id
int64 599M
3.26B
| number
int64 1
7.7k
| title
stringlengths 1
290
| body
stringlengths 0
228k
⌀ | state
stringclasses 2
values | html_url
stringlengths 46
51
| created_at
timestamp[s]date 2020-04-14 10:18:02
2025-07-23 08:04:53
| updated_at
timestamp[s]date 2020-04-27 16:04:17
2025-07-23 18:53:44
| closed_at
timestamp[s]date 2020-04-14 12:01:40
2025-07-23 16:44:42
⌀ | user
dict | labels
listlengths 0
4
| is_pull_request
bool 2
classes | comments
listlengths 0
0
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|
1,660,455,202
| 5,725
|
How to limit the number of examples in dataset, for testing?
|
### Describe the bug
I am using this command:
`data = load_dataset("json", data_files=data_path)`
However, I want to add a parameter, to limit the number of loaded examples to be 10, for development purposes, but can't find this simple parameter.
### Steps to reproduce the bug
In the description.
### Expected behavior
To be able to limit the number of examples
### Environment info
Nothing special
|
closed
|
https://github.com/huggingface/datasets/issues/5725
| 2023-04-10T08:41:43
| 2023-04-21T06:16:24
| 2023-04-21T06:16:24
|
{
"login": "ndvbd",
"id": 845175,
"type": "User"
}
|
[] | false
|
[] |
1,659,938,135
| 5,724
|
Error after shuffling streaming IterableDatasets with downloaded dataset
|
### Describe the bug
I downloaded the C4 dataset, and used streaming IterableDatasets to read it. Everything went normal until I used `dataset = dataset.shuffle(seed=42, buffer_size=10_000)` to shuffle the dataset. Shuffled dataset will throw the following error when it is used by `next(iter(dataset))`:
```
File "/data/miniconda3/lib/python3.9/site-packages/datasets/iterable_dataset.py", line 937, in __iter__
for key, example in ex_iterable:
File "/data/miniconda3/lib/python3.9/site-packages/datasets/iterable_dataset.py", line 627, in __iter__
for x in self.ex_iterable:
File "/data/miniconda3/lib/python3.9/site-packages/datasets/iterable_dataset.py", line 138, in __iter__
yield from self.generate_examples_fn(**kwargs_with_shuffled_shards)
File "/data/miniconda3/lib/python3.9/site-packages/datasets/iterable_dataset.py", line 763, in wrapper
for key, table in generate_tables_fn(**kwargs):
File "/data/miniconda3/lib/python3.9/site-packages/datasets/packaged_modules/json/json.py", line 101, in _generate_tables
batch = f.read(self.config.chunksize)
File "/data/miniconda3/lib/python3.9/site-packages/datasets/download/streaming_download_manager.py", line 372, in read_with_retries
out = read(*args, **kwargs)
File "/data/miniconda3/lib/python3.9/gzip.py", line 300, in read
return self._buffer.read(size)
File "/data/miniconda3/lib/python3.9/_compression.py", line 68, in readinto
data = self.read(len(byte_view))
File "/data/miniconda3/lib/python3.9/gzip.py", line 487, in read
if not self._read_gzip_header():
File "/data/miniconda3/lib/python3.9/gzip.py", line 435, in _read_gzip_header
raise BadGzipFile('Not a gzipped file (%r)' % magic)
gzip.BadGzipFile: Not a gzipped file (b've')
```
I found that there is no problem to use the dataset in this way without shuffling. Also, use `dataset = datasets.load_dataset('c4', 'en', split='train', streaming=True)`, which will download the dataset on-the-fly instead of loading from the local file, will also not have problems even after shuffle.
### Steps to reproduce the bug
1. Download C4 dataset from https://huggingface.co/datasets/allenai/c4
2.
```
import datasets
dataset = datasets.load_dataset('/path/to/your/data/dir', 'en', streaming=True, split='train')
dataset = dataset.shuffle(buffer_size=10_000, seed=42)
next(iter(dataset))
```
### Expected behavior
`next(iter(dataset))` should give me a sample from the dataset
### Environment info
- `datasets` version: 2.11.0
- Platform: Linux-5.4.32-1-tlinux4-0001-x86_64-with-glibc2.28
- Python version: 3.9.16
- Huggingface_hub version: 0.13.1
- PyArrow version: 11.0.0
- Pandas version: 1.5.3
|
closed
|
https://github.com/huggingface/datasets/issues/5724
| 2023-04-09T16:58:44
| 2023-04-20T20:37:30
| 2023-04-20T20:37:30
|
{
"login": "szxiangjn",
"id": 41177966,
"type": "User"
}
|
[] | false
|
[] |
1,659,837,510
| 5,722
|
Distributed Training Error on Customized Dataset
|
Hi guys, recently I tried to use `datasets` to train a dual encoder.
I finish my own datasets according to the nice [tutorial](https://huggingface.co/docs/datasets/v2.11.0/en/dataset_script)
Here are my code:
```python
class RetrivalDataset(datasets.GeneratorBasedBuilder):
"""CrossEncoder dataset."""
BUILDER_CONFIGS = [RetrivalConfig(name="DuReader")]
# DEFAULT_CONFIG_NAME = "DuReader"
def _info(self):
return datasets.DatasetInfo(
features=datasets.Features(
{
"id": datasets.Value("string"),
"question": datasets.Value("string"),
"documents": Sequence(datasets.Value("string")),
}
),
supervised_keys=None,
)
def _split_generators(self, dl_manager):
"""Returns SplitGenerators."""
train_file = self.config.data_dir + self.config.train_file
valid_file = self.config.data_dir + self.config.valid_file
logger.info(f"Training on {self.config.train_file}")
logger.info(f"Evaluating on {self.config.valid_file}")
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN, gen_kwargs={"file_path": train_file}
),
datasets.SplitGenerator(
name=datasets.Split.VALIDATION, gen_kwargs={"file_path": valid_file}
),
]
def _generate_examples(self, file_path):
with jsonlines.open(file_path, "r") as f:
for record in f:
label = record["label"]
question = record["question"]
# dual encoder
all_documents = record["all_documents"]
positive_paragraph = all_documents.pop(label)
all_documents = [positive_paragraph] + all_documents
u_id = "{}_#_{}".format(
md5_hash(question + "".join(all_documents)),
"".join(random.sample(string.ascii_letters + string.digits, 7)),
)
item = {
"question": question,
"documents": all_documents,
"id": u_id,
}
yield u_id, item
```
It works well on single GPU, but got errors as follows when used DDP:
```python
Detected mismatch between collectives on ranks. Rank 1 is running collective: CollectiveFingerPrint(OpType=BARRIER), but Rank 0 is running collective: CollectiveFingerPrint(OpType=ALLGATHER_COALESCED)
```
Here are my train script on a two A100 mechine:
```bash
export TORCH_DISTRIBUTED_DEBUG=DETAIL
export TORCH_SHOW_CPP_STACKTRACES=1
export NCCL_DEBUG=INFO
export NCCL_DEBUG_SUBSYS=INIT,COLL,ENV
nohup torchrun --nproc_per_node 2 train.py experiments/de-big.json >logs/de-big.log 2>&1&
```
I am not sure if this error below related to my dataset code when use DDP. And I notice the PR(#5369 ), but I don't know when and where should I used the function(`split_dataset_by_node`) .
@lhoestq hope you could help me?
|
closed
|
https://github.com/huggingface/datasets/issues/5722
| 2023-04-09T11:04:59
| 2023-07-24T14:50:46
| 2023-07-24T14:50:46
|
{
"login": "wlhgtc",
"id": 16603773,
"type": "User"
}
|
[] | false
|
[] |
1,659,680,682
| 5,721
|
Calling datasets.load_dataset("text" ...) results in a wrong split.
|
### Describe the bug
When creating a text dataset, the training split should have the bulk of the examples by default. Currently, testing does.
### Steps to reproduce the bug
I have a folder with 18K text files in it. Each text file essentially consists in a document or article scraped from online. Calling the following codeL
```
folder_path = "/home/cyril/Downloads/llama_dataset"
data = datasets.load_dataset("text", data_dir=folder_path)
data.save_to_disk("/home/cyril/Downloads/data.hf")
data = datasets.load_from_disk("/home/cyril/Downloads/data.hf")
print(data)
```
Results in the following split:
```
DatasetDict({
train: Dataset({
features: ['text'],
num_rows: 2114
})
test: Dataset({
features: ['text'],
num_rows: 200882
})
validation: Dataset({
features: ['text'],
num_rows: 152
})
})
```
It seems to me like the train/test/validation splits are in the wrong order since test split >>>> train_split
### Expected behavior
Train split should have the bulk of the training examples.
### Environment info
datasets 2.11.0, python 3.10.6
|
open
|
https://github.com/huggingface/datasets/issues/5721
| 2023-04-08T23:55:12
| 2023-04-08T23:55:12
| null |
{
"login": "cyrilzakka",
"id": 1841186,
"type": "User"
}
|
[] | false
|
[] |
1,659,610,705
| 5,720
|
Streaming IterableDatasets do not work with torch DataLoaders
|
### Describe the bug
When using streaming datasets set up with train/val split using `.skip()` and `.take()`, the following error occurs when iterating over a torch dataloader:
```
File "/Users/julian/miniconda3/envs/sims/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 363, in __iter__
self._iterator = self._get_iterator()
File "/Users/julian/miniconda3/envs/sims/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 314, in _get_iterator
return _MultiProcessingDataLoaderIter(self)
File "/Users/julian/miniconda3/envs/sims/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 927, in __init__
w.start()
File "/Users/julian/miniconda3/envs/sims/lib/python3.9/multiprocessing/process.py", line 121, in start
self._popen = self._Popen(self)
File "/Users/julian/miniconda3/envs/sims/lib/python3.9/multiprocessing/context.py", line 224, in _Popen
return _default_context.get_context().Process._Popen(process_obj)
File "/Users/julian/miniconda3/envs/sims/lib/python3.9/multiprocessing/context.py", line 284, in _Popen
return Popen(process_obj)
File "/Users/julian/miniconda3/envs/sims/lib/python3.9/multiprocessing/popen_spawn_posix.py", line 32, in __init__
super().__init__(process_obj)
File "/Users/julian/miniconda3/envs/sims/lib/python3.9/multiprocessing/popen_fork.py", line 19, in __init__
self._launch(process_obj)
File "/Users/julian/miniconda3/envs/sims/lib/python3.9/multiprocessing/popen_spawn_posix.py", line 47, in _launch
reduction.dump(process_obj, fp)
File "/Users/julian/miniconda3/envs/sims/lib/python3.9/multiprocessing/reduction.py", line 60, in dump
ForkingPickler(file, protocol).dump(obj)
AttributeError: Can't pickle local object '_generate_examples_from_tables_wrapper.<locals>.wrapper'
```
To reproduce, run the code
```
from datasets import load_dataset
data = load_dataset(args.dataset_name, split="train", streaming=True)
train_len = 5000
val_len = 100
train, val = data.take(train_len), data.skip(train_len).take(val_len)
traindata = IterableClipDataset(data, context_length=args.max_len, tokenizer=tokenizer, image_key="url", text_key="text")
traindata = DataLoader(traindata, batch_size=args.batch_size, num_workers=args.num_workers, persistent_workers=True)
```
Where the class IterableClipDataset is a simple wrapper to cast the dataset to a torch iterabledataset, defined via
```
from torch.utils.data import Dataset, IterableDataset
from torchvision.transforms import Compose, Resize, ToTensor
from transformers import AutoTokenizer
import requests
from PIL import Image
class IterableClipDataset(IterableDataset):
def __init__(self, dataset, context_length: int, image_transform=None, tokenizer=None, image_key="image", text_key="text"):
self.dataset = dataset
self.context_length = context_length
self.image_transform = Compose([Resize((224, 224)), ToTensor()]) if image_transform is None else image_transform
self.tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased") if tokenizer is None else tokenizer
self.image_key = image_key
self.text_key = text_key
def read_image(self, url: str):
try: # Try to read the image
image = Image.open(requests.get(url, stream=True).raw)
except:
image = Image.new("RGB", (224, 224), (0, 0, 0))
return image
def process_sample(self, image, text):
if isinstance(image, str):
image = self.read_image(image)
if self.image_transform is not None:
image = self.image_transform(image)
text = self.tokenizer.encode(
text, add_special_tokens=True, max_length=self.context_length, truncation=True, padding="max_length"
)
text = torch.tensor(text, dtype=torch.long)
return image, text
def __iter__(self):
for sample in self.dataset:
image, text = sample[self.image_key], sample[self.text_key]
yield self.process_sample(image, text)
```
### Steps to reproduce the bug
Steps to reproduce
1. Install `datasets`, `torch`, and `PIL` (if you want to reproduce exactly)
2. Run the code above
### Expected behavior
Batched data is produced from the dataloader
### Environment info
```
datasets == 2.9.0
python == 3.9.12
torch == 1.11.0
```
|
open
|
https://github.com/huggingface/datasets/issues/5720
| 2023-04-08T18:45:48
| 2025-03-19T14:06:47
| null |
{
"login": "jlehrer1",
"id": 29244648,
"type": "User"
}
|
[] | false
|
[] |
1,659,203,222
| 5,719
|
Array2D feature creates a list of list instead of a numpy array
|
### Describe the bug
I'm not sure if this is expected behavior or not. When I create a 2D array using `Array2D`, the data has list type instead of numpy array. I think it should not be the expected behavior especially when I feed a numpy array as input to the data creation function. Why is it converting my array into a list?
Also if I change the first dimension of the `Array2D` shape to None, it's returning array correctly.
### Steps to reproduce the bug
Run this code:
```py
from datasets import Dataset, Features, Array2D
import numpy as np
# you have to change the first dimension of the shape to None to make it return an array
features = Features(dict(seq=Array2D((2,2), 'float32')))
ds = Dataset.from_dict(dict(seq=[np.random.rand(2,2)]), features=features)
a = ds[0]['seq']
print(a)
print(type(a))
```
The following will be printed in stdout:
```
[[0.8127174377441406, 0.3760348856449127], [0.7510159611701965, 0.4322739541530609]]
<class 'list'>
```
### Expected behavior
Each indexed item should be a list or numpy array. Currently, `Array((2,2))` yields a list but `Array((None,2))` yields an array.
### Environment info
- `datasets` version: 2.11.0
- Platform: Windows-10-10.0.19045-SP0
- Python version: 3.9.13
- Huggingface_hub version: 0.13.4
- PyArrow version: 11.0.0
- Pandas version: 1.4.4
|
closed
|
https://github.com/huggingface/datasets/issues/5719
| 2023-04-07T21:04:08
| 2023-04-20T15:34:41
| 2023-04-20T15:34:41
|
{
"login": "offchan42",
"id": 15215732,
"type": "User"
}
|
[] | false
|
[] |
1,658,958,406
| 5,718
|
Reorder default data splits to have validation before test
|
This PR reorders data splits, so that by default validation appears before test.
The default order becomes: [train, validation, test] instead of [train, test, validation].
|
closed
|
https://github.com/huggingface/datasets/pull/5718
| 2023-04-07T16:01:26
| 2023-04-27T14:43:13
| 2023-04-27T14:35:52
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[] | true
|
[] |
1,658,729,866
| 5,717
|
Errror when saving to disk a dataset of images
|
### Describe the bug
Hello!
I have an issue when I try to save on disk my dataset of images. The error I get is:
```
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 1442, in save_to_disk
for job_id, done, content in Dataset._save_to_disk_single(**kwargs):
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/arrow_dataset.py", line 1473, in _save_to_disk_single
writer.write_table(pa_table)
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/arrow_writer.py", line 570, in write_table
pa_table = embed_table_storage(pa_table)
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/table.py", line 2268, in embed_table_storage
arrays = [
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/table.py", line 2269, in <listcomp>
embed_array_storage(table[name], feature) if require_storage_embed(feature) else table[name]
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/table.py", line 1817, in wrapper
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/table.py", line 1817, in <listcomp>
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/table.py", line 2142, in embed_array_storage
return feature.embed_storage(array)
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/features/image.py", line 269, in embed_storage
storage = pa.StructArray.from_arrays([bytes_array, path_array], ["bytes", "path"], mask=bytes_array.is_null())
File "pyarrow/array.pxi", line 2766, in pyarrow.lib.StructArray.from_arrays
File "pyarrow/array.pxi", line 2961, in pyarrow.lib.c_mask_inverted_from_obj
TypeError: Mask must be a pyarrow.Array of type boolean
```
My dataset is around 50K images, is this error might be due to a bad image?
Thanks for the help.
### Steps to reproduce the bug
```python
from datasets import load_dataset
dataset = load_dataset("imagefolder", data_dir="/path/to/dataset")
dataset["train"].save_to_disk("./myds", num_shards=40)
```
### Expected behavior
Having my dataset properly saved to disk.
### Environment info
- `datasets` version: 2.11.0
- Platform: Linux-5.15.90.1-microsoft-standard-WSL2-x86_64-with-glibc2.35
- Python version: 3.10.10
- Huggingface_hub version: 0.13.3
- PyArrow version: 11.0.0
- Pandas version: 2.0.0
|
open
|
https://github.com/huggingface/datasets/issues/5717
| 2023-04-07T11:59:17
| 2025-07-13T08:27:47
| null |
{
"login": "jplu",
"id": 959590,
"type": "User"
}
|
[] | false
|
[] |
1,658,613,092
| 5,716
|
Handle empty audio
|
Some audio paths exist, but they are empty, and an error will be reported when reading the audio path.How to use the filter function to avoid the empty audio path?
when a audio is empty, when do resample , it will break:
`array, sampling_rate = sf.read(f) array = librosa.resample(array, orig_sr=sampling_rate, target_sr=self.sampling_rate)`
|
closed
|
https://github.com/huggingface/datasets/issues/5716
| 2023-04-07T09:51:40
| 2023-09-27T17:47:08
| 2023-09-27T17:47:08
|
{
"login": "ben-8878",
"id": 38179632,
"type": "User"
}
|
[] | false
|
[] |
1,657,479,788
| 5,715
|
Return Numpy Array (fixed length) Mode, in __get_item__, Instead of List
|
### Feature request
There are old known issues, but they can be easily forgettable problems in multiprocessing with pytorch-dataloader:
Too high usage of RAM or shared-memory in pytorch when we set num workers > 1 and returning type of dataset or dataloader is "List" or "Dict".
https://github.com/pytorch/pytorch/issues/13246
With huggingface datasets, unfortunately, the default return type is the list, so the problem is raised too often if we do not set anything for the issue.
However, this issue can be released when the returning output is fixed in length.
Therefore, I request the mode, returning outputs with fixed length (e.g. numpy array) rather than list.
The design would be good when we load datasets as
```python
load_dataset(..., with_return_as_fixed_tensor=True)
```
### Motivation
The general solution for this issue is already in the comments: https://github.com/pytorch/pytorch/issues/13246#issuecomment-905703662
: Numpy or Pandas seems not to have problems, while both have the string type.
(I'm not sure that the sequence of huggingface datasets can solve this problem as well)
### Your contribution
I'll read it ! thanks
|
closed
|
https://github.com/huggingface/datasets/issues/5715
| 2023-04-06T13:57:48
| 2023-04-20T17:16:26
| 2023-04-20T17:16:26
|
{
"login": "jungbaepark",
"id": 34066771,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
}
] | false
|
[] |
1,657,388,033
| 5,714
|
Fix xnumpy_load for .npz files
|
PR:
- #5626
implemented support for streaming `.npy` files by using `numpy.load`.
However, it introduced a bug when used with `.npz` files, within a context manager:
```
ValueError: seek of closed file
```
or in streaming mode:
```
ValueError: I/O operation on closed file.
```
This PR fixes the bug and tests for both `.npy` and `.npz` files.
Fix #5711.
|
closed
|
https://github.com/huggingface/datasets/pull/5714
| 2023-04-06T13:01:45
| 2023-04-07T09:23:54
| 2023-04-07T09:16:57
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[] | true
|
[] |
1,657,141,251
| 5,713
|
ArrowNotImplementedError when loading dataset from the hub
|
### Describe the bug
Hello,
I have created a dataset by using the image loader. Once the dataset is created I try to download it and I get the error:
```
Traceback (most recent call last):
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/builder.py", line 1860, in _prepare_split_single
for _, table in generator:
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/packaged_modules/parquet/parquet.py", line 69, in _generate_tables
for batch_idx, record_batch in enumerate(
File "pyarrow/_parquet.pyx", line 1323, in iter_batches
File "pyarrow/error.pxi", line 121, in pyarrow.lib.check_status
pyarrow.lib.ArrowNotImplementedError: Nested data conversions not implemented for chunked array outputs
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/load.py", line 1791, in load_dataset
builder_instance.download_and_prepare(
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/builder.py", line 891, in download_and_prepare
self._download_and_prepare(
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/builder.py", line 986, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/builder.py", line 1748, in _prepare_split
for job_id, done, content in self._prepare_split_single(
File "/home/jplu/miniconda3/envs/image-xp/lib/python3.10/site-packages/datasets/builder.py", line 1893, 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
```
### Steps to reproduce the bug
Create the dataset and push it to the hub:
```python
from datasets import load_dataset
dataset = load_dataset("imagefolder", data_dir="/path/to/dataset")
dataset.push_to_hub("org/dataset-name", private=True, max_shard_size="1GB")
```
Then use it:
```python
from datasets import load_dataset
dataset = load_dataset("org/dataset-name")
```
### Expected behavior
To properly download and use the pushed dataset.
Something else to note is that I specified to have shards of 1GB max, but at the end, for the train set, it is an almost 7GB single file that is pushed.
### Environment info
- `datasets` version: 2.11.0
- Platform: Linux-5.15.90.1-microsoft-standard-WSL2-x86_64-with-glibc2.35
- Python version: 3.10.10
- Huggingface_hub version: 0.13.3
- PyArrow version: 11.0.0
- Pandas version: 2.0.0
|
closed
|
https://github.com/huggingface/datasets/issues/5713
| 2023-04-06T10:27:22
| 2023-04-06T13:06:22
| 2023-04-06T13:06:21
|
{
"login": "jplu",
"id": 959590,
"type": "User"
}
|
[] | false
|
[] |
1,655,972,106
| 5,712
|
load_dataset in v2.11.0 raises "ValueError: seek of closed file" in np.load()
|
### Describe the bug
Hi,
I have some `dataset_load()` code of a custom offline dataset that works with datasets v2.10.1.
