadata.jsonl', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_3/audiofolder_data_dir_with_metadata/train/audio_file.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_3/audiofolder_data_dir_with_metadata/train/audio_file2.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_3/audiofolder_data_dir_with_metadata/train/metadata.jsonl'])
ERROR tests/packaged_modules/test_audiofolder.py::test_data_files_with_metadata_and_multiple_splits[csv-False] - AssertionError: assert 6 == 3
+ where 6 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_4/audiofolder_data_dir_with_metadata/train/audio_file.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_4/audiofolder_data_dir_with_metadata/train/audio_file2.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_4/audiofolder_data_dir_with_metadata/train/metadata.csv', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_4/audiofolder_data_dir_with_metadata/train/audio_file.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_4/audiofolder_data_dir_with_metadata/train/audio_file2.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_4/audiofolder_data_dir_with_metadata/train/metadata.csv'])
ERROR tests/packaged_modules/test_audiofolder.py::test_data_files_with_metadata_and_multiple_splits[csv-True] - AssertionError: assert 6 == 3
+ where 6 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_5/audiofolder_data_dir_with_metadata/train/audio_file.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_5/audiofolder_data_dir_with_metadata/train/audio_file2.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_5/audiofolder_data_dir_with_metadata/train/metadata.csv', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_5/audiofolder_data_dir_with_metadata/train/audio_file.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_5/audiofolder_data_dir_with_metadata/train/audio_file2.wav', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_5/audiofolder_data_dir_with_metadata/train/metadata.csv'])
ERROR tests/packaged_modules/test_folder_based_builder.py::test_data_files_with_metadata_and_splits[1-False] - AssertionError: assert 6 == 3
+ where 6 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_3/autofolder_data_dir_with_metadata_two_splits/train/file.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_3/autofolder_data_dir_with_metadata_two_splits/train/file2.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_3/autofolder_data_dir_with_metadata_two_splits/train/metadata.jsonl', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_3/autofolder_data_dir_with_metadata_two_splits/train/file.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_3/autofolder_data_dir_with_metadata_two_splits/train/file2.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_3/autofolder_data_dir_with_metadata_two_splits/train/metadata.jsonl'])
ERROR tests/packaged_modules/test_folder_based_builder.py::test_data_files_with_metadata_and_splits[1-True] - AssertionError: assert 6 == 3
+ where 6 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_4/autofolder_data_dir_with_metadata_two_splits/train/file.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_4/autofolder_data_dir_with_metadata_two_splits/train/file2.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_4/autofolder_data_dir_with_metadata_two_splits/train/metadata.jsonl', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_4/autofolder_data_dir_with_metadata_two_splits/train/file.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_4/autofolder_data_dir_with_metadata_two_splits/train/file2.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_4/autofolder_data_dir_with_metadata_two_splits/train/metadata.jsonl'])
ERROR tests/packaged_modules/test_folder_based_builder.py::test_data_files_with_metadata_and_splits[2-False] - AssertionError: assert 6 == 3
+ where 6 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_5/autofolder_data_dir_with_metadata_two_splits/train/file.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_5/autofolder_data_dir_with_metadata_two_splits/train/file2.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_5/autofolder_data_dir_with_metadata_two_splits/train/metadata.jsonl', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_5/autofolder_data_dir_with_metadata_two_splits/train/file.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_5/autofolder_data_dir_with_metadata_two_splits/train/file2.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_5/autofolder_data_dir_with_metadata_two_splits/train/metadata.jsonl'])
ERROR tests/packaged_modules/test_imagefolder.py::test_data_files_with_metadata_and_multiple_splits[jsonl-False] - AssertionError: assert 6 == 3