```python
ds = datasets.load_dataset(path=dataset_dir,
name=configuration,
data_dir=dataset_dir,
cache_dir=cache_dir,
aux_dir=aux_dir,
# download_mode=datasets.DownloadMode.FORCE_REDOWNLOAD,
num_proc=18)
```
When upgrading datasets to 2.11.0, it fails with error
```
Traceback (most recent call last):
File "<string>", line 2, in <module>
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/site-packages/datasets/load.py", line 1791, in load_dataset
builder_instance.download_and_prepare(
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/site-packages/datasets/builder.py", line 891, in download_and_prepare
self._download_and_prepare(
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/site-packages/datasets/builder.py", line 1651, in _download_and_prepare
super()._download_and_prepare(
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/site-packages/datasets/builder.py", line 964, in _download_and_prepare
split_generators = self._split_generators(dl_manager, **split_generators_kwargs)
File "/home/ramon.casero/.cache/huggingface/modules/datasets_modules/datasets/71f67f69e6e00e139903a121f96b71f39b65a6b6aaeb0862e6a5da3a3f565b4c/mydataset.py", line 682, in _split_generators
self.some_function()
File "/home/ramon.casero/.cache/huggingface/modules/datasets_modules/datasets/71f67f69e6e00e139903a121f96b71f39b65a6b6aaeb0862e6a5da3a3f565b4c/mydataset.py", line 1314, in some_function()
x_df = pd.DataFrame({'cell_type_descriptor': fp['x'].tolist()})
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/site-packages/numpy/lib/npyio.py", line 248, in __getitem__
bytes = self.zip.open(key)
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/zipfile.py", line 1530, in open
fheader = zef_file.read(sizeFileHeader)
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/zipfile.py", line 744, in read
self._file.seek(self._pos)
ValueError: seek of closed file
```
### Steps to reproduce the bug
Sorry, I cannot share the data or code because they are not mine to share, but the point of failure is a call in `some_function()`
```python
with np.load(filename) as fp:
x_df = pd.DataFrame({'feature': fp['x'].tolist()})
```
I'll try to generate a short snippet that reproduces the error.
### Expected behavior
I would expect that `load_dataset` works on the custom datasets generation script for v2.11.0 the same way it works for 2.10.1, without making `np.load()` give a `ValueError: seek of closed file` error.
### Environment info
- `datasets` version: 2.11.0
- Platform: Linux-4.18.0-483.el8.x86_64-x86_64-with-glibc2.28
- Python version: 3.10.8
- Huggingface_hub version: 0.12.0
- PyArrow version: 11.0.0
- Pandas version: 1.5.2
- numpy: 1.24.2
- This is an offline dataset that uses `datasets.config.HF_DATASETS_OFFLINE = True` in the generation script.
|
closed
|
https://github.com/huggingface/datasets/issues/5712
| 2023-04-05T16:47:10
| 2023-04-06T08:32:37
| 2023-04-05T17:17:44
|
{
"login": "rcasero",
"id": 1219084,
"type": "User"
}
|
[] | false
|
[] |
1,655,971,647
| 5,711
|
load_dataset in v2.11.0 raises "ValueError: seek of closed file" in np.load()
|
### Describe the bug
Hi,
I have some `dataset_load()` code of a custom offline dataset that works with datasets v2.10.1.
```python
ds = datasets.load_dataset(path=dataset_dir,
name=configuration,
data_dir=dataset_dir,
cache_dir=cache_dir,
aux_dir=aux_dir,
# download_mode=datasets.DownloadMode.FORCE_REDOWNLOAD,
num_proc=18)
```
When upgrading datasets to 2.11.0, it fails with error
```
Traceback (most recent call last):
File "<string>", line 2, in <module>
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/site-packages/datasets/load.py", line 1791, in load_dataset
builder_instance.download_and_prepare(
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/site-packages/datasets/builder.py", line 891, in download_and_prepare
self._download_and_prepare(
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/site-packages/datasets/builder.py", line 1651, in _download_and_prepare
super()._download_and_prepare(
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/site-packages/datasets/builder.py", line 964, in _download_and_prepare
split_generators = self._split_generators(dl_manager, **split_generators_kwargs)
File "/home/ramon.casero/.cache/huggingface/modules/datasets_modules/datasets/71f67f69e6e00e139903a121f96b71f39b65a6b6aaeb0862e6a5da3a3f565b4c/mydataset.py", line 682, in _split_generators
self.some_function()
File "/home/ramon.casero/.cache/huggingface/modules/datasets_modules/datasets/71f67f69e6e00e139903a121f96b71f39b65a6b6aaeb0862e6a5da3a3f565b4c/mydataset.py", line 1314, in some_function()
x_df = pd.DataFrame({'cell_type_descriptor': fp['x'].tolist()})
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/site-packages/numpy/lib/npyio.py", line 248, in __getitem__
bytes = self.zip.open(key)
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/zipfile.py", line 1530, in open
fheader = zef_file.read(sizeFileHeader)
File "/home/ramon.casero/opt/miniconda3/envs/myenv/lib/python3.10/zipfile.py", line 744, in read
self._file.seek(self._pos)
ValueError: seek of closed file
```
### Steps to reproduce the bug
Sorry, I cannot share the data or code because they are not mine to share, but the point of failure is a call in `some_function()`
```python
with np.load(embedding_filename) as fp:
x_df = pd.DataFrame({'feature': fp['x'].tolist()})
```
I'll try to generate a short snippet that reproduces the error.
### Expected behavior
I would expect that `load_dataset` works on the custom datasets generation script for v2.11.0 the same way it works for 2.10.1, without making `np.load()` give a `ValueError: seek of closed file` error.
### Environment info
- `datasets` version: 2.11.0
- Platform: Linux-4.18.0-483.el8.x86_64-x86_64-with-glibc2.28
- Python version: 3.10.8
- Huggingface_hub version: 0.12.0
- PyArrow version: 11.0.0
- Pandas version: 1.5.2
- numpy: 1.24.2
- This is an offline dataset that uses `datasets.config.HF_DATASETS_OFFLINE = True` in the generation script.
|
closed
|
https://github.com/huggingface/datasets/issues/5711
| 2023-04-05T16:46:49
| 2023-04-07T09:16:59
| 2023-04-07T09:16:59
|
{
"login": "rcasero",
"id": 1219084,
"type": "User"
}
|
[] | false
|
[] |
1,655,703,534
| 5,710
|
OSError: Memory mapping file failed: Cannot allocate memory
|
### Describe the bug
Hello, I have a series of datasets each of 5 GB, 600 datasets in total. So together this makes 3TB.
When I trying to load all the 600 datasets into memory, I get the above error message.
Is this normal because I'm hitting the max size of memory mapping of the OS?
Thank you
```terminal
0_21/cache-e9c42499f65b1881.arrow
load_hf_datasets_from_disk: 82%|████████████████████████████████████████████████████████████████████████████████████████████████████▍ | 494/600 [07:26<01:35, 1.11it/s]
Traceback (most recent call last):
File "example_load_genkalm_dataset.py", line 35, in <module>
multi_ds.post_process(max_node_num=args.max_node_num,max_seq_length=args.max_seq_length,delay=args.delay)
File "/home/geng/GenKaLM/src/dataloader/dataset.py", line 142, in post_process
genkalm_dataset = GenKaLM_Dataset.from_hf_dataset(path_or_name=ds_path, max_seq_length=self.max_seq_length,
File "/home/geng/GenKaLM/src/dataloader/dataset.py", line 47, in from_hf_dataset
hf_ds = load_from_disk(path_or_name)
File "/home/geng/.conda/envs/genkalm/lib/python3.8/site-packages/datasets/load.py", line 1848, in load_from_disk
return Dataset.load_from_disk(dataset_path, keep_in_memory=keep_in_memory, storage_options=storage_options)
File "/home/geng/.conda/envs/genkalm/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 1549, in load_from_disk
arrow_table = concat_tables(
File "/home/geng/.conda/envs/genkalm/lib/python3.8/site-packages/datasets/table.py", line 1805, in concat_tables
tables = list(tables)
File "/home/geng/.conda/envs/genkalm/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 1550, in <genexpr>
table_cls.from_file(Path(dataset_path, data_file["filename"]).as_posix())
File "/home/geng/.conda/envs/genkalm/lib/python3.8/site-packages/datasets/table.py", line 1065, in from_file
table = _memory_mapped_arrow_table_from_file(filename)
File "/home/geng/.conda/envs/genkalm/lib/python3.8/site-packages/datasets/table.py", line 50, in _memory_mapped_arrow_table_from_file
memory_mapped_stream = pa.memory_map(filename)
File "pyarrow/io.pxi", line 950, in pyarrow.lib.memory_map
File "pyarrow/io.pxi", line 911, in pyarrow.lib.MemoryMappedFile._open
File "pyarrow/error.pxi", line 144, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 115, in pyarrow.lib.check_status
OSError: Memory mapping file failed: Cannot allocate memory
```
### Steps to reproduce the bug
Sorry I can not provide a reproducible code as the data is stored on my server and it's too large to share.
### Expected behavior
I expect the 3TB of data can be fully mapped to memory
### Environment info
- `datasets` version: 2.9.0
- Platform: Linux-4.15.0-204-generic-x86_64-with-debian-buster-sid
- Python version: 3.7.6
- PyArrow version: 11.0.0
- Pandas version: 1.0.1
|
closed
|
https://github.com/huggingface/datasets/issues/5710
| 2023-04-05T14:11:26
| 2023-04-20T17:16:40
| 2023-04-20T17:16:40
|
{
"login": "Saibo-creator",
"id": 53392976,
"type": "User"
}
|
[] | false
|
[] |
1,655,423,503
| 5,709
|
Manually dataset info made not taken into account
|
### Describe the bug
Hello,
I'm manually building an image dataset with the `from_dict` approach. I also build the features with the `cast_features` methods. Once the dataset is created I push it on the hub, and a default `dataset_infos.json` file seems to have been automatically added to the repo in same time. Hence I update it manually with all the missing info, but when I download the dataset the info are never updated.
Former `dataset_infos.json` file:
```
{"default": {
"description": "",
"citation": "",
"homepage": "",
"license": "",
"features": {
"image": {
"_type": "Image"
},
"labels": {
"names": [
"Fake",
"Real"
],
"_type": "ClassLabel"
}
},
"splits": {
"validation": {
"name": "validation",
"num_bytes": 901010094.0,
"num_examples": 3200,
"dataset_name": null
},
"train": {
"name": "train",
"num_bytes": 901010094.0,
"num_examples": 3200,
"dataset_name": null
}
},
"download_size": 1802008414,
"dataset_size": 1802020188.0,
"size_in_bytes": 3604028602.0
}}
```
After I update it manually it looks like:
```
{
"bstrai--deepfake-detection":{
"description":"",
"citation":"",
"homepage":"",
"license":"",
"features":{
"image":{
"decode":true,
"id":null,
"_type":"Image"
},
"labels":{
"num_classes":2,
"names":[
"Fake",
"Real"
],
"id":null,
"_type":"ClassLabel"
}
},
"supervised_keys":{
"input":"image",
"output":"labels"
},
"task_templates":[
{
"task":"image-classification",
"image_column":"image",
"label_column":"labels"
}
],
"config_name":null,
"splits":{
"validation":{
"name":"validation",
"num_bytes":36627822,
"num_examples":123,
"dataset_name":"deepfake-detection"
},
"train":{
"name":"train",
"num_bytes":901023694,
"num_examples":3200,
"dataset_name":"deepfake-detection"
}
},
"download_checksums":null,
"download_size":937562209,
"dataset_size":937651516,
"size_in_bytes":1875213725
}
}
```
Anything I should do to have the new infos in the `dataset_infos.json` to be taken into account? Or it is not possible yet?
Thanks!
### Steps to reproduce the bug
-
### Expected behavior
-
### Environment info
- `datasets` version: 2.11.0
- Platform: Linux-5.15.90.1-microsoft-standard-WSL2-x86_64-with-glibc2.35
- Python version: 3.10.10
- Huggingface_hub version: 0.13.3
- PyArrow version: 11.0.0
- Pandas version: 2.0.0
|
closed
|
https://github.com/huggingface/datasets/issues/5709
| 2023-04-05T11:15:17
| 2023-04-06T08:52:20
| 2023-04-06T08:52:19
|
{
"login": "jplu",
"id": 959590,
"type": "User"
}
|
[] | false
|
[] |
1,655,023,642
| 5,708
|
Dataset sizes are in MiB instead of MB in dataset cards
|
As @severo reported in an internal discussion (https://github.com/huggingface/moon-landing/issues/5929):
Now we show the dataset size:
- from the dataset card (in the side column)
- from the datasets-server (in the viewer)
But, even if the size is the same, we see a mismatch because the viewer shows MB, while the info from the README generally shows MiB (even if it's written MB -> https://huggingface.co/datasets/blimp/blob/main/README.md?code=true#L1932)
<img width="664" alt="Capture d’écran 2023-04-04 à 10 16 01" src="https://user-images.githubusercontent.com/1676121/229730887-0bd8fa6e-9462-46c6-bd4e-4d2c5784cabb.png">
TODO: Values to be fixed in: `Size of downloaded dataset files:`, `Size of the generated dataset:` and `Total amount of disk used:`
- [x] Bulk edit on the Hub to fix this in all canonical datasets
- [x] Bulk PR on the Hub to fix ancient canonical datasets that were moved to organizations
|
closed
|
https://github.com/huggingface/datasets/issues/5708
| 2023-04-05T06:36:03
| 2023-12-21T10:20:28
| 2023-12-21T10:20:27
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[
{
"name": "bug",
"color": "d73a4a"
},
{
"name": "dataset-viewer",
"color": "E5583E"
}
] | false
|
[] |
1,653,545,835
| 5,706
|
Support categorical data types for Parquet
|
### Feature request
Huggingface datasets does not seem to support categorical / dictionary data types for Parquet as of now. There seems to be a `TODO` in the code for this feature but no implementation yet. Below you can find sample code to reproduce the error that is currently thrown when attempting to read a Parquet file with categorical columns:
```python
import pandas as pd
import pyarrow.parquet as pq
from datasets import load_dataset
# Create categorical sample DataFrame
df = pd.DataFrame({'type': ['foo', 'bar']}).astype('category')
df.to_parquet('data.parquet')
# Read back as pyarrow table
table = pq.read_table('data.parquet')
print(table.schema)
# type: dictionary<values=string, indices=int32, ordered=0>
# Load with huggingface datasets
load_dataset('parquet', data_files='data.parquet')
```
Error:
```
Traceback (most recent call last):
File ".venv/lib/python3.10/site-packages/datasets/builder.py", line 1875, in _prepare_split_single
writer.write_table(table)
File ".venv/lib/python3.10/site-packages/datasets/arrow_writer.py", line 566, in write_table
self._build_writer(inferred_schema=pa_table.schema)
File ".venv/lib/python3.10/site-packages/datasets/arrow_writer.py", line 379, in _build_writer
inferred_features = Features.from_arrow_schema(inferred_schema)
File ".venv/lib/python3.10/site-packages/datasets/features/features.py", line 1622, in from_arrow_schema
obj = {field.name: generate_from_arrow_type(field.type) for field in pa_schema}
File ".venv/lib/python3.10/site-packages/datasets/features/features.py", line 1622, in <dictcomp>
obj = {field.name: generate_from_arrow_type(field.type) for field in pa_schema}
File ".venv/lib/python3.10/site-packages/datasets/features/features.py", line 1361, in generate_from_arrow_type
raise NotImplementedError # TODO(thom) this will need access to the dictionary as well (for labels). I.e. to the py_table
NotImplementedError
```
### Motivation
Categorical data types, as offered by Pandas and implemented with the `DictionaryType` dtype in `pyarrow` can significantly reduce dataset size and are a handy way to turn textual features into numerical representations and back. Lack of support in Huggingface datasets greatly reduces compatibility with a common Pandas / Parquet feature.
### Your contribution
I could provide a PR. However, it would be nice to have an initial complexity estimate from one of the core developers first.
|
closed
|
https://github.com/huggingface/datasets/issues/5706
| 2023-04-04T09:45:35
| 2024-06-07T12:20:43
| 2024-06-07T12:20:43
|
{
"login": "kklemon",
"id": 1430243,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
}
] | false
|
[] |
1,653,500,383
| 5,705
|
Getting next item from IterableDataset took forever.
|
### Describe the bug
I have a large dataset, about 500GB. The format of the dataset is parquet.
I then load the dataset and try to get the first item
```python
def get_one_item():
dataset = load_dataset("path/to/datafiles", split="train", cache_dir=".", streaming=True)
dataset = dataset.filter(lambda example: example['text'].startswith('Ar'))
print(next(iter(dataset)))
```
However, this function never finish. I waited ~10mins, the function was still running so I killed the process. I'm now using `line_profiler` to profile how long it would take to return one item. I'll be patient and wait for as long as it needs.
I suspect the filter operation is the reason why it took so long. Can I get some possible reasons behind this?
### Steps to reproduce the bug
Unfortunately without my data files, there is no way to reproduce this bug.
### Expected behavior
With `IteralbeDataset`, I expect the first item to be returned instantly.
### Environment info
- datasets version: 2.11.0
- python: 3.7.12
|
closed
|
https://github.com/huggingface/datasets/issues/5705
| 2023-04-04T09:16:17
| 2023-04-05T23:35:41
| 2023-04-05T23:35:41
|
{
"login": "HongtaoYang",
"id": 16588434,
"type": "User"
}
|
[] | false
|
[] |
1,653,471,356
| 5,704
|
5537 speedup load
|
I reimplemented fsspec.spec.glob() in `hffilesystem.py` as `_glob`, used it in `_resolve_single_pattern_in_dataset_repository` only, and saw a 20% speedup in times to load the config, on average.
That's not much when usually this step takes only 2-3 seconds for most datasets, but in this particular case, `bigcode/the-stack-dedup` , the loading time to get the config (not download the entire 6tb dataset, of course), went from ~170 secs to ~20 secs.
What makes this work is this code in `_glob`:
```
if self.dir_cache is not None:
allpaths = self.dir_cache
else:
allpaths = self.find(root, maxdepth=depth, withdirs=True, detail=True, **kwargs)
```
I also had to `import glob.has_magic( )` for `_glob()` (confusing, I know).
I hope there is no issue with copying most of the code from `fsspec.spec.glob`, as it is a BSD 3-Clause License,
and I left a comment about this in the docstring of` _glob()`, that we may want to delete.
As mentioned, I evaluated the speedup across a random selection of about 1000 datasets (not all 27k+), and verified that old_config.eq(new_method_config) with the build in method, but deleted this test and related code changes on the subsequent commit. It's in the commit history if anyone wants to see it. (Note this does not include the outlier of `bigcode/the-stack-dedup`
| | old_time | new _time | diff | pct_diff |
| -- | -- | -- | -- | -- |
| mean | 3.340 | 2.642 | 0.698 | 18.404 |
| min | 2.024 | 1.976 | -0.840 | -37.634 |
| max | 66.582 | 41.517 | 30.927 | 85.538 |
|
open
|
https://github.com/huggingface/datasets/pull/5704
| 2023-04-04T08:58:14
| 2023-04-07T16:10:55
| null |
{
"login": "semajyllek",
"id": 35013374,
"type": "User"
}
|
[] | true
|
[] |
1,653,158,955
| 5,703
|
[WIP][Test, Please ignore] Investigate performance impact of using multiprocessing only
| null |
closed
|
https://github.com/huggingface/datasets/pull/5703
| 2023-04-04T04:37:49
| 2023-04-20T03:17:37
| 2023-04-20T03:17:32
|
{
"login": "hvaara",
"id": 1535968,
"type": "User"
}
|
[] | true
|
[] |
1,653,104,720
| 5,702
|
Is it possible or how to define a `datasets.Sequence` that could potentially be either a dict, a str, or None?
|
### Feature request
Hello! Apologies if my question sounds naive:
I was wondering if it’s possible, or how one would go about defining a 'datasets.Sequence' element in datasets.Features that could potentially be either a dict, a str, or None?
Specifically, I’d like to define a feature for a list that contains 18 elements, each of which has been pre-defined as either a `dict or None` or `str or None` - as demonstrated in the slightly misaligned data provided below:
```json
[
[
{"text":"老妇人","idxes":[0,1,2]},null,{"text":"跪","idxes":[3]},null,null,null,null,{"text":"在那坑里","idxes":[4,5,6,7]},null,null,null,null,null,null,null,null,null,null],
[
{"text":"那些水","idxes":[13,14,15]},null,{"text":"舀","idxes":[11]},null,null,null,null,null,{"text":"在那坑里","idxes":[4,5,6,7]},null,{"text":"出","idxes":[12]},null,null,null,null,null,null,null],
[
{"text":"水","idxes":[38]},
null,
{"text":"舀","idxes":[40]},
"假", // note this is just a standalone string
null,null,null,{"text":"坑里","idxes":[35,36]},null,null,null,null,null,null,null,null,null,null]]
```
### Motivation
I'm currently working with a dataset of the following structure and I couldn't find a solution in the [documentation](https://huggingface.co/docs/datasets/v2.11.0/en/package_reference/main_classes#datasets.Features).
```json
{"qid":"3-train-1058","context":"桑桑害怕了。从玉米地里走到田埂上，他遥望着他家那幢草房子里的灯光，知道母亲没有让他回家的意思，很伤感，有点想哭。但没哭，转身朝阿恕家走去。","corefs":[[{"text":"桑桑","idxes":[0,1]},{"text":"他","idxes":[17]}]],"non_corefs":[],"outputs":[[{"text":"他","idxes":[17]},null,{"text":"走","idxes":[11]},null,null,null,null,null,{"text":"从玉米地里","idxes":[6,7,8,9,10]},{"text":"到田埂上","idxes":[12,13,14,15]},null,null,null,null,null,null,null,null],[{"text":"他","idxes":[17]},null,{"text":"走","idxes":[66]},null,null,null,null,null,null,null,{"text":"转身朝阿恕家去","idxes":[60,61,62,63,64,65,67]},null,null,null,null,null,null,null],[{"text":"灯光","idxes":[30,31]},null,null,null,null,null,null,{"text":"草房子里","idxes":[25,26,27,28]},null,null,null,null,null,null,null,null,null,null],[{"text":"他","idxes":[17]},{"text":"他家那幢草房子","idxes":[21,22,23,24,25,26,27]},null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,"远"],[{"text":"他","idxes":[17]},{"text":"阿恕家","idxes":[63,64,65]},null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,"变近"]]}
```
### Your contribution
I'm going to provide the dataset at https://huggingface.co/datasets/2030NLP/SpaCE2022 .
|
closed
|
https://github.com/huggingface/datasets/issues/5702
| 2023-04-04T03:20:43
| 2023-04-05T14:15:18
| 2023-04-05T14:15:17
|
{
"login": "gitforziio",
"id": 10508116,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
}
] | false
|
[] |
1,652,931,399
| 5,701
|
Add Dataset.from_spark
|
Adds static method Dataset.from_spark to create datasets from Spark DataFrames.
This approach alleviates users of the need to materialize their dataframe---a common use case is that the user loads their dataset into a dataframe, uses Spark to apply some transformation to some of the columns, and then wants to train on the dataset.
Related issue: https://github.com/huggingface/datasets/issues/5678
|
closed
|
https://github.com/huggingface/datasets/pull/5701
| 2023-04-03T23:51:29
| 2023-06-16T16:39:32
| 2023-04-26T15:43:39
|
{
"login": "maddiedawson",
"id": 106995444,
"type": "User"
}
|
[] | true
|
[] |
1,652,527,530
| 5,700
|
fix: fix wrong modification of the 'cache_file_name' -related paramet…
|
…ers values in 'train_test_split' + fix bad interaction between 'keep_in_memory' and 'cache_file_name' -related parameters (#5699)
|
open
|
https://github.com/huggingface/datasets/pull/5700
| 2023-04-03T18:05:26
| 2023-04-06T17:17:27
| null |
{
"login": "FrancoisNoyez",
"id": 47528215,
"type": "User"
}
|
[] | true
|
[] |
1,652,437,419
| 5,699
|
Issue when wanting to split in memory a cached dataset
|
### Describe the bug
**In the 'train_test_split' method of the Dataset class** (defined datasets/arrow_dataset.py), **if 'self.cache_files' is not empty**, then, **regarding the input parameters 'train_indices_cache_file_name' and 'test_indices_cache_file_name', if they are None**, we modify them to make them not None, to see if we can just provide back / work from cached data. But if we can't provide cached data, we move on with the call to the method, except those two values are not None anymore, which will conflict with the use of the 'keep_in_memory' parameter down the line.
Indeed, at some point we end up calling the 'select' method, **and if 'keep_in_memory' is True**, since the value of this method's parameter 'indices_cache_file_name' is now not None anymore, **an exception is raised, whose message is "Please use either 'keep_in_memory' or 'indices_cache_file_name' but not both.".**
Because of that, it's impossible to perform a train / test split of a cached dataset while requesting that the result not be cached. Which is inconvenient when one is just performing experiments, with no intention of caching the result.
Aside from this being inconvenient, **the code which lead up to that situation seems simply wrong** to me: the input variable should not be modified so as to change the user's intention just to perform a test, if that test can fail and respecting the user's intention is necessary to proceed in that case.
To fix this, I suggest to use other variables / other variable names, in order to host the value(s) needed to perform the test, so as not to change the originally input values needed by the rest of the method's code.
Also, **I don't see why an exception should be raised when the 'select' method is called with both 'keep_in_memory'=True and 'indices_cache_file_name'!=None**: should the use of 'keep_in_memory' not prevail anyway, specifying that the user does not want to perform caching, and so making irrelevant the value of 'indices_cache_file_name'? This is indeed what happens when we look further in the code, in the '\_select_with_indices_mapping' method: when 'keep_in_memory' is True, then the value of indices_cache_file_name does not matter, the data will be written to a stream buffer anyway.
Hence I suggest to remove the raising of exception in those circumstances. Notably, to remove the raising of it in the 'select', '\_select_with_indices_mapping', 'shuffle' and 'map' methods.
### Steps to reproduce the bug
```python
import datasets
def generate_examples():
for i in range(10):
yield {"id": i}
dataset_ = datasets.Dataset.from_generator(
generate_examples,
keep_in_memory=False,
)
dataset_.train_test_split(
test_size=3,
shuffle=False,
keep_in_memory=True,
train_indices_cache_file_name=None,
test_indices_cache_file_name=None,
)
```
### Expected behavior
The result of the above code should be a DatasetDict instance.
Instead, we get the following exception stack:
```python
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Cell In[3], line 1
----> 1 dataset_.train_test_split(
2 test_size=3,
3 shuffle=False,
4 keep_in_memory=True,
5 train_indices_cache_file_name=None,
6 test_indices_cache_file_name=None,
7 )
File ~/Work/Developments/datasets/src/datasets/arrow_dataset.py:528, in transmit_format.<locals>.wrapper(*args, **kwargs)
521 self_format = {
522 "type": self._format_type,
523 "format_kwargs": self._format_kwargs,
524 "columns": self._format_columns,
525 "output_all_columns": self._output_all_columns,
526 }
527 # apply actual function
--> 528 out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
529 datasets: List["Dataset"] = list(out.values()) if isinstance(out, dict) else [out]
530 # re-apply format to the output
File ~/Work/Developments/datasets/src/datasets/fingerprint.py:511, in fingerprint_transform.<locals>._fingerprint.<locals>.wrapper(*args, **kwargs)
507 validate_fingerprint(kwargs[fingerprint_name])
509 # Call actual function
--> 511 out = func(dataset, *args, **kwargs)
513 # Update fingerprint of in-place transforms + update in-place history of transforms
515 if inplace: # update after calling func so that the fingerprint doesn't change if the function fails
File ~/Work/Developments/datasets/src/datasets/arrow_dataset.py:4428, in Dataset.train_test_split(self, test_size, train_size, shuffle, stratify_by_column, seed, generator, keep_in_memory, load_from_cache_file, train_indices_cache_file_name, test_indices_cache_file_name, writer_batch_size, train_new_fingerprint, test_new_fingerprint)
4425 test_indices = permutation[:n_test]
4426 train_indices = permutation[n_test : (n_test + n_train)]
-> 4428 train_split = self.select(
4429 indices=train_indices,
4430 keep_in_memory=keep_in_memory,
4431 indices_cache_file_name=train_indices_cache_file_name,
4432 writer_batch_size=writer_batch_size,
4433 new_fingerprint=train_new_fingerprint,
4434 )
4435 test_split = self.select(
4436 indices=test_indices,
4437 keep_in_memory=keep_in_memory,
(...)
4440 new_fingerprint=test_new_fingerprint,
4441 )
4443 return DatasetDict({"train": train_split, "test": test_split})
File ~/Work/Developments/datasets/src/datasets/arrow_dataset.py:528, in transmit_format.<locals>.wrapper(*args, **kwargs)
521 self_format = {
522 "type": self._format_type,
523 "format_kwargs": self._format_kwargs,
524 "columns": self._format_columns,
525 "output_all_columns": self._output_all_columns,
526 }
527 # apply actual function
--> 528 out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
529 datasets: List["Dataset"] = list(out.values()) if isinstance(out, dict) else [out]
530 # re-apply format to the output
File ~/Work/Developments/datasets/src/datasets/fingerprint.py:511, in fingerprint_transform.<locals>._fingerprint.<locals>.wrapper(*args, **kwargs)
507 validate_fingerprint(kwargs[fingerprint_name])
509 # Call actual function
--> 511 out = func(dataset, *args, **kwargs)
513 # Update fingerprint of in-place transforms + update in-place history of transforms
515 if inplace: # update after calling func so that the fingerprint doesn't change if the function fails
File ~/Work/Developments/datasets/src/datasets/arrow_dataset.py:3679, in Dataset.select(self, indices, keep_in_memory, indices_cache_file_name, writer_batch_size, new_fingerprint)
3645 """Create a new dataset with rows selected following the list/array of indices.
3646
3647 Args:
(...)
3676 ```
3677 """
3678 if keep_in_memory and indices_cache_file_name is not None:
-> 3679 raise ValueError("Please use either `keep_in_memory` or `indices_cache_file_name` but not both.")
3681 if len(self.list_indexes()) > 0:
3682 raise DatasetTransformationNotAllowedError(
3683 "Using `.select` on a dataset with attached indexes is not allowed. You can first run `.drop_index() to remove your index and then re-add it."
3684 )
ValueError: Please use either `keep_in_memory` or `indices_cache_file_name` but not both.
```
### Environment info
- `datasets` version: 2.11.1.dev0
- Platform: Linux-5.4.236-1-MANJARO-x86_64-with-glibc2.2.5
- Python version: 3.8.12
- Huggingface_hub version: 0.13.3
- PyArrow version: 11.0.0
- Pandas version: 2.0.0
***
***
EDIT:
Now with a pull request to fix this [here](https://github.com/huggingface/datasets/pull/5700)
|
open
|
https://github.com/huggingface/datasets/issues/5699
| 2023-04-03T17:00:07
| 2024-05-15T13:12:18
| null |
{
"login": "FrancoisNoyez",
"id": 47528215,
"type": "User"
}
|
[] | false
|
[] |
1,652,183,611
| 5,698
|
Add Qdrant as another search index
|
### Feature request
I'd suggest adding Qdrant (https://qdrant.tech) as another search index available, so users can directly build an index from a dataset. Currently, FAISS and ElasticSearch are only supported: https://huggingface.co/docs/datasets/faiss_es
### Motivation
ElasticSearch is a keyword-based search system, while FAISS is a vector search library. Vector database, such as Qdrant, is a different tool based on similarity (like FAISS) but is not limited to a single machine. It makes the vector database well-suited for bigger datasets and collaboration if several people want to access a particular dataset.
### Your contribution
I can provide a PR implementing that functionality on my own.
|
open
|
https://github.com/huggingface/datasets/issues/5698
| 2023-04-03T14:25:19
| 2023-04-11T10:28:40
| null |
{
"login": "kacperlukawski",
"id": 2649301,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
}
] | false
|
[] |
1,651,812,614
| 5,697
|
Raise an error on missing distributed seed
|
close https://github.com/huggingface/datasets/issues/5696
|
closed
|
https://github.com/huggingface/datasets/pull/5697
| 2023-04-03T10:44:58
| 2023-04-04T15:05:24
| 2023-04-04T14:58:16
|
{
"login": "lhoestq",
"id": 42851186,
"type": "User"
}
|
[] | true
|
[] |
1,651,707,008
| 5,696
|
Shuffle a sharded iterable dataset without seed can lead to duplicate data
|
As reported in https://github.com/huggingface/datasets/issues/5360
If `seed=None` in `.shuffle()`, shuffled datasets don't use the same shuffling seed across nodes.
Because of that, the lists of shards is not shuffled the same way across nodes, and therefore some shards may be assigned to multiple nodes instead of exactly one.
This can happen only when you have a number of shards that is a factor of the number of nodes.
The current workaround is to always set a `seed` in `.shuffle()`
|
closed
|
https://github.com/huggingface/datasets/issues/5696
| 2023-04-03T09:40:03
| 2023-04-04T14:58:18
| 2023-04-04T14:58:18
|
{
"login": "lhoestq",
"id": 42851186,
"type": "User"
}
|
[
{
"name": "bug",
"color": "d73a4a"
}
] | false
|
[] |
1,650,974,156
| 5,695
|
Loading big dataset raises pyarrow.lib.ArrowNotImplementedError
|
### Describe the bug
Calling `datasets.load_dataset` to load the (publicly available) dataset `theodor1289/wit` fails with `pyarrow.lib.ArrowNotImplementedError`.
### Steps to reproduce the bug
Steps to reproduce this behavior:
1. `!pip install datasets`
2. `!huggingface-cli login`
3. This step will throw the error (it might take a while as the dataset has ~170GB):
```python
from datasets import load_dataset
dataset = load_dataset("theodor1289/wit", "train", use_auth_token=True)
```
Stack trace:
```
(torch-multimodal) bash-4.2$ python test.py
Downloading and preparing dataset None/None to /cluster/work/cotterell/tamariucai/HuggingfaceDatasets/theodor1289___parquet/theodor1289--wit-7a3e984414a86a0f/0.0.0/2a3b91fbd88a2c90d1dbbb32b460cf621d31bd5b05b934492fdef7d8d6f236ec...
Downloading data files: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 491.68it/s]
Extracting data files: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 16.93it/s]
Traceback (most recent call last):
File "/cluster/home/tamariucai/.local/lib/python3.10/site-packages/datasets/builder.py", line 1860, in _prepare_split_single
for _, table in generator:
File "/cluster/home/tamariucai/.local/lib/python3.10/site-packages/datasets/packaged_modules/parquet/parquet.py", line 69, in _generate_tables
for batch_idx, record_batch in enumerate(
File "pyarrow/_parquet.pyx", line 1323, in iter_batches
File "pyarrow/error.pxi", line 121, in pyarrow.lib.check_status
pyarrow.lib.ArrowNotImplementedError: Nested data conversions not implemented for chunked array outputs
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/cluster/work/cotterell/tamariucai/multimodal-mirror/examples/test.py", line 2, in <module>
dataset = load_dataset("theodor1289/wit", "train", use_auth_token=True)
File "/cluster/home/tamariucai/.local/lib/python3.10/site-packages/datasets/load.py", line 1791, in load_dataset
builder_instance.download_and_prepare(
File "/cluster/home/tamariucai/.local/lib/python3.10/site-packages/datasets/builder.py", line 891, in download_and_prepare
self._download_and_prepare(
File "/cluster/home/tamariucai/.local/lib/python3.10/site-packages/datasets/builder.py", line 986, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/cluster/home/tamariucai/.local/lib/python3.10/site-packages/datasets/builder.py", line 1748, in _prepare_split
for job_id, done, content in self._prepare_split_single(
File "/cluster/home/tamariucai/.local/lib/python3.10/site-packages/datasets/builder.py", line 1893, 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
The dataset is loaded in variable `dataset`.
### Environment info
- `datasets` version: 2.11.0
- Platform: Linux-3.10.0-1160.80.1.el7.x86_64-x86_64-with-glibc2.17
- Python version: 3.10.4
- Huggingface_hub version: 0.13.3
- PyArrow version: 11.0.0
- Pandas version: 1.5.3
|
closed
|
https://github.com/huggingface/datasets/issues/5695
| 2023-04-02T14:42:44
| 2024-05-15T12:04:47
| 2023-04-10T08:04:04
|
{
"login": "amariucaitheodor",
"id": 32778667,
"type": "User"
}
|
[] | false
|
[] |
1,650,467,793
| 5,694
|
Dataset configuration
|
Following discussions from https://github.com/huggingface/datasets/pull/5331
We could have something like `config.json` to define the configuration of a dataset.
```json
{
"data_dir": "data"
"data_files": {
"train": "train-[0-9][0-9][0-9][0-9]-of-[0-9][0-9][0-9][0-9][0-9]*.*"
}
}
```
we could also support a list for several configs with a 'config_name' field.
The alternative was to use YAML in the README.md.
I think it could also support a `dataset_type` field to specify which dataset builder class to use, and the other parameters would be the builder's parameters. Some parameters exist for all builders like `data_files` and `data_dir`, but some parameters are builder specific like `sep` for csv.
This format would be used in `push_to_hub` to be able to push multiple configs.
cc @huggingface/datasets
EDIT: actually we're going for the YAML approach in README.md
|
open
|
https://github.com/huggingface/datasets/issues/5694
| 2023-04-01T13:08:05
| 2023-04-04T14:54:37
| null |
{
"login": "lhoestq",
"id": 42851186,
"type": "User"
}
|
[
{
"name": "generic discussion",
"color": "c5def5"
}
] | false
|
[] |
1,649,934,749
| 5,693
|
[docs] Split pattern search order
|
This PR addresses #5681 about the order of split patterns 🤗 Datasets searches for when generating dataset splits.
|
closed
|
https://github.com/huggingface/datasets/pull/5693
| 2023-03-31T19:51:38
| 2023-04-03T18:43:30
| 2023-04-03T18:29:58
|
{
"login": "stevhliu",
"id": 59462357,
"type": "User"
}
|
[] | true
|
[] |
1,649,818,644
| 5,692
|
pyarrow.lib.ArrowInvalid: Unable to merge: Field <field> has incompatible types
|
### Describe the bug
When loading the dataset [wikianc-en](https://huggingface.co/datasets/cyanic-selkie/wikianc-en) which I created using [this](https://github.com/cyanic-selkie/wikianc) code, I get the following error:
```
Traceback (most recent call last):
File "/home/sven/code/rector/answer-detection/train.py", line 106, in <module>
(dataset, weights) = get_dataset(args.dataset, tokenizer, labels, args.padding)
File "/home/sven/code/rector/answer-detection/dataset.py", line 106, in get_dataset
dataset = load_dataset("cyanic-selkie/wikianc-en")
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/load.py", line 1794, in load_dataset
ds = builder_instance.as_dataset(split=split, verification_mode=verification_mode, in_memory=keep_in_memory)
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/builder.py", line 1106, in as_dataset
datasets = map_nested(
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/utils/py_utils.py", line 443, in map_nested
mapped = [
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/utils/py_utils.py", line 444, in <listcomp>
_single_map_nested((function, obj, types, None, True, None))
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/utils/py_utils.py", line 346, in _single_map_nested
return function(data_struct)
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/builder.py", line 1136, in _build_single_dataset
ds = self._as_dataset(
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/builder.py", line 1207, in _as_dataset
dataset_kwargs = ArrowReader(cache_dir, self.info).read(
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/arrow_reader.py", line 239, in read
return self.read_files(files=files, original_instructions=instructions, in_memory=in_memory)
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/arrow_reader.py", line 260, in read_files
pa_table = self._read_files(files, in_memory=in_memory)
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/arrow_reader.py", line 203, in _read_files
pa_table = concat_tables(pa_tables) if len(pa_tables) != 1 else pa_tables[0]
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/table.py", line 1808, in concat_tables
return ConcatenationTable.from_tables(tables, axis=axis)
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/table.py", line 1514, in from_tables
return cls.from_blocks(blocks)
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/table.py", line 1427, in from_blocks
table = cls._concat_blocks(blocks, axis=0)
File "/home/sven/.cache/pypoetry/virtualenvs/rector-Z2mdKRnn-py3.10/lib/python3.10/site-packages/datasets/table.py", line 1373, in _concat_blocks
return pa.concat_tables(pa_tables, promote=True)
File "pyarrow/table.pxi", line 5224, in pyarrow.lib.concat_tables
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: Unable to merge: Field paragraph_anchors has incompatible types: list<: struct<start: uint32 not null, end: uint32 not null, qid: uint32, pageid: uint32, title: string not null> not null> vs list<item: struct<start: uint32, end: uint32, qid: uint32, pageid: uint32, title: string>>
```
This only happens when I load the `train` split, indicating that the size of the dataset is the deciding factor.
### Steps to reproduce the bug
```python
from datasets import load_dataset
dataset = load_dataset("cyanic-selkie/wikianc-en", split="train")
```
### Expected behavior
The dataset should load normally without any errors.
### Environment info
- `datasets` version: 2.10.1
- Platform: Linux-6.2.8-arch1-1-x86_64-with-glibc2.37
- Python version: 3.10.10
- PyArrow version: 11.0.0
- Pandas version: 1.5.3
|
open
|
https://github.com/huggingface/datasets/issues/5692
| 2023-03-31T18:19:40
| 2024-01-14T07:24:21
| null |
{
"login": "cyanic-selkie",
"id": 32219669,
"type": "User"
}
|
[] | false
|
[] |
1,649,737,526
| 5,691
|
[docs] Compress data files
|
This PR addresses the comments in #5687 about compressing text file extensions before uploading to the Hub. Also clarified what "too large" means based on the GitLFS [docs](https://docs.github.com/en/repositories/working-with-files/managing-large-files/about-git-large-file-storage).
|
closed
|
https://github.com/huggingface/datasets/pull/5691
| 2023-03-31T17:17:26
| 2023-04-19T13:37:32
| 2023-04-19T07:25:58
|
{
"login": "stevhliu",
"id": 59462357,
"type": "User"
}
|
[] | true
|
[] |
1,648,956,349
| 5,689
|
Support streaming Beam datasets from HF GCS preprocessed data
|
This PR implements streaming Apache Beam datasets that are already preprocessed by us and stored in the HF Google Cloud Storage:
- natural_questions
- wiki40b
- wikipedia
This is done by streaming from the prepared Arrow files in HF Google Cloud Storage.
This will fix their corresponding dataset viewers. Related to:
- https://github.com/huggingface/datasets-server/pull/988#discussion_r1150767138
Related to:
- https://huggingface.co/datasets/natural_questions/discussions/4
- https://huggingface.co/datasets/wiki40b/discussions/2
- https://huggingface.co/datasets/wikipedia/discussions/9
CC: @severo
|
closed
|
https://github.com/huggingface/datasets/pull/5689
| 2023-03-31T08:44:24
| 2023-04-12T05:57:55
| 2023-04-12T05:50:31
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[] | true
|
[] |
1,649,289,883
| 5,690
|
raise AttributeError(f"No {package_name} attribute {name}") AttributeError: No huggingface_hub attribute hf_api
|
### Describe the bug
rta.sh
Traceback (most recent call last):
File "run.py", line 7, in <module>
import datasets
File "/home/appuser/miniconda3/envs/pt2/lib/python3.8/site-packages/datasets/__init__.py", line 37, in <module>
from .builder import ArrowBasedBuilder, BeamBasedBuilder, BuilderConfig, DatasetBuilder, GeneratorBasedBuilder
File "/home/appuser/miniconda3/envs/pt2/lib/python3.8/site-packages/datasets/builder.py", line 44, in <module>
from .data_files import DataFilesDict, _sanitize_patterns
File "/home/appuser/miniconda3/envs/pt2/lib/python3.8/site-packages/datasets/data_files.py", line 120, in <module>
dataset_info: huggingface_hub.hf_api.DatasetInfo,
File "/home/appuser/miniconda3/envs/pt2/lib/python3.8/site-packages/huggingface_hub/__init__.py", line 290, in __getattr__
raise AttributeError(f"No {package_name} attribute {name}")
AttributeError: No huggingface_hub attribute hf_api
### Reproduction
_No response_
### Logs
```shell
Traceback (most recent call last):
File "run.py", line 7, in <module>
import datasets
File "/home/appuser/miniconda3/envs/pt2/lib/python3.8/site-packages/datasets/__init__.py", line 37, in <module>
from .builder import ArrowBasedBuilder, BeamBasedBuilder, BuilderConfig, DatasetBuilder, GeneratorBasedBuilder
File "/home/appuser/miniconda3/envs/pt2/lib/python3.8/site-packages/datasets/builder.py", line 44, in <module>
from .data_files import DataFilesDict, _sanitize_patterns
File "/home/appuser/miniconda3/envs/pt2/lib/python3.8/site-packages/datasets/data_files.py", line 120, in <module>
dataset_info: huggingface_hub.hf_api.DatasetInfo,
File "/home/appuser/miniconda3/envs/pt2/lib/python3.8/site-packages/huggingface_hub/__init__.py", line 290, in __getattr__
raise AttributeError(f"No {package_name} attribute {name}")
AttributeError: No huggingface_hub attribute hf_api
```
### System info
```shell
- huggingface_hub version: 0.13.2
- Platform: Linux-5.4.0-144-generic-x86_64-with-glibc2.10
- Python version: 3.8.5
- Running in iPython ?: No
- Running in notebook ?: No
- Running in Google Colab ?: No
- Token path ?: /home/appuser/.cache/huggingface/token
- Has saved token ?: False
- Configured git credential helpers:
- FastAI: N/A
- Tensorflow: N/A
- Torch: 1.7.1
- Jinja2: N/A
- Graphviz: N/A
- Pydot: N/A
- Pillow: 9.3.0
- hf_transfer: N/A
- ENDPOINT: https://huggingface.co
- HUGGINGFACE_HUB_CACHE: /home/appuser/.cache/huggingface/hub
- HUGGINGFACE_ASSETS_CACHE: /home/appuser/.cache/huggingface/assets
- HF_TOKEN_PATH: /home/appuser/.cache/huggingface/token
- HF_HUB_OFFLINE: False
- HF_HUB_DISABLE_TELEMETRY: False
- HF_HUB_DISABLE_PROGRESS_BARS: None
- HF_HUB_DISABLE_SYMLINKS_WARNING: False
- HF_HUB_DISABLE_IMPLICIT_TOKEN: False
```
|
closed
|
https://github.com/huggingface/datasets/issues/5690
| 2023-03-31T08:22:22
| 2023-07-21T14:21:57
| 2023-07-21T14:21:57
|
{
"login": "wccccp",
"id": 55964850,
"type": "User"
}
|
[
{
"name": "bug",
"color": "d73a4a"
}
] | false
|
[] |
1,648,463,504
| 5,688
|
Wikipedia download_and_prepare for GCS
|
### Describe the bug
I am unable to download the wikipedia dataset onto GCS.