+ where 6 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_12/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_12/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb2.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_12/imagefolder_data_dir_with_metadata_two_splits/train/metadata.jsonl', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_12/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_12/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb2.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_12/imagefolder_data_dir_with_metadata_two_splits/train/metadata.jsonl'])
ERROR tests/packaged_modules/test_imagefolder.py::test_data_files_with_metadata_and_multiple_splits[jsonl-True] - AssertionError: assert 6 == 3
+ where 6 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_13/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_13/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb2.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_13/imagefolder_data_dir_with_metadata_two_splits/train/metadata.jsonl', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_13/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_13/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb2.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_13/imagefolder_data_dir_with_metadata_two_splits/train/metadata.jsonl'])
ERROR tests/packaged_modules/test_folder_based_builder.py::test_data_files_with_metadata_and_splits[2-True] - AssertionError: assert 6 == 3
+ where 6 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_6/autofolder_data_dir_with_metadata_two_splits/train/file.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_6/autofolder_data_dir_with_metadata_two_splits/train/file2.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_6/autofolder_data_dir_with_metadata_two_splits/train/metadata.jsonl', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_6/autofolder_data_dir_with_metadata_two_splits/train/file.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_6/autofolder_data_dir_with_metadata_two_splits/train/file2.txt', '/tmp/pytest-of-runner/pytest-0/popen-gw0/test_data_files_with_metadata_6/autofolder_data_dir_with_metadata_two_splits/train/metadata.jsonl'])
ERROR tests/packaged_modules/test_imagefolder.py::test_data_files_with_metadata_and_multiple_splits[csv-False] - AssertionError: assert 6 == 3
+ where 6 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_14/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_14/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb2.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_14/imagefolder_data_dir_with_metadata_two_splits/train/metadata.csv', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_14/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_14/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb2.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_14/imagefolder_data_dir_with_metadata_two_splits/train/metadata.csv'])
ERROR tests/packaged_modules/test_imagefolder.py::test_data_files_with_metadata_and_multiple_splits[csv-True] - AssertionError: assert 6 == 3
+ where 6 = len(['/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_15/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_15/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb2.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_15/imagefolder_data_dir_with_metadata_two_splits/train/metadata.csv', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_15/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_15/imagefolder_data_dir_with_metadata_two_splits/train/image_rgb2.jpg', '/tmp/pytest-of-runner/pytest-0/popen-gw1/test_data_files_with_metadata_15/imagefolder_data_dir_with_metadata_two_splits/train/metadata.csv'])
= 3 failed, 2383 passed, 26 skipped, 9 warnings, 12 errors in 280.79s (0:04:40) =
``` | 6,209 |
https://github.com/huggingface/datasets/issues/6207 | No-script datasets with ZIP files do not load | [] | While investigating an issue on a Hub dataset, I have discovered the no-script datasets containing ZIP files do not load.
For example, that no-script dataset containing ZIP files, raises NonMatchingSplitsSizesError:
```python
In [2]: ds = load_dataset("sidovic/LearningQ-qg")
NonMatchingSplitsSizesError: [
{
'expected': SplitInfo(name='train', num_bytes=0, num_examples=188660, shard_lengths=None, dataset_name=None),
'recorded': SplitInfo(name='train', num_bytes=0, num_examples=0, shard_lengths=None, dataset_name='learning_q-qg')
}, {
'expected': SplitInfo(name='validation', num_bytes=0, num_examples=20630, shard_lengths=None, dataset_name=None),
'recorded': SplitInfo(name='validation', num_bytes=0, num_examples=0, shard_lengths=None, dataset_name='learning_q-qg')