When I run the script provided the memory firstly gets eaten up, then it crashes.
I tried running this on a VM with 128GB RAM and all I got was a two empty files: _data_builder.lock_, _data.incomplete/beam-temp-wikipedia-train-1ab2039acf3611ed87a9893475de0093_
I have troubleshot this for two straight days now, but I am just unable to get the dataset into storage.
### Steps to reproduce the bug
Run this and insert a path:
```
import datasets
builder = datasets.load_dataset_builder(
"wikipedia", language="en", date="20230320", beam_runner="DirectRunner")
builder.download_and_prepare({path}, file_format="parquet")
```
This is where the problem of it eating RAM occurs.
I have also tried several versions of this, based on the docs:
```
import gcsfs
import datasets
storage_options = {"project": "tdt4310", "token": "cloud"}
fs = gcsfs.GCSFileSystem(**storage_options)
output_dir = "gcs://wikipediadata/"
builder = datasets.load_dataset_builder(
"wikipedia", date="20230320", language="en", beam_runner="DirectRunner")
builder.download_and_prepare(
output_dir, storage_options=storage_options, file_format="parquet")
```
The error message that is received here is:
> ValueError: Unable to get filesystem from specified path, please use the correct path or ensure the required dependency is installed, e.g., pip install apache-beam[gcp]. Path specified: gcs://wikipediadata/wikipedia-train [while running 'train/Save to parquet/Write/WriteImpl/InitializeWrite']
I have ran `pip install apache-beam[gcp]`
### Expected behavior
The wikipedia data loaded into GCS
Everything worked when testing with a smaller demo dataset found somewhere in the docs
### Environment info
Newest published version of datasets. Python 3.9. Also tested with Python 3.7. 128GB RAM Google Cloud VM instance.
|
closed
|
https://github.com/huggingface/datasets/issues/5688
| 2023-03-30T23:43:22
| 2024-03-15T15:59:18
| 2024-03-15T15:59:18
|
{
"login": "adrianfagerland",
"id": 25522531,
"type": "User"
}
|
[] | false
|
[] |
1,647,009,018
| 5,687
|
Document to compress data files before uploading
|
In our docs to [Share a dataset to the Hub](https://huggingface.co/docs/datasets/upload_dataset), we tell users to upload directly their data files, like CSV, JSON, JSON-Lines, text,... However, these extensions are not tracked by Git LFS by default, as they are not in the `.giattributes` file. Therefore, if they are too large, Git will fail to commit/upload them.
I think for those file extensions (.csv, .json, .jsonl, .txt), we should better recommend to **compress** their data files (using ZIP for example) before uploading them to the Hub.
- Compressed files are tracked by Git LFS in our default `.gitattributes` file
What do you think?
CC: @stevhliu
See related issue:
- https://huggingface.co/datasets/tcor0005/langchain-docs-400-chunksize/discussions/1
|
closed
|
https://github.com/huggingface/datasets/issues/5687
| 2023-03-30T06:41:07
| 2023-04-19T07:25:59
| 2023-04-19T07:25:59
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[
{
"name": "documentation",
"color": "0075ca"
}
] | false
|
[] |
1,646,308,228
| 5,686
|
set dev version
| null |
closed
|
https://github.com/huggingface/datasets/pull/5686
| 2023-03-29T18:24:13
| 2023-03-29T18:33:49
| 2023-03-29T18:24:22
|
{
"login": "lhoestq",
"id": 42851186,
"type": "User"
}
|
[] | true
|
[] |
1,646,048,667
| 5,685
|
Broken Image render on the hub website
|
### Describe the bug
Hi :wave:
Not sure if this is the right place to ask, but I am trying to load a huge amount of datasets on the hub (:partying_face: ) but I am facing a little issue with the `image` type
See this [dataset](https://huggingface.co/datasets/Francesco/cell-towers), basically for some reason the first image has numerical bytes inside, not sure if that is okay, but the image render feature **doesn't work**
So the dataset is stored in the following way
```python
builder.download_and_prepare(output_dir=str(output_dir))
ds = builder.as_dataset(split="train")
# [NOTE] no idea how to push it from the builder folder
ds.push_to_hub(repo_id=repo_id)
builder.as_dataset(split="validation").push_to_hub(repo_id=repo_id)
ds = builder.as_dataset(split="test")
ds.push_to_hub(repo_id=repo_id)
```
The build is this class
```python
class COCOLikeDatasetBuilder(datasets.GeneratorBasedBuilder):
VERSION = datasets.Version("1.0.0")
def _info(self):
features = datasets.Features(
{
"image_id": datasets.Value("int64"),
"image": datasets.Image(),
"width": datasets.Value("int32"),
"height": datasets.Value("int32"),
"objects": datasets.Sequence(
{
"id": datasets.Value("int64"),
"area": datasets.Value("int64"),
"bbox": datasets.Sequence(
datasets.Value("float32"), length=4
),
"category": datasets.ClassLabel(names=categories),
}
),
}
)
return datasets.DatasetInfo(
description=description,
features=features,
homepage=homepage,
license=license,
citation=citation,
)
def _split_generators(self, dl_manager):
archive = dl_manager.download(url)
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN,
gen_kwargs={
"annotation_file_path": "train/_annotations.coco.json",
"files": dl_manager.iter_archive(archive),
},
),
datasets.SplitGenerator(
name=datasets.Split.VALIDATION,
gen_kwargs={
"annotation_file_path": "test/_annotations.coco.json",
"files": dl_manager.iter_archive(archive),
},
),
datasets.SplitGenerator(
name=datasets.Split.TEST,
gen_kwargs={
"annotation_file_path": "valid/_annotations.coco.json",
"files": dl_manager.iter_archive(archive),
},
),
]
def _generate_examples(self, annotation_file_path, files):
def process_annot(annot, category_id_to_category):
return {
"id": annot["id"],
"area": annot["area"],
"bbox": annot["bbox"],
"category": category_id_to_category[annot["category_id"]],
}
image_id_to_image = {}
idx = 0
# This loop relies on the ordering of the files in the archive:
# Annotation files come first, then the images.
for path, f in files:
file_name = os.path.basename(path)
if annotation_file_path in path:
annotations = json.load(f)
category_id_to_category = {
category["id"]: category["name"]
for category in annotations["categories"]
}
print(category_id_to_category)
image_id_to_annotations = collections.defaultdict(list)
for annot in annotations["annotations"]:
image_id_to_annotations[annot["image_id"]].append(annot)
image_id_to_image = {
annot["file_name"]: annot for annot in annotations["images"]
}
elif file_name in image_id_to_image:
image = image_id_to_image[file_name]
objects = [
process_annot(annot, category_id_to_category)
for annot in image_id_to_annotations[image["id"]]
]
print(file_name)
yield idx, {
"image_id": image["id"],
"image": {"path": path, "bytes": f.read()},
"width": image["width"],
"height": image["height"],
"objects": objects,
}
idx += 1
```
Basically, I want to add to the hub every dataset I come across on coco format
Thanks
Fra
### Steps to reproduce the bug
In this case, you can just navigate on the [dataset](https://huggingface.co/datasets/Francesco/cell-towers)
### Expected behavior
I was expecting the image rendering feature to work
### Environment info
Not a lot to share, I am using `datasets` from a fresh venv
|
closed
|
https://github.com/huggingface/datasets/issues/5685
| 2023-03-29T15:25:30
| 2023-03-30T07:54:25
| 2023-03-30T07:54:25
|
{
"login": "FrancescoSaverioZuppichini",
"id": 15908060,
"type": "User"
}
|
[] | false
|
[] |
1,646,013,226
| 5,684
|
Release: 2.11.0
| null |
closed
|
https://github.com/huggingface/datasets/pull/5684
| 2023-03-29T15:06:07
| 2023-03-29T18:30:34
| 2023-03-29T18:15:54
|
{
"login": "lhoestq",
"id": 42851186,
"type": "User"
}
|
[] | true
|
[] |
1,646,001,197
| 5,683
|
Fix verification_mode when ignore_verifications is passed
|
This PR fixes the values assigned to `verification_mode` when passing `ignore_verifications` to `load_dataset`.
Related to:
- #5303
Fix #5682.
|
closed
|
https://github.com/huggingface/datasets/pull/5683
| 2023-03-29T15:00:50
| 2023-03-29T17:36:06
| 2023-03-29T17:28:57
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[] | true
|
[] |
1,646,000,571
| 5,682
|
ValueError when passing ignore_verifications
|
When passing `ignore_verifications=True` to `load_dataset`, we get a ValueError:
```
ValueError: 'none' is not a valid VerificationMode
```
|
closed
|
https://github.com/huggingface/datasets/issues/5682
| 2023-03-29T15:00:30
| 2023-03-29T17:28:58
| 2023-03-29T17:28:58
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[
{
"name": "bug",
"color": "d73a4a"
}
] | false
|
[] |
1,645,630,784
| 5,681
|
Add information about patterns search order to the doc about structuring repo
|
Following [this](https://github.com/huggingface/datasets/issues/5650) issue I think we should add a note about the order of patterns that is used to find splits, see [my comment](https://github.com/huggingface/datasets/issues/5650#issuecomment-1488412527). Also we should reference this page in pages about packaged loaders.
I have a déjà vu that it had already been discussed as some point but I don't remember....
|
closed
|
https://github.com/huggingface/datasets/issues/5681
| 2023-03-29T11:44:49
| 2023-04-03T18:31:11
| 2023-04-03T18:31:11
|
{
"login": "polinaeterna",
"id": 16348744,
"type": "User"
}
|
[
{
"name": "documentation",
"color": "0075ca"
}
] | false
|
[] |
1,645,430,103
| 5,680
|
Fix a description error for interleave_datasets.
|
There is a description mistake in the annotation of interleave_dataset with "all_exhausted" stopping_strategy.
``` python
d1 = Dataset.from_dict({"a": [0, 1, 2]})
d2 = Dataset.from_dict({"a": [10, 11, 12, 13]})
d3 = Dataset.from_dict({"a": [20, 21, 22, 23, 24]})
dataset = interleave_datasets([d1, d2, d3], stopping_strategy="all_exhausted")
```
According to the interleave way, the correct output of `dataset["a"]` is `[0, 10, 20, 1, 11, 21, 2, 12, 22, 0, 13, 23, 1, 10, 24]`, not `[0, 10, 20, 1, 11, 21, 2, 12, 22, 0, 13, 23, 1, 0, 24]`
|
closed
|
https://github.com/huggingface/datasets/pull/5680
| 2023-03-29T09:50:23
| 2023-03-30T13:14:19
| 2023-03-30T13:07:18
|
{
"login": "QizhiPei",
"id": 55624066,
"type": "User"
}
|
[] | true
|
[] |
1,645,184,622
| 5,679
|
Allow load_dataset to take a working dir for intermediate data
|
### Feature request
As a user, I can set a working dir for intermediate data creation. The processed files will be moved to the cache dir, like
```
load_dataset(…, working_dir=”/temp/dir”, cache_dir=”/cloud_dir”).
```
### Motivation
This will help the use case for using datasets with cloud storage as cache. It will help boost the performance.
### Your contribution
I can provide a PR to fix this if the proposal seems reasonable.
|
open
|
https://github.com/huggingface/datasets/issues/5679
| 2023-03-29T07:21:09
| 2023-04-12T22:30:25
| null |
{
"login": "lu-wang-dl",
"id": 38018689,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
}
] | false
|
[] |
1,645,018,359
| 5,678
|
Add support to create a Dataset from spark dataframe
|
### Feature request
Add a new API `Dataset.from_spark` to create a Dataset from Spark DataFrame.
### Motivation
Spark is a distributed computing framework that can handle large datasets. By supporting loading Spark DataFrames directly into Hugging Face Datasets, we enable take the advantages of spark to processing the data in parallel.
By providing a seamless integration between these two frameworks, we make it easier for data scientists and developers to work with both Spark and Hugging Face in the same workflow.
### Your contribution
We can discuss about the ideas and I can help preparing a PR for this feature.
|
closed
|
https://github.com/huggingface/datasets/issues/5678
| 2023-03-29T04:36:28
| 2024-08-27T14:43:19
| 2023-07-21T14:15:38
|
{
"login": "lu-wang-dl",
"id": 38018689,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
}
] | false
|
[] |
1,644,828,606
| 5,677
|
Dataset.map() crashes when any column contains more than 1000 empty dictionaries
|
### Describe the bug
`Dataset.map()` crashes any time any column contains more than `writer_batch_size` (default 1000) empty dictionaries, regardless of whether the column is being operated on. The error does not occur if the dictionaries are non-empty.
### Steps to reproduce the bug
Example:
```
import datasets
def add_one(example):
example["col2"] += 1
return example
n = 1001 # crashes
# n = 999 # works
ds = datasets.Dataset.from_dict({"col1": [{}] * n, "col2": [1] * n})
ds = ds.map(add_one, writer_batch_size=1000)
```
### Expected behavior
Above code should not crash
### Environment info
- `datasets` version: 2.10.1
- Platform: Linux-5.4.0-120-generic-x86_64-with-glibc2.10
- Python version: 3.8.15
- PyArrow version: 9.0.0
- Pandas version: 1.5.3
|
closed
|
https://github.com/huggingface/datasets/issues/5677
| 2023-03-29T00:01:31
| 2023-07-07T14:01:14
| 2023-07-07T14:01:14
|
{
"login": "mtoles",
"id": 7139344,
"type": "User"
}
|
[] | false
|
[] |
1,641,763,478
| 5,675
|
Filter datasets by language code
|
Hi! I use the language search field on https://huggingface.co/datasets
However, some of the datasets tagged by ISO language code are not accessible by this search form.
For example, [myv_ru_2022](https://huggingface.co/datasets/slone/myv_ru_2022) is has `myv` language tag but it is not included in Languages search form.
I've also noticed the same problem with `mhr` (see https://huggingface.co/datasets/AigizK/mari-russian-parallel-corpora)
|
closed
|
https://github.com/huggingface/datasets/issues/5675
| 2023-03-27T09:42:28
| 2023-03-30T08:08:15
| 2023-03-30T08:08:15
|
{
"login": "named-entity",
"id": 5658496,
"type": "User"
}
|
[] | false
|
[] |
1,641,084,105
| 5,674
|
Stored XSS
|
x
|
closed
|
https://github.com/huggingface/datasets/issues/5674
| 2023-03-26T20:55:58
| 2024-04-30T22:56:41
| 2023-03-27T21:01:55
|
{
"login": "Fadavvi",
"id": 21213484,
"type": "User"
}
|
[] | false
|
[] |
1,641,066,352
| 5,673
|
Pass down storage options
|
Remove implementation-specific kwargs from `file_utils.fsspec_get` and `file_utils.fsspec_head`, instead allowing them to be passed down via `storage_options`. This fixes an issue where s3fs did not recognize a timeout arg as well as fixes an issue mentioned in https://github.com/huggingface/datasets/issues/5281 by allowing users to pass down `storage_options` all the way from `datasets.load_dataset` to support implementation-specific credentials
Supports something like the following to provide credentials explicitly instead of relying on boto's methods of locating them
```
load_dataset(..., data_files=["s3://..."], storage_options={"profile": "..."})
```
|
closed
|
https://github.com/huggingface/datasets/pull/5673
| 2023-03-26T20:09:37
| 2023-03-28T15:03:38
| 2023-03-28T14:54:17
|
{
"login": "dwyatte",
"id": 2512762,
"type": "User"
}
|
[] | true
|
[] |
1,641,005,322
| 5,672
|
Pushing dataset to hub crash
|
### Describe the bug
Uploading a dataset with `push_to_hub()` fails without error description.
### Steps to reproduce the bug
Hey there,
I've built a image dataset of 100k images + text pair as described here https://huggingface.co/docs/datasets/image_dataset#imagefolder
Now I'm trying to push it to the hub but I'm running into issues. First, I tried doing it via git directly, I added all the files in git lfs and pushed but I got hit with an error saying huggingface only accept up to 10k files in a folder.
So I'm now trying with the `push_to_hub()` func as follow:
```python
from datasets import load_dataset
import os
dataset = load_dataset("imagefolder", data_dir="./data", split="train")
dataset.push_to_hub("tzvc/organization-logos", token=os.environ.get('HF_TOKEN'))
```
But again, this produces an error:
```
Resolving data files: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████| 100212/100212 [00:00<00:00, 439108.61it/s]
Downloading and preparing dataset imagefolder/default to /home/contact_theochampion/.cache/huggingface/datasets/imagefolder/default-20567ffc703aa314/0.0.0/37fbb85cc714a338bea574ac6c7d0b5be5aff46c1862c1989b20e0771199e93f...
Downloading data files: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████| 100211/100211 [00:00<00:00, 149323.73it/s]
Downloading data files: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 15947.92it/s]
Extracting data files: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 2245.34it/s]
Dataset imagefolder downloaded and prepared to /home/contact_theochampion/.cache/huggingface/datasets/imagefolder/default-20567ffc703aa314/0.0.0/37fbb85cc714a338bea574ac6c7d0b5be5aff46c1862c1989b20e0771199e93f. Subsequent calls will reuse this data.
Resuming upload of the dataset shards.
Pushing dataset shards to the dataset hub: 100%|██████████████████████████████████████████████████████████████████████████████████████████████| 14/14 [00:31<00:00, 2.24s/it]
Downloading metadata: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 118/118 [00:00<00:00, 225kB/s]
Traceback (most recent call last):
File "/home/contact_theochampion/organization-logos/push_to_hub.py", line 5, in <module>
dataset.push_to_hub("tzvc/organization-logos", token=os.environ.get('HF_TOKEN'))
File "/home/contact_theochampion/.local/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 5245, in push_to_hub
repo_info = dataset_infos[next(iter(dataset_infos))]
StopIteration
```
What could be happening here ?
### Expected behavior
The dataset is pushed to the hub
### Environment info
- `datasets` version: 2.10.1
- Platform: Linux-5.10.0-21-cloud-amd64-x86_64-with-glibc2.31
- Python version: 3.9.2
- PyArrow version: 11.0.0
- Pandas version: 1.5.3
|
closed
|
https://github.com/huggingface/datasets/issues/5672
| 2023-03-26T17:42:13
| 2023-03-30T08:11:05
| 2023-03-30T08:11:05
|
{
"login": "tzvc",
"id": 14275989,
"type": "User"
}
|
[] | false
|
[] |
1,640,840,012
| 5,671
|
How to use `load_dataset('glue', 'cola')`
|
### Describe the bug
I'm new to use HuggingFace datasets but I cannot use `load_dataset('glue', 'cola')`.
- I was stacked by the following problem:
```python
from datasets import load_dataset
cola_dataset = load_dataset('glue', 'cola')
---------------------------------------------------------------------------
InvalidVersion Traceback (most recent call last)
File <timed exec>:1
(Omit because of long error message)
File /usr/local/lib/python3.8/site-packages/packaging/version.py:197, in Version.__init__(self, version)
195 match = self._regex.search(version)
196 if not match:
--> 197 raise InvalidVersion(f"Invalid version: '{version}'")
199 # Store the parsed out pieces of the version
200 self._version = _Version(
201 epoch=int(match.group("epoch")) if match.group("epoch") else 0,
202 release=tuple(int(i) for i in match.group("release").split(".")),
(...)
208 local=_parse_local_version(match.group("local")),
209 )
InvalidVersion: Invalid version: '0.10.1,<0.11'
```
- You can check this full error message in my repository: [MLOps-Basics/week_0_project_setup/experimental_notebooks/data_exploration.ipynb](https://github.com/makinzm/MLOps-Basics/blob/eabab4b837880607d9968d3fa687c70177b2affd/week_0_project_setup/experimental_notebooks/data_exploration.ipynb)
### Steps to reproduce the bug
- This is my repository to reproduce: [MLOps-Basics/week_0_project_setup](https://github.com/makinzm/MLOps-Basics/tree/eabab4b837880607d9968d3fa687c70177b2affd/week_0_project_setup)
1. cd `/DockerImage` and command `docker build . -t week0`
2. cd `/` and command `docker-compose up`
3. Run `experimental_notebooks/data_exploration.ipynb`
----
Just to be sure, I wrote down Dockerfile and requirements.txt
- Dockerfile
```Dockerfile
FROM python:3.8
WORKDIR /root/working
RUN apt-get update && \
apt-get install -y python3-dev python3-pip python3-venv && \
apt-get clean && \
rm -rf /var/lib/apt/lists/*
COPY requirements.txt .
RUN pip3 install --no-cache-dir jupyter notebook && pip install --no-cache-dir -r requirements.txt
CMD ["bash"]
```
- requirements.txt
```txt
pytorch-lightning==1.2.10
datasets==1.6.2
transformers==4.5.1
scikit-learn==0.24.2
```
### Expected behavior
There is no bug to implement `load_dataset('glue', 'cola')`
### Environment info
I already wrote it.
|
closed
|
https://github.com/huggingface/datasets/issues/5671
| 2023-03-26T09:40:34
| 2023-03-28T07:43:44
| 2023-03-28T07:43:43
|
{
"login": "makinzm",
"id": 40193664,
"type": "User"
}
|
[] | false
|
[] |
1,640,607,045
| 5,670
|
Unable to load multi class classification datasets
|
### Describe the bug
I've been playing around with huggingface library, mostly with `datasets` and wanted to download the multi class classification datasets to fine tune BERT on this task. ([link](https://huggingface.co/docs/transformers/training#train-with-pytorch-trainer)).
While loading the dataset, I'm getting the following error snippet.
```
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
Cell In[44], line 3
1 from datasets import load_dataset
----> 3 imdb_dataset = load_dataset("yelp_review_full")
4 imdb_dataset
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/load.py:1719, 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)
1716 ignore_verifications = ignore_verifications or save_infos
1718 # Create a dataset builder
-> 1719 builder_instance = load_dataset_builder(
1720 path=path,
1721 name=name,
1722 data_dir=data_dir,
1723 data_files=data_files,
1724 cache_dir=cache_dir,
1725 features=features,
1726 download_config=download_config,
1727 download_mode=download_mode,
1728 revision=revision,
1729 use_auth_token=use_auth_token,
1730 **config_kwargs,
1731 )
1733 # Return iterable dataset in case of streaming
1734 if streaming:
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/load.py:1523, in load_dataset_builder(path, name, data_dir, data_files, cache_dir, features, download_config, download_mode, revision, use_auth_token, **config_kwargs)
1520 raise ValueError(error_msg)
1522 # Instantiate the dataset builder
-> 1523 builder_instance: DatasetBuilder = builder_cls(
1524 cache_dir=cache_dir,
1525 config_name=config_name,
1526 data_dir=data_dir,
1527 data_files=data_files,
1528 hash=hash,
1529 features=features,
1530 use_auth_token=use_auth_token,
1531 **builder_kwargs,
1532 **config_kwargs,
1533 )
1535 return builder_instance
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/builder.py:1292, in GeneratorBasedBuilder.__init__(self, writer_batch_size, *args, **kwargs)
1291 def __init__(self, *args, writer_batch_size=None, **kwargs):
-> 1292 super().__init__(*args, **kwargs)
1293 # Batch size used by the ArrowWriter
1294 # It defines the number of samples that are kept in memory before writing them
1295 # and also the length of the arrow chunks
1296 # None means that the ArrowWriter will use its default value
1297 self._writer_batch_size = writer_batch_size or self.DEFAULT_WRITER_BATCH_SIZE
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/builder.py:312, in DatasetBuilder.__init__(self, cache_dir, config_name, hash, base_path, info, features, use_auth_token, repo_id, data_files, data_dir, name, **config_kwargs)
309 # prepare info: DatasetInfo are a standardized dataclass across all datasets
310 # Prefill datasetinfo
311 if info is None:
--> 312 info = self.get_exported_dataset_info()
313 info.update(self._info())
314 info.builder_name = self.name
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/builder.py:412, in DatasetBuilder.get_exported_dataset_info(self)
400 def get_exported_dataset_info(self) -> DatasetInfo:
401 """Empty DatasetInfo if doesn't exist
402
403 Example:
(...)
410 ```
411 """
--> 412 return self.get_all_exported_dataset_infos().get(self.config.name, DatasetInfo())
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/builder.py:398, in DatasetBuilder.get_all_exported_dataset_infos(cls)
385 @classmethod
386 def get_all_exported_dataset_infos(cls) -> DatasetInfosDict:
387 """Empty dict if doesn't exist
388
389 Example:
(...)
396 ```
397 """
--> 398 return DatasetInfosDict.from_directory(cls.get_imported_module_dir())
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/info.py:370, in DatasetInfosDict.from_directory(cls, dataset_infos_dir)
368 dataset_metadata = DatasetMetadata.from_readme(Path(dataset_infos_dir) / "README.md")
369 if "dataset_info" in dataset_metadata:
--> 370 return cls.from_metadata(dataset_metadata)
371 if os.path.exists(os.path.join(dataset_infos_dir, config.DATASETDICT_INFOS_FILENAME)):
372 # this is just to have backward compatibility with dataset_infos.json files
373 with open(os.path.join(dataset_infos_dir, config.DATASETDICT_INFOS_FILENAME), encoding="utf-8") as f:
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/info.py:396, in DatasetInfosDict.from_metadata(cls, dataset_metadata)
387 return cls(
388 {
389 dataset_info_yaml_dict.get("config_name", "default"): DatasetInfo._from_yaml_dict(
(...)
393 }
394 )
395 else:
--> 396 dataset_info = DatasetInfo._from_yaml_dict(dataset_metadata["dataset_info"])
397 dataset_info.config_name = dataset_metadata["dataset_info"].get("config_name", "default")
398 return cls({dataset_info.config_name: dataset_info})
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/info.py:332, in DatasetInfo._from_yaml_dict(cls, yaml_data)
330 yaml_data = copy.deepcopy(yaml_data)
331 if yaml_data.get("features") is not None:
--> 332 yaml_data["features"] = Features._from_yaml_list(yaml_data["features"])
333 if yaml_data.get("splits") is not None:
334 yaml_data["splits"] = SplitDict._from_yaml_list(yaml_data["splits"])
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/features/features.py:1745, in Features._from_yaml_list(cls, yaml_data)
1742 else:
1743 raise TypeError(f"Expected a dict or a list but got {type(obj)}: {obj}")
-> 1745 return cls.from_dict(from_yaml_inner(yaml_data))
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/features/features.py:1741, in Features._from_yaml_list.<locals>.from_yaml_inner(obj)
1739 elif isinstance(obj, list):
1740 names = [_feature.pop("name") for _feature in obj]
-> 1741 return {name: from_yaml_inner(_feature) for name, _feature in zip(names, obj)}
1742 else:
1743 raise TypeError(f"Expected a dict or a list but got {type(obj)}: {obj}")
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/features/features.py:1741, in <dictcomp>(.0)
1739 elif isinstance(obj, list):
1740 names = [_feature.pop("name") for _feature in obj]
-> 1741 return {name: from_yaml_inner(_feature) for name, _feature in zip(names, obj)}
1742 else:
1743 raise TypeError(f"Expected a dict or a list but got {type(obj)}: {obj}")
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/features/features.py:1736, in Features._from_yaml_list.<locals>.from_yaml_inner(obj)
1734 return {"_type": snakecase_to_camelcase(obj["dtype"])}
1735 else:
-> 1736 return from_yaml_inner(obj["dtype"])
1737 else:
1738 return {"_type": snakecase_to_camelcase(_type), **unsimplify(obj)[_type]}
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/features/features.py:1738, in Features._from_yaml_list.<locals>.from_yaml_inner(obj)
1736 return from_yaml_inner(obj["dtype"])
1737 else:
-> 1738 return {"_type": snakecase_to_camelcase(_type), **unsimplify(obj)[_type]}
1739 elif isinstance(obj, list):
1740 names = [_feature.pop("name") for _feature in obj]
File /work/pi_adrozdov_umass_edu/syerawar_umass_edu/envs/vadops/lib/python3.10/site-packages/datasets/features/features.py:1706, in Features._from_yaml_list.<locals>.unsimplify(feature)
1704 if isinstance(feature.get("class_label"), dict) and isinstance(feature["class_label"].get("names"), dict):
1705 label_ids = sorted(feature["class_label"]["names"])
-> 1706 if label_ids and label_ids != list(range(label_ids[-1] + 1)):
1707 raise ValueError(
1708 f"ClassLabel expected a value for all label ids [0:{label_ids[-1] + 1}] but some ids are missing."
1709 )
1710 feature["class_label"]["names"] = [feature["class_label"]["names"][label_id] for label_id in label_ids]
TypeError: can only concatenate str (not "int") to str
```
The same issue happens when I try to load `go-emotions` multi class classification dataset. Could somebody guide me on how to fix this issue?
### Steps to reproduce the bug
Run the following code snippet in a python script/ notebook cell:
```
from datasets import load_dataset
yelp_dataset = load_dataset("yelp_review_full")
yelp_dataset
```
### Expected behavior
The dataset should be loaded perfectly, which showing the train, test and unsupervised splits with the basic data statistics
### Environment info
- `datasets` version: 2.6.1
- Platform: Linux-5.4.0-124-generic-x86_64-with-glibc2.31
- Python version: 3.10.9
- PyArrow version: 8.0.0
- Pandas version: 1.5.3
|
closed
|
https://github.com/huggingface/datasets/issues/5670
| 2023-03-25T18:06:15
| 2023-03-27T22:54:56
| 2023-03-27T22:54:56
|
{
"login": "ysahil97",
"id": 19690506,
"type": "User"
}
|
[] | false
|
[] |
1,638,070,046
| 5,669
|
Almost identical datasets, huge performance difference
|
### Describe the bug
I am struggling to understand (huge) performance difference between two datasets that are almost identical.
### Steps to reproduce the bug
# Fast (normal) dataset speed:
```python
import cv2
from datasets import load_dataset
from torch.utils.data import DataLoader
dataset = load_dataset("beans", split="train")
for x in DataLoader(dataset.with_format("torch"), batch_size=16, shuffle=True, num_workers=8):
pass
```
The above pass over the dataset takes about 1.5 seconds on my computer.
However, if I re-create (almost) the same dataset, the sweep takes HUGE amount of time: 15 minutes. Steps to reproduce:
```python
def transform(example):
example["image2"] = cv2.imread(example["image_file_path"])
return example
dataset2 = dataset.map(transform, remove_columns=["image"])
for x in DataLoader(dataset2.with_format("torch"), batch_size=16, shuffle=True, num_workers=8):
pass
```
### Expected behavior
Same timings
### Environment info
python==3.10.9
datasets==2.10.1
|
open
|
https://github.com/huggingface/datasets/issues/5669
| 2023-03-23T18:20:20
| 2023-04-09T18:56:23
| null |
{
"login": "eli-osherovich",
"id": 2437102,
"type": "User"
}
|
[] | false
|
[] |
1,638,018,598
| 5,668
|
Support for downloading only provided split
|
We can pass split to `_split_generators()`.
But I'm not sure if it's possible to solve cache issues, mostly with `dataset_info.json`
|
open
|
https://github.com/huggingface/datasets/pull/5668
| 2023-03-23T17:53:39
| 2023-03-24T06:43:14
| null |
{
"login": "polinaeterna",
"id": 16348744,
"type": "User"
}
|
[] | true
|
[] |
1,637,789,361
| 5,667
|
Jax requires jaxlib
|
close https://github.com/huggingface/datasets/issues/5666
|
closed
|
https://github.com/huggingface/datasets/pull/5667
| 2023-03-23T15:41:09
| 2023-03-23T16:23:11
| 2023-03-23T16:14:52
|
{
"login": "lhoestq",
"id": 42851186,
"type": "User"
}
|
[] | true
|
[] |
1,637,675,062
| 5,666
|
Support tensorflow 2.12.0 in CI
|
Once we find out the root cause of:
- #5663
we should revert the temporary pin on tensorflow introduced by:
- #5664
|
closed
|
https://github.com/huggingface/datasets/issues/5666
| 2023-03-23T14:37:51
| 2023-03-23T16:14:54
| 2023-03-23T16:14:54
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
}
] | false
|
[] |
1,637,193,648
| 5,665
|
Feature request: IterableDataset.push_to_hub
|
### Feature request
It'd be great to have a lazy push to hub, similar to the lazy loading we have with `IterableDataset`.
Suppose you'd like to filter [LAION](https://huggingface.co/datasets/laion/laion400m) based on certain conditions, but as LAION doesn't fit into your disk, you'd like to leverage streaming:
```
from datasets import load_dataset
dataset = load_dataset("laion/laion400m", streaming=True, split="train")
```
Then you could filter the dataset based on certain conditions:
```
filtered_dataset = dataset.filter(lambda example: example['HEIGHT'] > 400)
```
In order to persist this dataset and push it back to the hub, one currently needs to first load the entire filtered dataset on disk and then push:
```
from datasets import Dataset
Dataset.from_generator(filtered_dataset.__iter__).push_to_hub(...)
```
It would be great if we can instead lazy push to the data to the hub (basically stream the data to the hub), not being limited by our disk size:
```
filtered_dataset.push_to_hub("my-filtered-dataset")
```
### Motivation
This feature would be very useful for people that want to filter huge datasets without having to load the entire dataset or a filtered version thereof on their local disk.
### Your contribution
Happy to test out a PR :)
|
closed
|
https://github.com/huggingface/datasets/issues/5665
| 2023-03-23T09:53:04
| 2025-06-06T16:13:22
| 2025-06-06T16:12:36
|
{
"login": "NielsRogge",
"id": 48327001,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
}
] | false
|
[] |
1,637,192,684
| 5,664
|
Fix CI by temporarily pinning tensorflow < 2.12.0
|
As a hotfix for our CI, temporarily pin `tensorflow` upper version:
- In Python 3.10, tensorflow-2.12.0 also installs `jax`
Fix #5663
Until root cause is fixed.
|
closed
|
https://github.com/huggingface/datasets/pull/5664
| 2023-03-23T09:52:26
| 2023-03-23T10:17:11
| 2023-03-23T10:09:54
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[] | true
|
[] |
1,637,173,248
| 5,663
|
CI is broken: ModuleNotFoundError: jax requires jaxlib to be installed
|
CI test_py310 is broken: see https://github.com/huggingface/datasets/actions/runs/4498945505/jobs/7916194236?pr=5662
```
FAILED tests/test_arrow_dataset.py::BaseDatasetTest::test_map_jax_in_memory - ModuleNotFoundError: jax requires jaxlib to be installed. See https://github.com/google/jax#installation for installation instructions.
FAILED tests/test_arrow_dataset.py::BaseDatasetTest::test_map_jax_on_disk - ModuleNotFoundError: jax requires jaxlib to be installed. See https://github.com/google/jax#installation for installation instructions.
FAILED tests/test_formatting.py::FormatterTest::test_jax_formatter - ModuleNotFoundError: jax requires jaxlib to be installed. See https://github.com/google/jax#installation for installation instructions.
FAILED tests/test_formatting.py::FormatterTest::test_jax_formatter_audio - ModuleNotFoundError: jax requires jaxlib to be installed. See https://github.com/google/jax#installation for installation instructions.
FAILED tests/test_formatting.py::FormatterTest::test_jax_formatter_device - ModuleNotFoundError: jax requires jaxlib to be installed. See https://github.com/google/jax#installation for installation instructions.
FAILED tests/test_formatting.py::FormatterTest::test_jax_formatter_image - ModuleNotFoundError: jax requires jaxlib to be installed. See https://github.com/google/jax#installation for installation instructions.
FAILED tests/test_formatting.py::FormatterTest::test_jax_formatter_jnp_array_kwargs - ModuleNotFoundError: jax requires jaxlib to be installed. See https://github.com/google/jax#installation for installation instructions.
FAILED tests/features/test_features.py::CastToPythonObjectsTest::test_cast_to_python_objects_jax - ModuleNotFoundError: jax requires jaxlib to be installed. See https://github.com/google/jax#installation for installation instructions.
===== 8 failed, 2147 passed, 10 skipped, 37 warnings in 228.69s (0:03:48) ======
```
|
closed
|
https://github.com/huggingface/datasets/issues/5663
| 2023-03-23T09:39:43
| 2023-03-23T10:09:55
| 2023-03-23T10:09:55
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[
{
"name": "bug",
"color": "d73a4a"
}
] | false
|
[] |
1,637,140,813
| 5,662
|
Fix unnecessary dict comprehension
|
After ruff-0.0.258 release, the C416 rule was updated with unnecessary dict comprehensions. See:
- https://github.com/charliermarsh/ruff/releases/tag/v0.0.258
- https://github.com/charliermarsh/ruff/pull/3605
This PR fixes one unnecessary dict comprehension in our code: no need to unpack and re-pack the tuple values.
Fix #5661
|
closed
|
https://github.com/huggingface/datasets/pull/5662
| 2023-03-23T09:18:58
| 2023-03-23T09:46:59
| 2023-03-23T09:37:49
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[] | true
|
[] |
1,637,129,445
| 5,661
|
CI is broken: Unnecessary `dict` comprehension
|
CI check_code_quality is broken:
```
src/datasets/arrow_dataset.py:3267:35: C416 [*] Unnecessary `dict` comprehension (rewrite using `dict()`)
Found 1 error.
```
|
closed
|
https://github.com/huggingface/datasets/issues/5661
| 2023-03-23T09:13:01
| 2023-03-23T09:37:51
| 2023-03-23T09:37:51
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[
{
"name": "bug",
"color": "d73a4a"
}
] | false
|
[] |
1,635,543,646
| 5,660
|
integration with imbalanced-learn
|
### Feature request
Wouldn't it be great if the various class balancing operations from imbalanced-learn were available as part of datasets?
### Motivation
I'm trying to use imbalanced-learn to balance a dataset, but it's not clear how to get the two to interoperate - what would be great would be some examples. I've looked online, asked gpt-4, but so far not making much progress.
### Your contribution
If I can get this working myself I can submit a PR with example code to go in the docs
|
closed
|
https://github.com/huggingface/datasets/issues/5660
| 2023-03-22T11:05:17
| 2023-07-06T18:10:15
| 2023-07-06T18:10:15
|
{
"login": "tansaku",
"id": 30216,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
},
{
"name": "wontfix",
"color": "ffffff"
}
] | false
|
[] |
1,635,447,540
| 5,659
|
[Audio] Soundfile/libsndfile requirements too stringent for decoding mp3 files
|
### Describe the bug
I'm encountering several issues trying to load mp3 audio files using `datasets` on a TPU v4.
The PR https://github.com/huggingface/datasets/pull/5573 updated the audio loading logic to rely solely on the `soundfile`/`libsndfile` libraries for loading audio samples, regardless of their file type.
The installation guide suggests that `libsndfile` is bundled in when `soundfile` is pip installed:
https://github.com/huggingface/datasets/blob/e1af108015e43f9df8734a1faeeaeb9eafce3971/docs/source/installation.md?plain=1#L70-L71
However, just pip installing `soundfile==0.12.1` throws an error that `libsndfile` is missing:
```
pip install soundfile==0.12.1
```
Then:
```python
>>> soundfile
>>> soundfile.__libsndfile_version__
```
<details>
<summary> Traceback (most recent call last): </summary>
```
File "/home/sanchitgandhi/hf/lib/python3.8/site-packages/soundfile.py", line 161, in <module>
import _soundfile_data # ImportError if this doesn't exist
ModuleNotFoundError: No module named '_soundfile_data'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/home/sanchitgandhi/hf/lib/python3.8/site-packages/soundfile.py", line 170, in <module>
raise OSError('sndfile library not found using ctypes.util.find_library')
OSError: sndfile library not found using ctypes.util.find_library
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "<string>", line 1, in <module>
File "/home/sanchitgandhi/hf/lib/python3.8/site-packages/soundfile.py", line 192, in <module>
_snd = _ffi.dlopen(_explicit_libname)
OSError: cannot load library 'libsndfile.so': libsndfile.so: cannot open shared object file: No such file or directory
```
</details>
Thus, I've followed the official instructions for installing the `soundfile` package from https://github.com/bastibe/python-soundfile#installation, which states that `libsndfile` needs to be installed separately as:
```
pip install --upgrade soundfile
sudo apt install libsndfile1
```
We can now import `soundfile`:
```python
>>> import soundfile
>>> soundfile.__version__
'0.12.1'
>>> soundfile.__libsndfile_version__
'1.0.28'
```
We see that we have `soundfile==0.12.1`, which matches the `datasets[audio]` package constraints:
https://github.com/huggingface/datasets/blob/e1af108015e43f9df8734a1faeeaeb9eafce3971/setup.py#L144-L147
But we have `libsndfile==1.0.28`, which is too low for decoding mp3 files:
https://github.com/huggingface/datasets/blob/e1af108015e43f9df8734a1faeeaeb9eafce3971/src/datasets/config.py#L136-L138
Updating/upgrading the `libsndfile` doesn't change this:
```
sudo apt-get update
sudo apt-get upgrade
```
Is there any other suggestion for how to get a compatible `libsndfile` version? Currently, the version bundled with Ubuntu `apt-get` is too low for decoding mp3 files.
Maybe we could add this under `setup.py` such that we install the correct `libsndfile` version when we do `pip install datasets[audio]`? IMO this would help circumvent such version issues.
### Steps to reproduce the bug
Environment described above. Loading mp3 files:
```python
from datasets import load_dataset
common_voice_es = load_dataset("common_voice", "es", split="validation", streaming=True)
print(next(iter(common_voice_es)))
```
```python
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
Cell In[4], line 2
1 common_voice_es = load_dataset("common_voice", "es", split="validation", streaming=True)
----> 2 print(next(iter(common_voice_es)))
File ~/datasets/src/datasets/iterable_dataset.py:941, in IterableDataset.__iter__(self)
937 for key, example in ex_iterable:
938 if self.features:
939 # `IterableDataset` automatically fills missing columns with None.
940 # This is done with `_apply_feature_types_on_example`.
--> 941 yield _apply_feature_types_on_example(
942 example, self.features, token_per_repo_id=self._token_per_repo_id
943 )
944 else:
945 yield example
File ~/datasets/src/datasets/iterable_dataset.py:700, in _apply_feature_types_on_example(example, features, token_per_repo_id)
698 encoded_example = features.encode_example(example)
699 # Decode example for Audio feature, e.g.
--> 700 decoded_example = features.decode_example(encoded_example, token_per_repo_id=token_per_repo_id)
701 return decoded_example
File ~/datasets/src/datasets/features/features.py:1864, in Features.decode_example(self, example, token_per_repo_id)
1850 def decode_example(self, example: dict, token_per_repo_id: Optional[Dict[str, Union[str, bool, None]]] = None):
1851 """Decode example with custom feature decoding.
1852
1853 Args:
(...)
1861 `dict[str, Any]`
1862 """
-> 1864 return {
1865 column_name: decode_nested_example(feature, value, token_per_repo_id=token_per_repo_id)
1866 if self._column_requires_decoding[column_name]
1867 else value
1868 for column_name, (feature, value) in zip_dict(
1869 {key: value for key, value in self.items() if key in example}, example
1870 )
1871 }
File ~/datasets/src/datasets/features/features.py:1865, in <dictcomp>(.0)
1850 def decode_example(self, example: dict, token_per_repo_id: Optional[Dict[str, Union[str, bool, None]]] = None):
1851 """Decode example with custom feature decoding.
1852
1853 Args:
(...)
1861 `dict[str, Any]`
1862 """
1864 return {
-> 1865 column_name: decode_nested_example(feature, value, token_per_repo_id=token_per_repo_id)
1866 if self._column_requires_decoding[column_name]
1867 else value
1868 for column_name, (feature, value) in zip_dict(
1869 {key: value for key, value in self.items() if key in example}, example
1870 )
1871 }
File ~/datasets/src/datasets/features/features.py:1308, in decode_nested_example(schema, obj, token_per_repo_id)
1305 elif isinstance(schema, (Audio, Image)):
1306 # we pass the token to read and decode files from private repositories in streaming mode
1307 if obj is not None and schema.decode:
-> 1308 return schema.decode_example(obj, token_per_repo_id=token_per_repo_id)
1309 return obj
File ~/datasets/src/datasets/features/audio.py:167, in Audio.decode_example(self, value, token_per_repo_id)
162 raise RuntimeError(
163 "Decoding 'opus' files requires system library 'libsndfile'>=1.0.31, "
164 'You can try to update `soundfile` python library: `pip install "soundfile>=0.12.1"`. '
165 )
166 elif not config.IS_MP3_SUPPORTED and audio_format == "mp3":
--> 167 raise RuntimeError(
168 "Decoding 'mp3' files requires system library 'libsndfile'>=1.1.0, "
169 'You can try to update `soundfile` python library: `pip install "soundfile>=0.12.1"`. '
170 )
172 if file is None:
173 token_per_repo_id = token_per_repo_id or {}
RuntimeError: Decoding 'mp3' files requires system library 'libsndfile'>=1.1.0, You can try to update `soundfile` python library: `pip install "soundfile>=0.12.1"`.
```
### Expected behavior
Load mp3 files!
### Environment info
- `datasets` version: 2.10.2.dev0
- Platform: Linux-5.13.0-1023-gcp-x86_64-with-glibc2.29
- Python version: 3.8.10
- Huggingface_hub version: 0.13.1
- PyArrow version: 11.0.0
- Pandas version: 1.5.3
- Soundfile version: 0.12.1
- Libsndfile version: 1.0.28
|
closed
|
https://github.com/huggingface/datasets/issues/5659
| 2023-03-22T10:07:33
| 2024-07-12T01:35:01
| 2023-04-07T08:51:28
|
{
"login": "sanchit-gandhi",
"id": 93869735,
"type": "User"
}
|
[] | false
|
[] |
1,634,867,204
| 5,658
|
docs: Update num_shards docs to mention num_proc on Dataset and DatasetDict
|
Closes #5653
@mariosasko
|
closed
|
https://github.com/huggingface/datasets/pull/5658
| 2023-03-22T00:12:18
| 2023-03-24T16:43:34
| 2023-03-24T16:36:21
|
{
"login": "connor-henderson",
"id": 78612354,
"type": "User"
}
|
[] | true
|
[] |
1,634,156,563
| 5,656
|
Fix `fsspec.open` when using an HTTP proxy
|
Most HTTP(S) downloads from this library support proxy automatically by reading the `HTTP_PROXY` environment variable (et al.) because `requests` is widely used. However, in some parts of the code, `fsspec` is used, which in turn uses `aiohttp` for HTTP(S) requests (as opposed to `requests`), which in turn doesn't support reading proxy env variables by default. This PR enables reading them automatically.