}, {
'expected': SplitInfo(name='test', num_bytes=0, num_examples=18227, shard_lengths=None, dataset_name=None),
'recorded': SplitInfo(name='test', num_bytes=0, num_examples=0, shard_lengths=None, dataset_name='learning_q-qg')
}
]
```
As another example, a no-script dataset containing just a (CSV)-ZIP file, raises a DatasetGenerationError:
```
> num_examples, num_bytes = writer.finalize()
src/datasets/builder.py:1949:
> raise SchemaInferenceError("Please pass `features` or at least one example when writing data")
E datasets.arrow_writer.SchemaInferenceError: Please pass `features` or at least one example when writing data
src/datasets/arrow_writer.py:598: SchemaInferenceError
The above exception was the direct cause of the following exception:
src/datasets/load.py:2143: in load_dataset
builder_instance.download_and_prepare(
src/datasets/builder.py:954: in download_and_prepare
self._download_and_prepare(
src/datasets/builder.py:1049: in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
src/datasets/builder.py:1813: in _prepare_split
for job_id, done, content in self._prepare_split_single(
> raise DatasetGenerationError("An error occurred while generating the dataset") from e
E datasets.builder.DatasetGenerationError: An error occurred while generating the dataset
src/datasets/builder.py:1958: DatasetGenerationError
```
After investigating, I think this bug was introduced in this PR:
- #5972
Related to:
- https://huggingface.co/datasets/sidovic/LearningQ-qg/discussions/1
CC: @lhoestq | 6,207 |
https://github.com/huggingface/datasets/issues/6206 | When calling load_dataset, raise error: pyarrow.lib.ArrowInvalid: offset overflow while concatenating arrays | [
"I solved the problem by modifying the \"self DEFAULT_WRITER_BATCH_SIZE\" in \"class MyDataset (datasets. GeneratorBasedBuilder) : __init__\"",
"same problem, and this solution worked me also - you can set this var by setting the keyword argument `writer_batch_size=...` in `load_dataset(...,writer_batch_size=...)`"
] | ### Describe the bug
When calling load_dataset, raise error
```
Traceback (most recent call last):
File "/home/aihao/miniconda3/envs/torch/lib/python3.11/site-packages/datasets/builder.py", line 1694, in _pre
pare_split_single
writer.write(example, key)
File "/home/aihao/miniconda3/envs/torch/lib/python3.11/site-packages/datasets/arrow_writer.py", line 490, in
write
self.write_examples_on_file()
File "/home/aihao/miniconda3/envs/torch/lib/python3.11/site-packages/datasets/arrow_writer.py", line 448, in
write_examples_on_file
self.write_batch(batch_examples=batch_examples)
File "/home/aihao/miniconda3/envs/torch/lib/python3.11/site-packages/datasets/arrow_writer.py", line 559, in
write_batch
self.write_table(pa_table, writer_batch_size)
File "/home/aihao/miniconda3/envs/torch/lib/python3.11/site-packages/datasets/arrow_writer.py", line 571, in
write_table
pa_table = pa_table.combine_chunks()
^^^^^^^^^^^^^^^^^^^^^^^^^
File "pyarrow/table.pxi", line 3439, in pyarrow.lib.Table.combine_chunks
File "pyarrow/error.pxi", line 144, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 100, in pyarrow.lib.check_status
pyarrow.lib.ArrowInvalid: offset overflow while concatenating arrays
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
dataset = load_dataset(
^^^^^^^^^^^^^
File "/home/aihao/miniconda3/envs/torch/lib/python3.11/site-packages/datasets/load.py", line 2133, in load_da
taset
builder_instance.download_and_prepare(
File "/home/aihao/miniconda3/envs/torch/lib/python3.11/site-packages/datasets/builder.py", line 954, in downl
oad_and_prepare
self._download_and_prepare(
File "/home/aihao/miniconda3/envs/torch/lib/python3.11/site-packages/datasets/builder.py", line 1717, in _dow
nload_and_prepare
super()._download_and_prepare(
File "/home/aihao/miniconda3/envs/torch/lib/python3.11/site-packages/datasets/builder.py", line 1049, in _dow
nload_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/home/aihao/miniconda3/envs/torch/lib/python3.11/site-packages/datasets/builder.py", line 1555, in _pre
pare_split
for job_id, done, content in self._prepare_split_single(
File "/home/aihao/miniconda3/envs/torch/lib/python3.11/site-packages/datasets/builder.py", line 1712, in _pre
pare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.builder.DatasetGenerationError: An error occurred while generating the dataset
Setting num_proc from 8 back to 1 for the train split to disable multiprocessing as it only contains one shard.
09/04/2023 12:02:04 - WARNING - datasets.builder - Setting num_proc from 8 back to 1 for the train split to dis
able multiprocessing as it only contains one shard.