Read [aiohttp docs on using proxies](https://docs.aiohttp.org/en/stable/client_advanced.html?highlight=trust_env#proxy-support).
For context, [the Python library requests](https://requests.readthedocs.io/en/latest/user/advanced/?highlight=http_proxy#proxies) and [the official Python library via `urllib.urlopen` support this automatically by default](https://docs.python.org/3/library/urllib.request.html#urllib.request.urlopen). Many (most common ones?) programs also do the same, including cURL, APT, Wget, and many others.
|
closed
|
https://github.com/huggingface/datasets/pull/5656
| 2023-03-21T15:23:29
| 2023-03-23T14:14:50
| 2023-03-23T13:15:46
|
{
"login": "bryant1410",
"id": 3905501,
"type": "User"
}
|
[] | true
|
[] |
1,634,030,017
| 5,655
|
Improve features decoding in to_iterable_dataset
|
Following discussion at https://github.com/huggingface/datasets/pull/5589
Right now `to_iterable_dataset` on images/audio hurts iterable dataset performance a lot (e.g. x4 slower because it encodes+decodes images/audios unnecessarily).
I fixed it by providing a generator that yields undecoded examples
|
closed
|
https://github.com/huggingface/datasets/pull/5655
| 2023-03-21T14:18:09
| 2023-03-23T13:19:27
| 2023-03-23T13:12:25
|
{
"login": "lhoestq",
"id": 42851186,
"type": "User"
}
|
[] | true
|
[] |
1,633,523,705
| 5,654
|
Offset overflow when executing Dataset.map
|
### Describe the bug
Hi, I'm trying to use `.map` method to cache multiple random crops from the image to speed up data processing during training, as the image size is too big.
The map function executes all iterations, and then returns the following error:
```bash
Traceback (most recent call last):
File "/home/ubuntu/miniconda3/envs/enhancement/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 3353, in _map_single
writer.finalize() # close_stream=bool(buf_writer is None)) # We only close if we are writing in a file
File "/home/ubuntu/miniconda3/envs/enhancement/lib/python3.8/site-packages/datasets/arrow_writer.py", line 582, in finalize
self.write_examples_on_file()
File "/home/ubuntu/miniconda3/envs/enhancement/lib/python3.8/site-packages/datasets/arrow_writer.py", line 446, in write_examples_on_file
self.write_batch(batch_examples=batch_examples)
File "/home/ubuntu/miniconda3/envs/enhancement/lib/python3.8/site-packages/datasets/arrow_writer.py", line 555, in write_batch
self.write_table(pa_table, writer_batch_size)
File "/home/ubuntu/miniconda3/envs/enhancement/lib/python3.8/site-packages/datasets/arrow_writer.py", line 567, in write_table
pa_table = pa_table.combine_chunks()
File "pyarrow/table.pxi", line 3315, 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
```
Here is the minimal code (`/home/datasets/DIV2K_train_HR` is just a folder of images that can be replaced by any appropriate):
### Steps to reproduce the bug
```python
from glob import glob
import torch
from datasets import Dataset, Image
from torchvision.transforms import PILToTensor, RandomCrop
file_paths = glob("/home/datasets/DIV2K_train_HR/*")
to_tensor = PILToTensor()
crop_transf = RandomCrop(size=256)
def prepare_data(example):
tensor = to_tensor(example["image"].convert("RGB"))
return {"hr": torch.stack([crop_transf(tensor) for _ in range(25)])}
train_data = Dataset.from_dict({"image": file_paths}).cast_column("image", Image())
train_data = train_data.map(
prepare_data,
cache_file_name="/home/datasets/DIV2K_train_HR_crops.tmp",
desc="Caching multiple random crops of image",
remove_columns="image",
)
print(train_data[0].keys(), train_data[0]["hr"].shape)
```
### Expected behavior
Cached file is stored at `"/home/datasets/DIV2K_train_HR_crops.tmp"`, output is `dict_keys(['hr']) torch.Size([25, 3, 256, 256])`
### Environment info
- `datasets` version: 2.10.1
- Platform: Linux-5.15.0-67-generic-x86_64-with-glibc2.10
- Python version: 3.8.16
- PyArrow version: 11.0.0
- Pandas version: 1.5.3
- Pytorch version: 2.0.0+cu117
- torchvision version: 0.15.1+cu117
|
open
|
https://github.com/huggingface/datasets/issues/5654
| 2023-03-21T09:33:27
| 2023-03-21T10:32:07
| null |
{
"login": "jan-pair",
"id": 118280608,
"type": "User"
}
|
[] | false
|
[] |
1,633,254,159
| 5,653
|
Doc: save_to_disk, `num_proc` will affect `num_shards`, but it's not documented
|
### Describe the bug
[`num_proc`](https://huggingface.co/docs/datasets/main/en/package_reference/main_classes#datasets.DatasetDict.save_to_disk.num_proc) will affect `num_shards`, but it's not documented
### Steps to reproduce the bug
Nothing to reproduce
### Expected behavior
[document of `num_shards`](https://huggingface.co/docs/datasets/main/en/package_reference/main_classes#datasets.DatasetDict.save_to_disk.num_shards) explicitly says that it depends on `max_shard_size`, it should also mention `num_proc`.
### Environment info
datasets main document
|
closed
|
https://github.com/huggingface/datasets/issues/5653
| 2023-03-21T05:25:35
| 2023-03-24T16:36:23
| 2023-03-24T16:36:23
|
{
"login": "RmZeta2718",
"id": 42400165,
"type": "User"
}
|
[
{
"name": "documentation",
"color": "0075ca"
},
{
"name": "good first issue",
"color": "7057ff"
}
] | false
|
[] |
1,632,546,073
| 5,652
|
Copy features
|
Some users (even internally at HF) are doing
```python
dset_features = dset.features
dset_features.pop(col_to_remove)
dset = dset.map(..., features=dset_features)
```
Right now this causes issues because it modifies the features dict in place before the map.
In this PR I modified `dset.features` to return a copy of the features, so that users can modify it if they want.
|
closed
|
https://github.com/huggingface/datasets/pull/5652
| 2023-03-20T17:17:23
| 2023-03-23T13:19:19
| 2023-03-23T13:12:08
|
{
"login": "lhoestq",
"id": 42851186,
"type": "User"
}
|
[] | true
|
[] |
1,631,967,509
| 5,651
|
expanduser in save_to_disk
|
### Describe the bug
save_to_disk() does not expand `~`
1. `dataset = load_datasets("any dataset")`
2. `dataset.save_to_disk("~/data")`
3. a folder named "~" created in current folder
4. FileNotFoundError is raised, because the expanded path does not exist (`/home/<user>/data`)
related issue https://github.com/huggingface/transformers/issues/10628
### Steps to reproduce the bug
As described above.
### Expected behavior
expanduser correctly
### Environment info
- datasets 2.10.1
- python 3.10
|
closed
|
https://github.com/huggingface/datasets/issues/5651
| 2023-03-20T12:02:18
| 2023-10-27T14:04:37
| 2023-10-27T14:04:37
|
{
"login": "RmZeta2718",
"id": 42400165,
"type": "User"
}
|
[
{
"name": "good first issue",
"color": "7057ff"
}
] | false
|
[] |
1,630,336,919
| 5,650
|
load_dataset can't work correct with my image data
|
I have about 20000 images in my folder which divided into 4 folders with class names.
When i use load_dataset("my_folder_name", split="train") this function create dataset in which there are only 4 images, the remaining 19000 images were not added there. What is the problem and did not understand. Tried converting images and the like but absolutely nothing worked
|
closed
|
https://github.com/huggingface/datasets/issues/5650
| 2023-03-18T13:59:13
| 2023-07-24T14:13:02
| 2023-07-24T14:13:01
|
{
"login": "WiNE-iNEFF",
"id": 41611046,
"type": "User"
}
|
[] | false
|
[] |
1,630,173,460
| 5,649
|
The index column created with .to_sql() is dependent on the batch_size when writing
|
### Describe the bug
It seems like the "index" column is designed to be unique? The values are only unique per batch. The SQL index is not a unique index.
This can be a problem, for instance, when building a faiss index on a dataset and then trying to match up ids with a sql export.
### Steps to reproduce the bug
```
from datasets import Dataset
import sqlite3
db = sqlite3.connect(":memory:")
nice_numbers = Dataset.from_dict({"nice_number": range(101,106)})
nice_numbers.to_sql("nice1", db, batch_size=1)
nice_numbers.to_sql("nice2", db, batch_size=2)
print(db.execute("select * from nice1").fetchall()) # [(0, 101), (0, 102), (0, 103), (0, 104), (0, 105)]
print(db.execute("select * from nice2").fetchall()) # [(0, 101), (1, 102), (0, 103), (1, 104), (0, 105)]
```
### Expected behavior
I expected the "index" column to be unique
### Environment info
```
% datasets-cli env
Copy-and-paste the text below in your GitHub issue.
- `datasets` version: 2.10.1
- Platform: macOS-13.2.1-arm64-arm-64bit
- Python version: 3.9.6
- PyArrow version: 7.0.0
- Pandas version: 1.5.2
zsh: segmentation fault datasets-cli env
```
|
closed
|
https://github.com/huggingface/datasets/issues/5649
| 2023-03-18T05:25:17
| 2023-06-17T07:01:57
| 2023-06-17T07:01:57
|
{
"login": "lsb",
"id": 45281,
"type": "User"
}
|
[] | false
|
[] |
1,629,253,719
| 5,648
|
flatten_indices doesn't work with pandas format
|
### Describe the bug
Hi,
I noticed that `flatten_indices` throws an error when the batch format is `pandas`. This is probably due to the fact that flatten_indices uses map internally which doesn't accept dataframes as the transformation function output
### Steps to reproduce the bug
tabular_data = pd.DataFrame(np.random.randn(10,10))
tabular_data = datasets.arrow_dataset.Dataset.from_pandas(tabular_data)
tabular_data.with_format("pandas").select([0,1,2,3]).flatten_indices()
### Expected behavior
No error thrown
### Environment info
- `datasets` version: 2.10.1
- Python version: 3.9.5
- PyArrow version: 11.0.0
- Pandas version: 1.4.1
|
open
|
https://github.com/huggingface/datasets/issues/5648
| 2023-03-17T12:44:25
| 2023-03-21T13:12:03
| null |
{
"login": "alialamiidrissi",
"id": 14365168,
"type": "User"
}
|
[
{
"name": "bug",
"color": "d73a4a"
}
] | false
|
[] |
1,628,225,544
| 5,647
|
Make all print statements optional
|
### Feature request
Make all print statements optional to speed up the development
### Motivation
Im loading multiple tiny datasets and all the print statements make the loading slower
### Your contribution
I can help contribute
|
closed
|
https://github.com/huggingface/datasets/issues/5647
| 2023-03-16T20:30:07
| 2023-07-21T14:20:25
| 2023-07-21T14:20:24
|
{
"login": "gagan3012",
"id": 49101362,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
}
] | false
|
[] |
1,627,838,762
| 5,646
|
Allow self as key in `Features`
|
Fix #5641
|
closed
|
https://github.com/huggingface/datasets/pull/5646
| 2023-03-16T16:17:03
| 2023-03-16T17:21:58
| 2023-03-16T17:14:50
|
{
"login": "mariosasko",
"id": 47462742,
"type": "User"
}
|
[] | true
|
[] |
1,627,108,278
| 5,645
|
Datasets map and select(range()) is giving dill error
|
### Describe the bug
I'm using Huggingface Datasets library to load the dataset in google colab
When I do,
> data = train_dataset.select(range(10))
or
> train_datasets = train_dataset.map(
> process_data_to_model_inputs,
> batched=True,
> batch_size=batch_size,
> remove_columns=["article", "abstract"],
> )
I get following error: `module 'dill._dill' has no attribute 'log'`
I've tried downgrading the dill version from latest to 0.2.8, but no luck.
Stack trace:
> ---------------------------------------------------------------------------
> ModuleNotFoundError Traceback (most recent call last)
> /usr/local/lib/python3.9/dist-packages/datasets/utils/py_utils.py in _no_cache_fields(obj)
> 367 try:
> --> 368 import transformers as tr
> 369
>
> ModuleNotFoundError: No module named 'transformers'
>
> During handling of the above exception, another exception occurred:
>
> AttributeError Traceback (most recent call last)
> 17 frames
> <ipython-input-13-dd14813880a6> in <module>
> ----> 1 test = train_dataset.select(range(10))
>
> /usr/local/lib/python3.9/dist-packages/datasets/arrow_dataset.py in wrapper(*args, **kwargs)
> 155 }
> 156 # apply actual function
> --> 157 out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
> 158 datasets: List["Dataset"] = list(out.values()) if isinstance(out, dict) else [out]
> 159 # re-apply format to the output
>
> /usr/local/lib/python3.9/dist-packages/datasets/fingerprint.py in wrapper(*args, **kwargs)
> 155 if kwargs.get(fingerprint_name) is None:
> 156 kwargs_for_fingerprint["fingerprint_name"] = fingerprint_name
> --> 157 kwargs[fingerprint_name] = update_fingerprint(
> 158 self._fingerprint, transform, kwargs_for_fingerprint
> 159 )
>
> /usr/local/lib/python3.9/dist-packages/datasets/fingerprint.py in update_fingerprint(fingerprint, transform, transform_args)
> 103 for key in sorted(transform_args):
> 104 hasher.update(key)
> --> 105 hasher.update(transform_args[key])
> 106 return hasher.hexdigest()
> 107
>
> /usr/local/lib/python3.9/dist-packages/datasets/fingerprint.py in update(self, value)
> 55 def update(self, value):
> 56 self.m.update(f"=={type(value)}==".encode("utf8"))
> ---> 57 self.m.update(self.hash(value).encode("utf-8"))
> 58
> 59 def hexdigest(self):
>
> /usr/local/lib/python3.9/dist-packages/datasets/fingerprint.py in hash(cls, value)
> 51 return cls.dispatch[type(value)](cls, value)
> 52 else:
> ---> 53 return cls.hash_default(value)
> 54
> 55 def update(self, value):
>
> /usr/local/lib/python3.9/dist-packages/datasets/fingerprint.py in hash_default(cls, value)
> 44 @classmethod
> 45 def hash_default(cls, value):
> ---> 46 return cls.hash_bytes(dumps(value))
> 47
> 48 @classmethod
>
> /usr/local/lib/python3.9/dist-packages/datasets/utils/py_utils.py in dumps(obj)
> 387 file = StringIO()
> 388 with _no_cache_fields(obj):
> --> 389 dump(obj, file)
> 390 return file.getvalue()
> 391
>
> /usr/local/lib/python3.9/dist-packages/datasets/utils/py_utils.py in dump(obj, file)
> 359 def dump(obj, file):
> 360 """pickle an object to a file"""
> --> 361 Pickler(file, recurse=True).dump(obj)
> 362 return
> 363
>
> /usr/local/lib/python3.9/dist-packages/dill/_dill.py in dump(self, obj)
> 392 return
> 393
> --> 394 def load_session(filename='/tmp/session.pkl', main=None):
> 395 """update the __main__ module with the state from the session file"""
> 396 if main is None: main = _main_module
>
> /usr/lib/python3.9/pickle.py in dump(self, obj)
> 485 if self.proto >= 4:
> 486 self.framer.start_framing()
> --> 487 self.save(obj)
> 488 self.write(STOP)
> 489 self.framer.end_framing()
>
> /usr/local/lib/python3.9/dist-packages/dill/_dill.py in save(self, obj, save_persistent_id)
> 386 pickler._byref = False # disable pickling by name reference
> 387 pickler._recurse = False # disable pickling recursion for globals
> --> 388 pickler._session = True # is best indicator of when pickling a session
> 389 pickler.dump(main)
> 390 finally:
>
> /usr/lib/python3.9/pickle.py in save(self, obj, save_persistent_id)
> 558 f = self.dispatch.get(t)
> 559 if f is not None:
> --> 560 f(self, obj) # Call unbound method with explicit self
> 561 return
> 562
>
> /usr/local/lib/python3.9/dist-packages/dill/_dill.py in save_singleton(pickler, obj)
>
> /usr/lib/python3.9/pickle.py in save_reduce(self, func, args, state, listitems, dictitems, state_setter, obj)
> 689 write(NEWOBJ)
> 690 else:
> --> 691 save(func)
> 692 save(args)
> 693 write(REDUCE)
>
> /usr/local/lib/python3.9/dist-packages/dill/_dill.py in save(self, obj, save_persistent_id)
> 386 pickler._byref = False # disable pickling by name reference
> 387 pickler._recurse = False # disable pickling recursion for globals
> --> 388 pickler._session = True # is best indicator of when pickling a session
> 389 pickler.dump(main)
> 390 finally:
>
> /usr/lib/python3.9/pickle.py in save(self, obj, save_persistent_id)
> 558 f = self.dispatch.get(t)
> 559 if f is not None:
> --> 560 f(self, obj) # Call unbound method with explicit self
> 561 return
> 562
>
> /usr/local/lib/python3.9/dist-packages/datasets/utils/py_utils.py in save_function(pickler, obj)
> 583 dill._dill.log.info("# F1")
> 584 else:
> --> 585 dill._dill.log.info("F2: %s" % obj)
> 586 name = getattr(obj, "__qualname__", getattr(obj, "__name__", None))
> 587 dill._dill.StockPickler.save_global(pickler, obj, name=name)
>
> AttributeError: module 'dill._dill' has no attribute 'log'
### Steps to reproduce the bug
After loading the dataset(eg: https://huggingface.co/datasets/scientific_papers) in google colab
do either
> data = train_dataset.select(range(10))
or
> train_datasets = train_dataset.map(
> process_data_to_model_inputs,
> batched=True,
> batch_size=batch_size,
> remove_columns=["article", "abstract"],
> )
### Expected behavior
The map and select function should work
### Environment info
dataset: https://huggingface.co/datasets/scientific_papers
dill = 0.3.6
python= 3.9.16
transformer = 4.2.0
|
closed
|
https://github.com/huggingface/datasets/issues/5645
| 2023-03-16T10:01:28
| 2023-03-17T04:24:51
| 2023-03-17T04:24:51
|
{
"login": "Tanya-11",
"id": 90728105,
"type": "User"
}
|
[] | false
|
[] |
1,626,204,046
| 5,644
|
Allow direct cast from binary to Audio/Image
|
To address https://github.com/huggingface/datasets/discussions/5593.
|
closed
|
https://github.com/huggingface/datasets/pull/5644
| 2023-03-15T20:02:54
| 2023-03-16T14:20:44
| 2023-03-16T14:12:55
|
{
"login": "mariosasko",
"id": 47462742,
"type": "User"
}
|
[] | true
|
[] |
1,626,160,220
| 5,643
|
Support PyArrow arrays as column values in `from_dict`
|
For consistency with `pa.Table.from_pydict`, which supports both Python lists and PyArrow arrays as column values.
"Fixes" https://discuss.huggingface.co/t/pyarrow-lib-floatarray-did-not-recognize-python-value-type-when-inferring-an-arrow-data-type/33417
|
closed
|
https://github.com/huggingface/datasets/pull/5643
| 2023-03-15T19:32:40
| 2023-03-16T17:23:06
| 2023-03-16T17:15:40
|
{
"login": "mariosasko",
"id": 47462742,
"type": "User"
}
|
[] | true
|
[] |
1,626,043,177
| 5,642
|
Bump hfh to 0.11.0
|
to fix errors like
```
requests.exceptions.HTTPError: 400 Client Error: Bad Request for url: https://hub-ci.huggingface.co/api/datasets/__DUMMY_TRANSFORMERS_USER__/...
```
(e.g. from this [failing CI](https://github.com/huggingface/datasets/actions/runs/4428956210/jobs/7769160997))
0.11.0 is the current minimum version in `transformers`
around 5% of users are currently using versions `<0.11.0`
|
closed
|
https://github.com/huggingface/datasets/pull/5642
| 2023-03-15T18:26:07
| 2023-03-20T12:34:09
| 2023-03-20T12:26:58
|
{
"login": "lhoestq",
"id": 42851186,
"type": "User"
}
|
[] | true
|
[] |
1,625,942,730
| 5,641
|
Features cannot be named "self"
|
### Describe the bug
Hi,
I noticed that we cannot create a HuggingFace dataset from Pandas DataFrame with a column named `self`.
The error seems to be coming from arguments validation in the `Features.from_dict` function.
### Steps to reproduce the bug
```python
import datasets
dummy_pandas = pd.DataFrame([0,1,2,3], columns = ["self"])
datasets.arrow_dataset.Dataset.from_pandas(dummy_pandas)
```
### Expected behavior
No error thrown
### Environment info
- `datasets` version: 2.8.0
- Python version: 3.9.5
- PyArrow version: 6.0.1
- Pandas version: 1.4.1
|
closed
|
https://github.com/huggingface/datasets/issues/5641
| 2023-03-15T17:16:40
| 2023-03-16T17:14:51
| 2023-03-16T17:14:51
|
{
"login": "alialamiidrissi",
"id": 14365168,
"type": "User"
}
|
[] | false
|
[] |
1,625,896,057
| 5,640
|
Less zip false positives
|
`zipfile.is_zipfile` return false positives for some Parquet files. It causes errors when loading certain parquet datasets, where some files are considered ZIP files by `zipfile.is_zipfile`
This is a known issue: https://github.com/python/cpython/issues/72680
At first I wanted to rely only on magic numbers, but then I found that someone contributed a [fix to is_zipfile](https://github.com/python/cpython/pull/5053) - do you think we should use it @albertvillanova or not ?
IMO it's ok to rely on magic numbers only for now, since in streaming mode we've had no issue checking only the magic number so far.
Close https://github.com/huggingface/datasets/issues/5639
|
closed
|
https://github.com/huggingface/datasets/pull/5640
| 2023-03-15T16:48:59
| 2023-03-16T13:47:37
| 2023-03-16T13:40:12
|
{
"login": "lhoestq",
"id": 42851186,
"type": "User"
}
|
[] | true
|
[] |
1,625,737,098
| 5,639
|
Parquet file wrongly recognized as zip prevents loading a dataset
|
### Describe the bug
When trying to `load_dataset_builder` for `HuggingFaceGECLM/StackExchange_Mar2023`, extraction fails, because parquet file [devops-00000-of-00001-22fe902fd8702892.parquet](https://huggingface.co/datasets/HuggingFaceGECLM/StackExchange_Mar2023/resolve/1f8c9a2ab6f7d0f9ae904b8b922e4384592ae1a5/data/devops-00000-of-00001-22fe902fd8702892.parquet) is wrongly identified by python as being a zip not a parquet.
(Full thread on [Slack](https://huggingface.slack.com/archives/C02V51Q3800/p1678890880803599))
### Steps to reproduce the bug
```python
from datasets import load_dataset_builder
ds = load_dataset_builder("HuggingFaceGECLM/StackExchange_Mar2023")
```
### Expected behavior
Loading the file normally.
### Environment info
- `datasets` version: 2.3.2
- Platform: Linux-5.14.0-1058-oem-x86_64-with-glibc2.29
- Python version: 3.8.10
- PyArrow version: 8.0.0
- Pandas version: 1.4.3
|
closed
|
https://github.com/huggingface/datasets/issues/5639
| 2023-03-15T15:20:45
| 2023-03-16T13:40:14
| 2023-03-16T13:40:14
|
{
"login": "clefourrier",
"id": 22726840,
"type": "User"
}
|
[] | false
|
[] |
1,625,564,471
| 5,638
|
xPath to implement all operations for Path
|
### Feature request
Current xPath implementation is a great extension of Path in order to work with remote objects. However some methods such as `mkdir` are not implemented correctly. It should instead rely on `fsspec` methods, instead of defaulting do `Path` methods which only work locally.
### Motivation
I'm using xPath to interact with remote objects.
### Your contribution
I could try to make a PR. I'm a bit unfamiliar with chaining right now.
|
closed
|
https://github.com/huggingface/datasets/issues/5638
| 2023-03-15T13:47:11
| 2023-03-17T13:21:12
| 2023-03-17T13:21:12
|
{
"login": "thomasw21",
"id": 24695242,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
}
] | false
|
[] |
1,625,295,691
| 5,637
|
IterableDataset with_format does not support 'device' keyword for jax
|
### Describe the bug
As seen here: https://huggingface.co/docs/datasets/use_with_jax dataset.with_format() supports the keyword 'device', to put data on a specific device when loaded as jax. However, when called on an IterableDataset, I got the error `TypeError: with_format() got an unexpected keyword argument 'device'`
Looking over the code, it seems IterableDataset support only pytorch and no support for jax device keyword?
https://github.com/huggingface/datasets/blob/fc5c84f36684343bff3e424cb0fd1ac5ecdd66da/src/datasets/iterable_dataset.py#L1029
### Steps to reproduce the bug
1. Load an IterableDataset (tested in streaming mode)
2. Call with_format('jax',device=device)
### Expected behavior
I expect to call `with_format('jax', device=device)` as per [documentation](https://huggingface.co/docs/datasets/use_with_jax) without error
### Environment info
Tested with installing newest (dev) and also pip release (2.10.1).
- `datasets` version: 2.10.2.dev0
- Platform: Linux-5.15.89+-x86_64-with-debian-bullseye-sid
- Python version: 3.7.12
- Huggingface_hub version: 0.12.1
- PyArrow version: 11.0.0
- Pandas version: 1.3.5
|
open
|
https://github.com/huggingface/datasets/issues/5637
| 2023-03-15T11:04:12
| 2025-01-07T06:59:33
| null |
{
"login": "Lime-Cakes",
"id": 91322985,
"type": "User"
}
|
[] | false
|
[] |
1,623,721,577
| 5,636
|
Fix CI: ignore C901 ("some_func" is to complex) in `ruff`
|
idk if I should have added this ignore to `ruff` too, but I added :)
|
closed
|
https://github.com/huggingface/datasets/pull/5636
| 2023-03-14T15:29:11
| 2023-03-14T16:37:06
| 2023-03-14T16:29:52
|
{
"login": "polinaeterna",
"id": 16348744,
"type": "User"
}
|
[] | true
|
[] |
1,623,682,558
| 5,635
|
Pass custom metadata filename to Image/Audio folders
|
This is a quick fix.
Now it requires to pass data via `data_files` parameters and include a required metadata file there and pass its filename as `metadata_filename` parameter.
For example, with the structure like:
```
data
images_dir/
im1.jpg
im2.jpg
...
metadata_dir/
meta_file1.jsonl
meta_file2.jsonl
...
```
to load data with `metadata_file1.jsonl` do:
```python
ds = load_dataset("imagefolder", data_files=["data/images_dir/**", "data/metadata_dir/meta_file1.jsonl"], metadata_filename="meta_file1.jsonl")
```
Note that if you have multiple splits, metadata file should be specified in each of them in `data_files`, smth like:
```python
data_files={
"train": ["data/train/**", "data/metadata_dir/meta_file1.jsonl"],
"test": ["data/train/**", "data/metadata_dir/meta_file1.jsonl"]
}
```
|
open
|
https://github.com/huggingface/datasets/pull/5635
| 2023-03-14T15:08:16
| 2023-03-22T17:50:31
| null |
{
"login": "polinaeterna",
"id": 16348744,
"type": "User"
}
|
[] | true
|
[] |
1,622,424,174
| 5,634
|
Not all progress bars are showing up when they should for downloading dataset
|
### Describe the bug
During downloading the rotten tomatoes dataset, not all progress bars are displayed properly. This might be related to [this ticket](https://github.com/huggingface/datasets/issues/5117) as it raised the same concern but its not clear if the fix solves this issue too.
ipywidgets
<img width="1243" alt="image" src="https://user-images.githubusercontent.com/110427462/224851138-13fee5b7-ab51-4883-b96f-1b9808782e3b.png">
tqdm
<img width="1251" alt="Screen Shot 2023-03-13 at 3 58 59 PM" src="https://user-images.githubusercontent.com/110427462/224851180-5feb7825-9250-4b1e-ad0c-f3172ac1eb78.png">
### Steps to reproduce the bug
1. Run this line
```
from datasets import load_dataset
rotten_tomatoes = load_dataset("rotten_tomatoes", split="train")
```
### Expected behavior
all progress bars for builder script, metadata, readme, training, validation, and test set
### Environment info
requirements.txt
```
aiofiles==22.1.0
aiohttp==3.8.4
aiosignal==1.3.1
aiosqlite==0.18.0
anyio==3.6.2
appnope==0.1.3
argon2-cffi==21.3.0
argon2-cffi-bindings==21.2.0
arrow==1.2.3
asttokens==2.2.1
async-generator==1.10
async-timeout==4.0.2
attrs==22.2.0
Babel==2.12.1
backcall==0.2.0
beautifulsoup4==4.11.2
bleach==6.0.0
brotlipy @ file:///Users/runner/miniforge3/conda-bld/brotlipy_1666764961872/work
certifi==2022.12.7
cffi @ file:///Users/runner/miniforge3/conda-bld/cffi_1671179414629/work
cfgv==3.3.1
charset-normalizer @ file:///home/conda/feedstock_root/build_artifacts/charset-normalizer_1661170624537/work
comm==0.1.2
conda==22.9.0
conda-package-handling @ file:///home/conda/feedstock_root/build_artifacts/conda-package-handling_1669907009957/work
conda_package_streaming @ file:///home/conda/feedstock_root/build_artifacts/conda-package-streaming_1669733752472/work
coverage==7.2.1
cryptography @ file:///Users/runner/miniforge3/conda-bld/cryptography_1669592251328/work
datasets==2.1.0
debugpy==1.6.6
decorator==5.1.1
defusedxml==0.7.1
dill==0.3.6
distlib==0.3.6
distro==1.4.0
entrypoints==0.4
exceptiongroup==1.1.0
executing==1.2.0
fastjsonschema==2.16.3
filelock==3.9.0
flaky==3.7.0
fqdn==1.5.1
frozenlist==1.3.3
fsspec==2023.3.0
huggingface-hub==0.10.1
identify==2.5.18
idna @ file:///home/conda/feedstock_root/build_artifacts/idna_1663625384323/work
iniconfig==2.0.0
ipykernel==6.12.1
ipyparallel==8.4.1
ipython==7.32.0
ipython-genutils==0.2.0
ipywidgets==8.0.4
isoduration==20.11.0
jedi==0.18.2
Jinja2==3.1.2
json5==0.9.11
jsonpointer==2.3
jsonschema==4.17.3
jupyter-events==0.6.3
jupyter-ydoc==0.2.2
jupyter_client==8.0.3
jupyter_core==5.2.0
jupyter_server==2.4.0
jupyter_server_fileid==0.8.0
jupyter_server_terminals==0.4.4
jupyter_server_ydoc==0.6.1
jupyterlab==3.6.1
jupyterlab-pygments==0.2.2
jupyterlab-widgets==3.0.5
jupyterlab_server==2.20.0
libmambapy @ file:///Users/runner/miniforge3/conda-bld/mamba-split_1671598370072/work/libmambapy
mamba @ file:///Users/runner/miniforge3/conda-bld/mamba-split_1671598370072/work/mamba
MarkupSafe==2.1.2
matplotlib-inline==0.1.6
mistune==2.0.5
multidict==6.0.4
multiprocess==0.70.14
nbclassic==0.5.3
nbclient==0.7.2
nbconvert==7.2.9
nbformat==5.7.3
nest-asyncio==1.5.6
nodeenv==1.7.0
notebook==6.5.3
notebook_shim==0.2.2
numpy==1.24.2
outcome==1.2.0
packaging==23.0
pandas==1.5.3
pandocfilters==1.5.0
parso==0.8.3
pexpect==4.8.0
pickleshare==0.7.5
platformdirs==3.0.0
plotly==5.13.1
pluggy==1.0.0
pre-commit==3.1.0
prometheus-client==0.16.0
prompt-toolkit==3.0.38
psutil==5.9.4
ptyprocess==0.7.0
pure-eval==0.2.2
pyarrow==11.0.0
pycosat @ file:///Users/runner/miniforge3/conda-bld/pycosat_1666836580084/work
pycparser @ file:///home/conda/feedstock_root/build_artifacts/pycparser_1636257122734/work
Pygments==2.14.0
pyOpenSSL @ file:///home/conda/feedstock_root/build_artifacts/pyopenssl_1665350324128/work
pyrsistent==0.19.3
PySocks @ file:///home/conda/feedstock_root/build_artifacts/pysocks_1661604839144/work
pytest==7.2.1
pytest-asyncio==0.20.3
pytest-cov==4.0.0
pytest-timeout==2.1.0
python-dateutil==2.8.2
python-json-logger==2.0.7
pytz==2022.7.1
PyYAML==6.0
pyzmq==25.0.0
requests @ file:///home/conda/feedstock_root/build_artifacts/requests_1661872987712/work
responses==0.18.0
rfc3339-validator==0.1.4
rfc3986-validator==0.1.1
ruamel-yaml-conda @ file:///Users/runner/miniforge3/conda-bld/ruamel_yaml_1666819760545/work
Send2Trash==1.8.0
simplegeneric==0.8.1
six==1.16.0
sniffio==1.3.0
sortedcontainers==2.4.0
soupsieve==2.4
stack-data==0.6.2
tenacity==8.2.2
terminado==0.17.1
tinycss2==1.2.1
tomli==2.0.1
toolz @ file:///home/conda/feedstock_root/build_artifacts/toolz_1657485559105/work
tornado==6.2
tqdm==4.64.1
traitlets==5.8.1
trio==0.22.0
typing_extensions==4.5.0
uri-template==1.2.0
urllib3 @ file:///home/conda/feedstock_root/build_artifacts/urllib3_1669259737463/work
virtualenv==20.19.0
wcwidth==0.2.6
webcolors==1.12
webencodings==0.5.1
websocket-client==1.5.1
widgetsnbextension==4.0.5
xxhash==3.2.0
y-py==0.5.9
yarl==1.8.2
ypy-websocket==0.8.2
zstandard==0.19.0
```
|
closed
|
https://github.com/huggingface/datasets/issues/5634
| 2023-03-13T23:04:18
| 2023-10-11T16:30:16
| 2023-10-11T16:30:16
|
{
"login": "garlandz-db",
"id": 110427462,
"type": "User"
}
|
[] | false
|
[] |
1,621,469,970
| 5,633
|
Cannot import datasets
|
### Describe the bug
Hi,
I cannot even import the library :( I installed it by running:
```
$ conda install datasets
```
Then I realized I should maybe use the huggingface channel, because I encountered the error below, so I ran:
```
$ conda remove datasets
$ conda install -c huggingface datasets
```
Please see 'steps to reproduce the bug' for the specific error, as steps to reproduce is just importing the library
### Steps to reproduce the bug
```
$ python3
Python 3.8.15 (default, Nov 24 2022, 15:19:38)
[GCC 11.2.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import datasets
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/jack/.conda/envs/jack_zpp/lib/python3.8/site-packages/datasets/__init__.py", line 33, in <module>
from .arrow_dataset import Dataset, concatenate_datasets
File "/home/jack/.conda/envs/jack_zpp/lib/python3.8/site-packages/datasets/arrow_dataset.py", line 59, in <module>
from .arrow_reader import ArrowReader
File "/home/jack/.conda/envs/jack_zpp/lib/python3.8/site-packages/datasets/arrow_reader.py", line 27, in <module>
import pyarrow.parquet as pq
File "/home/jack/.conda/envs/jack_zpp/lib/python3.8/site-packages/pyarrow/parquet/__init__.py", line 20, in <module>
from .core import *
File "/home/jack/.conda/envs/jack_zpp/lib/python3.8/site-packages/pyarrow/parquet/core.py", line 37, in <module>
from pyarrow._parquet import (ParquetReader, Statistics, # noqa
ImportError: cannot import name 'FileEncryptionProperties' from 'pyarrow._parquet' (/home/jack/.conda/envs/jack_zpp/lib/python3.8/site-packages/pyarrow/_parquet.cpython-38-x86_64-linux-gnu.so)
```
### Expected behavior
I would expect for the statement `import datasets` to cause no error
### Environment info
Output of `conda list`:
```
# packages in environment at /home/jack/.conda/envs/pbalawender_zpp:
#
# Name Version Build Channel
_libgcc_mutex 0.1 main
_openmp_mutex 5.1 1_gnu
abseil-cpp 20210324.2 h2531618_0
advertools 0.13.2 pypi_0 pypi
aiofiles 0.8.0 pypi_0 pypi
aiohttp 3.8.3 py38h5eee18b_0
aiosignal 1.2.0 pyhd3eb1b0_0
aiosqlite 0.17.0 pypi_0 pypi
anyio 3.6.2 pypi_0 pypi
aquirdturtle-collapsible-headings 3.1.0 pypi_0 pypi
argon2-cffi 21.3.0 pypi_0 pypi
argon2-cffi-bindings 21.2.0 pypi_0 pypi
arrow 1.2.3 pypi_0 pypi
arrow-cpp 3.0.0 py38h6b21186_4
asttokens 2.2.0 pypi_0 pypi
async-timeout 4.0.2 py38h06a4308_0
attrs 22.1.0 py38h06a4308_0
automat 22.10.0 pypi_0 pypi
aws-c-common 0.4.57 he6710b0_1
aws-c-event-stream 0.1.6 h2531618_5
aws-checksums 0.1.9 he6710b0_0
aws-sdk-cpp 1.8.185 hce553d0_0
babel 2.11.0 pypi_0 pypi
backcall 0.2.0 pyhd3eb1b0_0
beautifulsoup4 4.11.1 pypi_0 pypi
blas 1.0 mkl
bleach 5.0.1 pypi_0 pypi
boost-cpp 1.73.0 h27cfd23_11
bottleneck 1.3.5 py38h7deecbd_0
brotli 1.0.9 h5eee18b_7
brotli-bin 1.0.9 h5eee18b_7
brotlipy 0.7.0 py38h27cfd23_1003
bzip2 1.0.8 h7b6447c_0
c-ares 1.18.1 h7f8727e_0
ca-certificates 2023.01.10 h06a4308_0
certifi 2022.9.24 pypi_0 pypi
cffi 1.15.1 py38h5eee18b_3
charset-normalizer 2.1.1 pypi_0 pypi
click 8.1.3 pypi_0 pypi
constantly 15.1.0 pypi_0 pypi
contourpy 1.0.6 pypi_0 pypi
cryptography 38.0.4 pypi_0 pypi
cssselect 1.2.0 pypi_0 pypi
cudatoolkit 10.1.243 h8cb64d8_10 conda-forge
cycler 0.11.0 pypi_0 pypi
dacite 1.6.0 pypi_0 pypi
dataclasses 0.8 pyh6d0b6a4_7
datasets 1.18.4 py_0 huggingface
datetime 4.7 pypi_0 pypi
debugpy 1.6.4 pypi_0 pypi
decorator 5.1.1 pyhd3eb1b0_0
defusedxml 0.7.1 pypi_0 pypi
dill 0.3.6 py38h06a4308_0
docker-pycreds 0.4.0 pypi_0 pypi
double-conversion 3.1.5 he6710b0_1
entrypoints 0.4 py38h06a4308_0
executing 0.8.3 pyhd3eb1b0_0
filelock 3.8.0 pypi_0 pypi
flake8 6.0.0 pypi_0 pypi
flask 2.1.3 py38h06a4308_0
flit-core 3.6.0 pyhd3eb1b0_0
fonttools 4.38.0 pypi_0 pypi
fqdn 1.5.1 pypi_0 pypi
freetype 2.12.1 h4a9f257_0
frozenlist 1.3.3 py38h5eee18b_0
fsspec 2022.11.0 py38h06a4308_0
gensim 4.2.0 pypi_0 pypi
gflags 2.2.2 he6710b0_0
giflib 5.2.1 h5eee18b_3
gitdb 4.0.10 pypi_0 pypi
gitpython 3.1.30 pypi_0 pypi
glog 0.5.0 h2531618_0
grpc-cpp 1.39.0 hae934f6_5
huggingface-hub 0.11.1 pypi_0 pypi
huggingface_hub 0.13.1 py_0 huggingface
hyperlink 21.0.0 pypi_0 pypi
icu 58.2 he6710b0_3
idna 3.4 py38h06a4308_0
importlib-metadata 5.1.0 pypi_0 pypi
importlib_metadata 4.11.3 hd3eb1b0_0
importlib_resources 5.2.0 pyhd3eb1b0_1
incremental 22.10.0 pypi_0 pypi
intel-openmp 2021.4.0 h06a4308_3561
ipykernel 6.17.1 pyh210e3f2_0 conda-forge
ipython 8.7.0 pypi_0 pypi
ipython-genutils 0.2.0 pypi_0 pypi
ipywidgets 8.0.2 pyhd8ed1ab_1 conda-forge
isoduration 20.11.0 pypi_0 pypi
itemadapter 0.7.0 pypi_0 pypi
itemloaders 1.0.6 pypi_0 pypi
itsdangerous 2.0.1 pyhd3eb1b0_0
jedi 0.18.2 pypi_0 pypi
jinja2 3.1.2 py38h06a4308_0
jmespath 1.0.1 pypi_0 pypi
joblib 1.2.0 pypi_0 pypi
jpeg 9b h024ee3a_2
json5 0.9.10 pypi_0 pypi
jsonpickle 3.0.0 pypi_0 pypi
jsonpointer 2.3 pypi_0 pypi
jsonschema 4.17.3 py38h06a4308_0
jupyter-core 5.1.0 pypi_0 pypi
jupyter-events 0.5.0 pypi_0 pypi
jupyter-server 1.23.3 pypi_0 pypi
jupyter-server-fileid 0.6.0 pypi_0 pypi
jupyter-server-ydoc 0.4.0 pypi_0 pypi
jupyter-ydoc 0.2.2 pypi_0 pypi
jupyter_client 7.4.9 py38h06a4308_0
jupyter_core 5.2.0 py38h06a4308_0
jupyterlab 3.6.0a4 pypi_0 pypi
jupyterlab-pygments 0.2.2 pypi_0 pypi
jupyterlab-server 2.16.3 pypi_0 pypi
jupyterlab_widgets 3.0.3 pyhd8ed1ab_0 conda-forge
kiwisolver 1.4.4 pypi_0 pypi
krb5 1.19.4 h568e23c_0
lcms2 2.12 h3be6417_0
ld_impl_linux-64 2.38 h1181459_1
libboost 1.73.0 h3ff78a5_11
libbrotlicommon 1.0.9 h5eee18b_7
libbrotlidec 1.0.9 h5eee18b_7
libbrotlienc 1.0.9 h5eee18b_7
libcurl 7.88.1 h91b91d3_0
libedit 3.1.20221030 h5eee18b_0
libev 4.33 h7f8727e_1
libevent 2.1.12 h8f2d780_0
libffi 3.4.2 h6a678d5_6
libgcc-ng 11.2.0 h1234567_1
libgomp 11.2.0 h1234567_1
libnghttp2 1.46.0 hce63b2e_0
libpng 1.6.39 h5eee18b_0
libprotobuf 3.17.2 h4ff587b_1
libsodium 1.0.18 h7b6447c_0
libssh2 1.10.0 h8f2d780_0
libstdcxx-ng 11.2.0 h1234567_1
libthrift 0.14.2 hcc01f38_0
libtiff 4.1.0 h2733197_1
libuv 1.44.2 h5eee18b_0
libwebp 1.2.0 h89dd481_0
lz4-c 1.9.4 h6a678d5_0
markupsafe 2.1.1 py38h7f8727e_0
matplotlib 3.6.2 pypi_0 pypi
matplotlib-inline 0.1.6 py38h06a4308_0
mccabe 0.7.0 pypi_0 pypi
mistune 2.0.4 pypi_0 pypi
mkl 2021.4.0 h06a4308_640
mkl-service 2.4.0 py38h7f8727e_0
mkl_fft 1.3.1 py38hd3c417c_0
mkl_random 1.2.2 py38h51133e4_0
morfeusz2 1.99.6 pypi_0 pypi
multidict 6.0.2 py38h5eee18b_0
multiprocess 0.70.14 py38h06a4308_0
nbclassic 0.4.8 pypi_0 pypi
nbclient 0.7.2 pypi_0 pypi
nbconvert 7.2.5 pypi_0 pypi
nbformat 5.7.0 py38h06a4308_0
ncurses 6.4 h6a678d5_0
nest-asyncio 1.5.6 py38h06a4308_0
ninja 1.10.2 h06a4308_5
ninja-base 1.10.2 hd09550d_5
notebook 6.5.2 pypi_0 pypi
notebook-shim 0.2.2 pypi_0 pypi
numexpr 2.8.4 py38he184ba9_0
numpy 1.23.5 py38h14f4228_0
numpy-base 1.23.5 py38h31eccc5_0
oauthlib 3.2.2 pypi_0 pypi
opencv-python 4.6.0.66 pypi_0 pypi
openssl 1.1.1t h7f8727e_0
orc 1.6.9 ha97a36c_3
packaging 22.0 py38h06a4308_0
pandas 1.5.2 pypi_0 pypi
pandocfilters 1.5.0 pypi_0 pypi
parsel 1.7.0 pypi_0 pypi
parso 0.8.3 pyhd3eb1b0_0
pathlib 1.0.1 pypi_0 pypi
pathtools 0.1.2 pypi_0 pypi
pexpect 4.8.0 pyhd3eb1b0_3
pickleshare 0.7.5 pyhd3eb1b0_1003
pillow 9.3.0 pypi_0 pypi
pip 22.2.2 py38h06a4308_0
pkgutil-resolve-name 1.3.10 py38h06a4308_0
platformdirs 2.5.4 pypi_0 pypi
prometheus-client 0.15.0 pypi_0 pypi
promise 2.3 pypi_0 pypi
prompt-toolkit 3.0.33 pypi_0 pypi
protego 0.2.1 pypi_0 pypi
protobuf 4.21.12 pypi_0 pypi
psutil 5.9.0 py38h5eee18b_0
ptyprocess 0.7.0 pyhd3eb1b0_2
pure_eval 0.2.2 pyhd3eb1b0_0
pyarrow 10.0.1 pypi_0 pypi
pyasn1 0.4.8 pypi_0 pypi
pyasn1-modules 0.2.8 pypi_0 pypi
pycodestyle 2.10.0 pypi_0 pypi
pycparser 2.21 pyhd3eb1b0_0
pydispatcher 2.0.6 pypi_0 pypi
pyflakes 3.0.1 pypi_0 pypi
pygments 2.11.2 pyhd3eb1b0_0
pyopenssl 22.1.0 pypi_0 pypi
pyrsistent 0.18.0 py38heee7806_0
pysocks 1.7.1 py38h06a4308_0
python 3.8.15 h7a1cb2a_2
python-dateutil 2.8.2 pyhd3eb1b0_0
python-dotenv 0.21.0 pypi_0 pypi
python-fastjsonschema 2.16.2 py38h06a4308_0
python-json-logger 2.0.4 pypi_0 pypi
python-xxhash 2.0.2 py38h5eee18b_1
pytorch 1.7.1 py3.8_cuda10.1.243_cudnn7.6.3_0 pytorch
pytz 2022.6 pypi_0 pypi
pyyaml 6.0 py38h5eee18b_1
pyzmq 23.2.0 py38h6a678d5_0
queuelib 1.6.2 pypi_0 pypi
re2 2022.04.01 h295c915_0
readline 8.2 h5eee18b_0
regex 2022.10.31 pypi_0 pypi
requests 2.28.1 py38h06a4308_0
requests-file 1.5.1 pypi_0 pypi
requests-oauthlib 1.3.1 pypi_0 pypi
rfc3339-validator 0.1.4 pypi_0 pypi
rfc3986-validator 0.1.1 pypi_0 pypi
scikit-learn 1.1.3 pypi_0 pypi
scipy 1.9.3 pypi_0 pypi
scrapy 2.7.1 pypi_0 pypi
seaborn 0.12.1 pypi_0 pypi
send2trash 1.8.0 pypi_0 pypi
sentry-sdk 1.12.1 pypi_0 pypi
service-identity 21.1.0 pypi_0 pypi
setproctitle 1.3.2 pypi_0 pypi
setuptools 65.6.3 pypi_0 pypi
shortuuid 1.0.11 pypi_0 pypi
six 1.16.0 pyhd3eb1b0_1
smart-open 6.2.0 pypi_0 pypi
smmap 5.0.0 pypi_0 pypi
snappy 1.1.9 h295c915_0
sniffio 1.3.0 pypi_0 pypi
soupsieve 2.3.2.post1 pypi_0 pypi
sqlite 3.40.1 h5082296_0
stack-data 0.6.2 pypi_0 pypi
stack_data 0.2.0 pyhd3eb1b0_0
terminado 0.17.0 pypi_0 pypi
threadpoolctl 3.1.0 pypi_0 pypi
tinycss2 1.2.1 pypi_0 pypi
tk 8.6.12 h1ccaba5_0
tldextract 3.4.0 pypi_0 pypi
tokenizers 0.13.2 pypi_0 pypi
tomli 2.0.1 pypi_0 pypi
torchvision 0.8.2 py38_cu101 pytorch
tornado 6.2 py38h5eee18b_0
tqdm 4.64.1 py38h06a4308_0
traitlets 5.6.0 pypi_0 pypi
transformers 4.25.1 pypi_0 pypi
tweepy 4.12.1 pypi_0 pypi
twisted 22.10.0 pypi_0 pypi
twython 3.9.1 pypi_0 pypi
typing-extensions 4.4.0 py38h06a4308_0
typing_extensions 4.4.0 py38h06a4308_0
uri-template 1.2.0 pypi_0 pypi
uriparser 0.9.3 he6710b0_1
urllib3 1.26.13 pypi_0 pypi
utf8proc 2.6.1 h27cfd23_0
w3lib 2.1.0 pypi_0 pypi
wandb 0.13.7 pypi_0 pypi
wcwidth 0.2.5 pyhd3eb1b0_0
webcolors 1.12 pypi_0 pypi
webencodings 0.5.1 pypi_0 pypi
websocket-client 1.4.2 pypi_0 pypi
werkzeug 2.2.2 py38h06a4308_0
wheel 0.38.4 py38h06a4308_0
widgetsnbextension 4.0.3 py38h06a4308_0
xxhash 0.8.0 h7f8727e_3
xz 5.2.10 h5eee18b_1
y-py 0.5.4 pypi_0 pypi
yaml 0.2.5 h7b6447c_0
yarl 1.8.1 py38h5eee18b_0
ypy-websocket 0.5.0 pypi_0 pypi
zeromq 4.3.4 h2531618_0
zipp 3.11.0 py38h06a4308_0
zlib 1.2.13 h5eee18b_0
zope-interface 5.5.2 pypi_0 pypi
zstd 1.4.9 haebb681_0
```
|
closed
|
https://github.com/huggingface/datasets/issues/5633
| 2023-03-13T13:14:44
| 2023-03-13T17:54:19
| 2023-03-13T17:54:19
|
{
"login": "ruplet",
"id": 11250555,
"type": "User"
}
|
[] | false
|
[] |
1,621,177,391
| 5,632
|
Dataset cannot convert too large dictionnary
|
### Describe the bug
Hello everyone!