```
### Steps to reproduce the bug
Call load_dataset with the large image as feature
### Expected behavior
no error
### Environment info
- `datasets` version: 2.14.3
- Platform: Linux-6.2.0-31-generic-x86_64-with-glibc2.35
- Python version: 3.11.4
- Huggingface_hub version: 0.16.4
- PyArrow version: 12.0.1
- Pandas version: 2.0.3 | 6,206 |
https://github.com/huggingface/datasets/issues/6203 | Support loading from a DVC remote repository | [
"(cross-posting from the linked DVC issue)\r\n\r\nI think this should already work out of the box with the current `datasets` and `dvc.api` releases by passing the correct `storage_options` into the datasets calls. `storage_options` is essentially just the kwargs dict that gets passed to the fsspec fs constructor.\r\n\r\nThe main thing to note here is that the fsspec DVCFileSystem URL should be `dvc://folder/file.json` (i.e. this should be the DVCFileSystem path that is relative to the DVC repo root). You cannot use a URL like `https://gitlab.com/user/repo/folder/file.json`.\r\n\r\nI think something like this should work for you (in a venv where both DVC and datasets are installed):\r\n```python\r\nimport datasets\r\n\r\n# load a dataset from Git/DVC repository where Git repo is located at https://gitlab.com/user/repo.git\r\n# and path to dataset (relative to git/dvc repo root) is 'folder/file.json'\r\ndatasets.load_from_disk(\r\n \"dvc://folder/file.json\",\r\n storage_options={\"url\": \"https://gitlab.com/user/repo.git\"},\r\n)\r\n```\r\n\r\nbasically the `dvc://` is what tells fsspec to create a `DVCFileSystem` and it will construct it like\r\n```python\r\nfs = DVCFileSystem(**storage_options)\r\n```\r\n\r\nThen the subsequent calls use the rest of the `dvc://...` URL like \r\n```python\r\nfs.exists(\"folder/file.json\")\r\n```",
"Hi @pmrowla Thank you for your help, that's very helpful, I was indeed using `fsspec` incorrectly here. There is still an issue with `datasets`:\r\n\r\n```python\r\nimport datasets\r\ndataset = datasets.load_dataset(\"json\", data_files=\"dvc://folder/file.jsonl\", storage_options={\"url\": \"https://gitlab.com/repo/folder/\"})\r\n```\r\n\r\nresults in the following exception:\r\n\r\n```\r\nTraceback (most recent call last): \r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/scmrepo/fs.py\", line 217, in info\r\n ret = self.trie.info(key)\r\n ^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/scmrepo/git/objects.py\", line 141, in info\r\n obj = self.trie[key]\r\n ~~~~~~~~~^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/pygtrie.py\", line 937, in __getitem__\r\n node, _ = self._get_node(key_or_slice)\r\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/pygtrie.py\", line 630, in _get_node\r\n raise KeyError(key)\r\nKeyError: ('dvc:', 'datasets', 'spider', 'train.jsonl')\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/datasets/load.py\", line 2129, in load_dataset\r\n builder_instance = load_dataset_builder(\r\n ^^^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/datasets/load.py\", line 1815, in load_dataset_builder\r\n dataset_module = dataset_module_factory(\r\n ^^^^^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/datasets/load.py\", line 1430, in dataset_module_factory\r\n ).get_module()\r\n ^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/datasets/load.py\", line 958, in get_module\r\n data_files = DataFilesDict.from_patterns(\r\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/datasets/data_files.py\", line 674, in from_patterns\r\n DataFilesList.from_patterns(\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/datasets/data_files.py\", line 589, in from_patterns\r\n origin_metadata = _get_origin_metadata(data_files, download_config=download_config)\r\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/datasets/data_files.py\", line 504, in _get_origin_metadata\r\n return thread_map(\r\n ^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/tqdm/contrib/concurrent.py\", line 69, in thread_map\r\n return _executor_map(ThreadPoolExecutor, fn, *iterables, **tqdm_kwargs)\r\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/tqdm/contrib/concurrent.py\", line 51, in _executor_map\r\n return list(tqdm_class(ex.map(fn, *iterables, chunksize=chunksize), **kwargs))\r\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/.pyenv/versions/3.11.4/lib/python3.11/concurrent/futures/_base.py\", line 619, in result_iterator\r\n yield _result_or_cancel(fs.pop())\r\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/.pyenv/versions/3.11.4/lib/python3.11/concurrent/futures/_base.py\", line 317, in _result_or_cancel\r\n return fut.result(timeout)\r\n ^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/.pyenv/versions/3.11.4/lib/python3.11/concurrent/futures/_base.py\", line 456, in result\r\n return self.