I tried to build a new dataset with the command "dict_valid = datasets.Dataset.from_dict({'input_values': values_array})".
However, I have a very large dataset (~400Go) and it seems that dataset cannot handle this.
Indeed, I can create the dataset until a certain size of my dictionnary, and then I have the error "OverflowError: Python int too large to convert to C long".
Do you know how to solve this problem?
Unfortunately I cannot give a reproductible code because I cannot share a so large file, but you can find the code below (it's a test on only a part of the validation data ~10Go, but it's already the case).
Thank you!
### Steps to reproduce the bug
SAVE_DIR = './data/'
features = h5py.File(SAVE_DIR+'features.hdf5','r')
valid_data = features["validation"]["data/features"]
v_array_values = [np.float32(item[()]) for item in valid_data.values()]
for i in range(len(v_array_values)):
v_array_values[i] = v_array_values[i].round(decimals=5)
dict_valid = datasets.Dataset.from_dict({'input_values': v_array_values})
### Expected behavior
The code is expected to give me a Huggingface dataset.
### Environment info
python: 3.8.15
numpy: 1.22.3
datasets: 2.3.2
pyarrow: 8.0.0
|
open
|
https://github.com/huggingface/datasets/issues/5632
| 2023-03-13T10:14:40
| 2023-03-16T15:28:57
| null |
{
"login": "MaraLac",
"id": 108518627,
"type": "User"
}
|
[] | false
|
[] |
1,620,442,854
| 5,631
|
Custom split names
|
### Feature request
Hi,
I participated in multiple NLP tasks where there are more than just train, test, validation splits, there could be multiple validation sets or test sets. But it seems currently only those mentioned three splits supported. It would be nice to have the support for more splits on the hub. (currently i can have more splits when I am loading datasets from urls, but not hub)
### Motivation
Easier access to more splits
### Your contribution
No
|
closed
|
https://github.com/huggingface/datasets/issues/5631
| 2023-03-12T17:21:43
| 2023-03-24T14:13:00
| 2023-03-24T14:13:00
|
{
"login": "ErfanMoosaviMonazzah",
"id": 79091831,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
}
] | false
|
[] |
1,620,327,510
| 5,630
|
adds early exit if url is `PathLike`
|
Closes #4864
Should fix errors thrown when attempting to load `json` dataset using `pathlib.Path` in `data_files` argument.
|
open
|
https://github.com/huggingface/datasets/pull/5630
| 2023-03-12T11:23:28
| 2023-03-15T11:58:38
| null |
{
"login": "vvvm23",
"id": 44398246,
"type": "User"
}
|
[] | true
|
[] |
1,619,921,247
| 5,629
|
load_dataset gives "403" error when using Financial phrasebank
|
When I try to load this dataset, I receive the following error:
ConnectionError: Couldn't reach https://www.researchgate.net/profile/Pekka_Malo/publication/251231364_FinancialPhraseBank-v10/data/0c96051eee4fb1d56e000000/FinancialPhraseBank-v10.zip (error 403)
Has this been seen before? Thanks. The website loads when I try to access it manually.
|
open
|
https://github.com/huggingface/datasets/issues/5629
| 2023-03-11T07:46:39
| 2023-03-13T18:27:26
| null |
{
"login": "Jimchoo91",
"id": 67709789,
"type": "User"
}
|
[] | false
|
[] |
1,619,641,810
| 5,628
|
add kwargs to index search
|
This PR proposes to add kwargs to index search methods.
This is particularly useful for setting the timeout of a query on elasticsearch.
A typical use case would be:
```python
dset.add_elasticsearch_index("filename", es_client=es_client)
scores, examples = dset.get_nearest_examples("filename", "my_name-train_29", request_timeout=60)
```
|
closed
|
https://github.com/huggingface/datasets/pull/5628
| 2023-03-10T21:24:58
| 2023-03-15T14:48:47
| 2023-03-15T14:46:04
|
{
"login": "SaulLu",
"id": 55560583,
"type": "User"
}
|
[] | true
|
[] |
1,619,336,609
| 5,627
|
Unable to load AutoTrain-generated dataset from the hub
|
### Describe the bug
DatasetGenerationError: An error occurred while generating the dataset -> ValueError: Couldn't cast ... because column names don't match
```
ValueError: Couldn't cast
_data_files: list<item: struct<filename: string>>
child 0, item: struct<filename: string>
child 0, filename: string
_fingerprint: string
_format_columns: list<item: string>
child 0, item: string
_format_kwargs: struct<>
_format_type: null
_indexes: struct<>
_output_all_columns: bool
_split: null
to
{'citation': Value(dtype='string', id=None), 'description': Value(dtype='string', id=None), 'features': {'image': {'_type': Value(dtype='string', id=None)}, 'target': {'names': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None), '_type': Value(dtype='string', id=None)}}, 'homepage': Value(dtype='string', id=None), 'license': Value(dtype='string', id=None), 'splits': {'train': {'name': Value(dtype='string', id=None), 'num_bytes': Value(dtype='int64', id=None), 'num_examples': Value(dtype='int64', id=None), 'dataset_name': Value(dtype='null', id=None)}}}
because column names don't match
```
### Steps to reproduce the bug
Steps to reproduce:
1. `pip install datasets==2.10.1`
2. Attempt to load (private dataset). Note that I'm authenticated via ` huggingface-cli login`
```
from datasets import load_dataset
# load dataset
dataset = "ijmiller2/autotrain-data-betterbin-vision-10000"
dataset = load_dataset(dataset)
```
Here's the full traceback:
```Downloading and preparing dataset json/ijmiller2--autotrain-data-betterbin-vision-10000 to /Users/ian/.cache/huggingface/datasets/ijmiller2___json/ijmiller2--autotrain-data-betterbin-vision-10000-2eae034a9ff8a1a9/0.0.0/0f7e3662623656454fcd2b650f34e886a7db4b9104504885bd462096cc7a9f51...
Downloading data files: 100%|███████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 2383.80it/s]
Extracting data files: 100%|█████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 505.95it/s]
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
File ~/anaconda3/envs/betterbin/lib/python3.8/site-packages/datasets/builder.py:1874, in ArrowBasedBuilder._prepare_split_single(self, gen_kwargs, fpath, file_format, max_shard_size, job_id)
1868 writer = writer_class(
1869 features=writer._features,
1870 path=fpath.replace("SSSSS", f"{shard_id:05d}").replace("JJJJJ", f"{job_id:05d}"),
1871 storage_options=self._fs.storage_options,
1872 embed_local_files=embed_local_files,
1873 )
-> 1874 writer.write_table(table)
1875 num_examples_progress_update += len(table)
File ~/anaconda3/envs/betterbin/lib/python3.8/site-packages/datasets/arrow_writer.py:568, in ArrowWriter.write_table(self, pa_table, writer_batch_size)
567 pa_table = pa_table.combine_chunks()
--> 568 pa_table = table_cast(pa_table, self._schema)
569 if self.embed_local_files:
File ~/anaconda3/envs/betterbin/lib/python3.8/site-packages/datasets/table.py:2312, in table_cast(table, schema)
2311 if table.schema != schema:
-> 2312 return cast_table_to_schema(table, schema)
2313 elif table.schema.metadata != schema.metadata:
File ~/anaconda3/envs/betterbin/lib/python3.8/site-packages/datasets/table.py:2270, in cast_table_to_schema(table, schema)
2269 if sorted(table.column_names) != sorted(features):
-> 2270 raise ValueError(f"Couldn't cast\n{table.schema}\nto\n{features}\nbecause column names don't match")
2271 arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]
ValueError: Couldn't cast
_data_files: list<item: struct<filename: string>>
child 0, item: struct<filename: string>
child 0, filename: string
_fingerprint: string
_format_columns: list<item: string>
child 0, item: string
_format_kwargs: struct<>
_format_type: null
_indexes: struct<>
_output_all_columns: bool
_split: null
to
{'citation': Value(dtype='string', id=None), 'description': Value(dtype='string', id=None), 'features': {'image': {'_type': Value(dtype='string', id=None)}, 'target': {'names': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None), '_type': Value(dtype='string', id=None)}}, 'homepage': Value(dtype='string', id=None), 'license': Value(dtype='string', id=None), 'splits': {'train': {'name': Value(dtype='string', id=None), 'num_bytes': Value(dtype='int64', id=None), 'num_examples': Value(dtype='int64', id=None), 'dataset_name': Value(dtype='null', id=None)}}}
because column names don't match
The above exception was the direct cause of the following exception:
DatasetGenerationError Traceback (most recent call last)
Input In [8], in <cell line: 6>()
4 # load dataset
5 dataset = "ijmiller2/autotrain-data-betterbin-vision-10000"
----> 6 dataset = load_dataset(dataset)
File ~/anaconda3/envs/betterbin/lib/python3.8/site-packages/datasets/load.py:1782, 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, **config_kwargs)
1779 try_from_hf_gcs = path not in _PACKAGED_DATASETS_MODULES
1781 # Download and prepare data
-> 1782 builder_instance.download_and_prepare(
1783 download_config=download_config,
1784 download_mode=download_mode,
1785 verification_mode=verification_mode,
1786 try_from_hf_gcs=try_from_hf_gcs,
1787 num_proc=num_proc,
1788 )
1790 # Build dataset for splits
1791 keep_in_memory = (
1792 keep_in_memory if keep_in_memory is not None else is_small_dataset(builder_instance.info.dataset_size)
1793 )
File ~/anaconda3/envs/betterbin/lib/python3.8/site-packages/datasets/builder.py:872, 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)
870 if num_proc is not None:
871 prepare_split_kwargs["num_proc"] = num_proc
--> 872 self._download_and_prepare(
873 dl_manager=dl_manager,
874 verification_mode=verification_mode,
875 **prepare_split_kwargs,
876 **download_and_prepare_kwargs,
877 )
878 # Sync info
879 self.info.dataset_size = sum(split.num_bytes for split in self.info.splits.values())
File ~/anaconda3/envs/betterbin/lib/python3.8/site-packages/datasets/builder.py:967, in DatasetBuilder._download_and_prepare(self, dl_manager, verification_mode, **prepare_split_kwargs)
963 split_dict.add(split_generator.split_info)
965 try:
966 # Prepare split will record examples associated to the split
--> 967 self._prepare_split(split_generator, **prepare_split_kwargs)
968 except OSError as e:
969 raise OSError(
970 "Cannot find data file. "
971 + (self.manual_download_instructions or "")
972 + "\nOriginal error:\n"
973 + str(e)
974 ) from None
File ~/anaconda3/envs/betterbin/lib/python3.8/site-packages/datasets/builder.py:1749, in ArrowBasedBuilder._prepare_split(self, split_generator, file_format, num_proc, max_shard_size)
1747 job_id = 0
1748 with pbar:
-> 1749 for job_id, done, content in self._prepare_split_single(
1750 gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
1751 ):
1752 if done:
1753 result = content
File ~/anaconda3/envs/betterbin/lib/python3.8/site-packages/datasets/builder.py:1892, in ArrowBasedBuilder._prepare_split_single(self, gen_kwargs, fpath, file_format, max_shard_size, job_id)
1890 if isinstance(e, SchemaInferenceError) and e.__context__ is not None:
1891 e = e.__context__
-> 1892 raise DatasetGenerationError("An error occurred while generating the dataset") from e
1894 yield job_id, True, (total_num_examples, total_num_bytes, writer._features, num_shards, shard_lengths)
DatasetGenerationError: An error occurred while generating the dataset
```
### Expected behavior
I'm ultimately trying to generate my own performance metrics on validation data (before putting an endpoint into production) and so was hoping to load all or at least the validation subset from the hub.
I'm expecting the `load_dataset()` function to work as shown in the documentation [here](https://huggingface.co/docs/datasets/loading#hugging-face-hub):
```python
dataset = load_dataset(
"lhoestq/custom_squad",
revision="main" # tag name, or branch name, or commit hash
)
```
### Environment info
- `datasets` version: 2.10.1
- Platform: macOS-13.2.1-arm64-arm-64bit
- Python version: 3.8.13
- PyArrow version: 9.0.0
- Pandas version: 1.4.4
|
open
|
https://github.com/huggingface/datasets/issues/5627
| 2023-03-10T17:25:58
| 2023-03-11T15:44:42
| null |
{
"login": "ijmiller2",
"id": 8560151,
"type": "User"
}
|
[] | false
|
[] |
1,619,252,984
| 5,626
|
Support streaming datasets with numpy.load
|
Support streaming datasets with `numpy.load`.
See: https://huggingface.co/datasets/qgallouedec/gia_dataset/discussions/1
|
closed
|
https://github.com/huggingface/datasets/pull/5626
| 2023-03-10T16:33:39
| 2023-03-21T06:36:05
| 2023-03-21T06:28:54
|
{
"login": "albertvillanova",
"id": 8515462,
"type": "User"
}
|
[] | true
|
[] |
1,618,971,855
| 5,625
|
Allow "jsonl" data type signifier
|
### Feature request
`load_dataset` currently does not accept `jsonl` as type but only `json`.
### Motivation
I was working with one of the `run_translation` scripts and used my own datasets (`.jsonl`) as train_dataset. But the default code did not work because
```
FileNotFoundError: Couldn't find a dataset script at jsonl\jsonl.py or any data file in the same directory. Couldn't find 'jsonl' on the Hugging Face Hub either: FileNotFoundError: Dataset 'jsonl' doesn't exist on the Hub. If the repo is private or gated, make sure to log in with `huggingface-cli login`.
```
The reason is because the script has these lines to extract the data type by its extension. Therefore, the derived type is `jsonl` which is not recognized by datasets as the error above shows.
https://github.com/huggingface/transformers/blob/ade26bf9912f69e2110137443e4406d7dbe253e7/examples/pytorch/translation/run_translation.py#L342-L356
I suppose you could argue that this is the script's fault (in which case I'll do a PR over at `transformers`) but it makes sense to me to add `jsonl` as an alias to `json` in `datasets`.
### Your contribution
At the moment I cannot work on this. I think it can be as "easy" as having an alias for json, namely jsonl.
|
open
|
https://github.com/huggingface/datasets/issues/5625
| 2023-03-10T13:21:48
| 2023-03-11T10:35:39
| null |
{
"login": "BramVanroy",
"id": 2779410,
"type": "User"
}
|
[
{
"name": "enhancement",
"color": "a2eeef"
}
] | false
|
[] |
1,617,400,192
| 5,624
|
glue datasets returning -1 for test split
|
### Describe the bug
Downloading any dataset from GLUE has -1 as class labels for test split. Train and validation have regular 0/1 class labels. This is also present in the dataset card online.
### Steps to reproduce the bug
```
dataset = load_dataset("glue", "sst2")
for d in dataset:
# prints out -1
print(d["label"]
```
### Expected behavior
Expected behavior should be 0/1 instead of -1.
### Environment info
- `datasets` version: 2.4.0
- Platform: Linux-5.15.0-46-generic-x86_64-with-glibc2.17
- Python version: 3.8.16
- PyArrow version: 8.0.0
- Pandas version: 1.5.3
|
closed
|
https://github.com/huggingface/datasets/issues/5624
| 2023-03-09T14:47:18
| 2023-03-09T16:49:29
| 2023-03-09T16:49:29
|
{
"login": "lithafnium",
"id": 8939967,
"type": "User"
}
|
[] | false
|
[] |
1,616,712,665
| 5,623
|
Remove set_access_token usage + fail tests if FutureWarning
|
`set_access_token` is deprecated and will be removed in `huggingface_hub>=0.14`.
This PR removes it from the tests (it was not used in `datasets` source code itself). FYI, it was not needed since `set_access_token` was just setting git credentials and `datasets` doesn't seem to use git anywhere.
In the future, use `set_git_credential` if needed. It is a git-credential-agnostic helper, i.e. you can store your git token in `git-credential-cache`, `git-credential-store`, `osxkeychain`, etc. The legacy `set_access_token` could only set in `git-credential-store` no matter the user preference.
(for context, I found out about this while working on https://github.com/huggingface/huggingface_hub/pull/1381)
---
In addition to this, I have added
```
filterwarnings =
error::FutureWarning:huggingface_hub*
```
to the `setup.cfg` config file to fail on future warnings from `huggingface_hub`. In `hfh`'s CI we trigger on FutureWarning from any package but it's less robust (any package update leads can lead to a failure). No obligation to keep it like that (I can remove it if you prefer) but I think it's a good idea in order to track future FutureWarnings.
FYI, in `huggingface_hub` tests we use `-Werror::FutureWarning --log-cli-level=INFO -sv --durations=0`
- FutureWarning are processed as error
- verbose mode / INFO logs (and above) are captured for easier debugging in github report
- track each test duration, just to see where we can improve. We have a quite long CI (~10min) so it helped improve that.
|
closed
|
https://github.com/huggingface/datasets/pull/5623
| 2023-03-09T08:46:01
| 2023-03-09T15:39:00
| 2023-03-09T15:31:59
|
{
"login": "Wauplin",
"id": 11801849,
"type": "User"
}
|
[] | true
|
[] |
1,615,190,942
| 5,622
|
Update README template to better template
| null |
closed
|
https://github.com/huggingface/datasets/pull/5622
| 2023-03-08T12:30:23
| 2023-03-11T05:07:38
| 2023-03-11T05:07:38
|
{
"login": "emiltj",
"id": 54767532,
"type": "User"
}
|
[] | true
|
[] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.