__get_result()\r\n ^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/.pyenv/versions/3.11.4/lib/python3.11/concurrent/futures/_base.py\", line 401, in __get_result\r\n raise self._exception\r\n File \"/Users/bilelomrani/.pyenv/versions/3.11.4/lib/python3.11/concurrent/futures/thread.py\", line 58, in run\r\n result = self.fn(*self.args, **self.kwargs)\r\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/datasets/data_files.py\", line 491, in _get_single_origin_metadata\r\n info = fs.info(data_file)\r\n ^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/dvc/fs/dvc.py\", line 357, in info\r\n return self._info(key, path, ignore_subrepos=ignore_subrepos)\r\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/dvc/fs/dvc.py\", line 377, in _info\r\n fs_info = fs.info(fs_path)\r\n ^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/dvc_objects/fs/base.py\", line 501, in info\r\n return self.fs.info(path, **kwargs)\r\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r\n File \"/Users/bilelomrani/Documents/ILLUIN.nosync/instructions-finetuning/.venv/lib/python3.11/site-packages/scmrepo/fs.py\", line 221, in info\r\n raise FileNotFoundError(errno.ENOENT, os.strerror(errno.ENOENT), path)\r\nFileNotFoundError: [Errno 2] No such file or directory: '/dvc:/folder/file.jsonl'\r\n```\r\n\r\nSomehow the URL gets turned into `/dvc:/folder/file.jsonl` inside `datasets`. Otherwise I can confirm that using `fsspec` properly with DVC works as expected.\r\n",
"For the record, there was a `dvc.api.DVCFileSystem` bug which is fixed in DVC `main` and will be available in the next DVC release.\r\n\r\nTo use DVC with `datasets` you just need to pass the Git/DVC repo `url` in `storage_options` as discussed above.\r\n\r\n(note that this requires having both `datasets` and `dvc` installed in your python environment)\r\n```python\r\n>>> from datasets import load_dataset\r\n>>> load_dataset(\r\n... \"json\",\r\n... data_files=\"dvc://eval/metrics.json\",\r\n... storage_options={\"url\": \"https://github.com/iterative/example-get-started.git\"},\r\n... )\r\nDatasetDict({\r\n train: Dataset({\r\n features: ['avg_prec', 'roc_auc'],\r\n num_rows: 1\r\n })\r\n})\r\n```\r\n\r\nAny additional `DVCFileSystem` args can be passed in the same way, so to get a specific branch/tag/commit from the DVC repo you just need to specify the `rev` in `storage_options` like\r\n```\r\nstorage_options={\"url\": \"https://github.com/iterative/example-get-started.git\", \"rev\": \"main\"}\r\n```\r\n\r\nI think this issue can probably be closed now.",
"Thank you for your help, closing."
] | ### Feature request
Adding support for loading a file from a DVC repository, tracked remotely on a SCM.
### Motivation
DVC is a popular version control system to version and manage datasets. The files are stored on a remote object storage platform, but they are tracked using Git. Integration with DVC is possible through the `DVCFileSystem`.
I have a Gitlab repository where multiple files are tracked using DVC and stored in a GCP bucket. I would like to be able to load these files using `datasets` directly using an URL. My goal is to write a generic code that abstracts the storage layer, such that my users will only have to pass in an `fsspec`-compliant URL and the corresponding files will be loaded.
### Your contribution
I managed to instantiate a `DVCFileSystem` pointing to a Gitlab repo from a `fsspec` chained URL in [this pull request](https://github.com/iterative/dvc/pull/9903) to DVC.
```python
from fsspec.core import url_to_fs
fs, _ = url_to_fs("dvc::https://gitlab.com/repository/group/my-repo")
```
From now I'm not sure how to continue, it seems that `datasets` expects the URL to be fully qualified like so: `dvc::https://gitlab.com/repository/group/my-repo/my-folder/my-file.json` but this fails because `DVCFileSystem` expects the URL to point to the root of an SCM repo. Is there a way to make this work with `datasets`? | 6,203 |
https://github.com/huggingface/datasets/issues/6202 | avoid downgrading jax version | [
"https://github.com/huggingface/datasets/blob/main/setup.py#L236\r\nCurrently has the highest version at 0.3.25; Not sure if there is any reason for this, other than that was the tested version?"
] | ### Feature request
Whenever I `pip install datasets[jax]` it downgrades jax to version 0.3.25. I seem to be able to install this library first then upgrade jax back to version 0.4.13.
### Motivation
It would be nice to not overwrite currently installed version of jax if possible.
### Your contribution
I would be willing to beta test. Or maybe write some code if I could get pointed in the right direction, I'm not super familiar with this codebase. | 6,202 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.