Datasets:

amannagrawall002

/

github-issues

like 0

[ "Hugging Face's datasets library may prioritize remote configurations. Make sure there are no conflicting configurations causing the library to prefer downloading data\r\nMay be try debugging\r\nraw_datasets = load_dataset('json', data_files=data_files)\r\nprint(raw_datasets)\r\n", "It doesn't download them but writes them to the local HF cache. The logging could indeed be better. Does loading the dataset succeed? If it doesn't, can you share the error stack trace?" ]

[ "Thanks for reporting. We are investigating it.", "This issue is caused by latest `pandas` release 2.1.0 (released yesterday Aug 30).\r\n\r\nSee: https://github.com/huggingface/datasets/actions/runs/6035484010/job/16375932085?pr=6198\r\n", "People using previous releases of `datasets` should pin `pandas` in their local environment:\r\n```\r\npython -m pip install 'pandas<2.1.0'\r\n```" ]

[]

[ "realized that need to pass the path at `cache_file_name` like\r\n\r\n```python\r\ntokenized_datasets = raw_datasets[\"train\"].map(\r\n tokenize_function,\r\n batched=True,\r\n num_proc=data_args.preprocessing_num_workers,\r\n remove_columns=[text_column_name],\r\n load_from_cache_file=True,\r\n desc=\"Running tokenizer on dataset line_by_line\",\r\n # cache_file_names= {\"train\": \"cache-1982fea76aa54a13.arrow\"}\r\n cache_file_name=\"/project/huggingface_cache/datasets/..../cache-1982fea76aa54a13.arrow\",\r\n new_fingerprint=\"1982fea76aa54a13\"\r\n )\r\n```", "Thank you so much! I went through a lot of issues before finding similar experiences here. I have to say that the [docs](https://huggingface.co/docs/datasets/v2.11.0/en/package_reference/main_classes#datasets.Dataset.map) of `.map()` is really misleading, probably making people think that just assigning the file name to cache_file_name is enough." ]

[ "The `fingerprint` parameter serves a slightly different purpose - we use it to inject a new fingerprint after transforming a `Dataset` (computed from the previous fingerprint + transform + transform args), e.g., to be able to compute the cache file for a transform. There is no concept of `fingerprint` before a `Dataset` is fully initialized, but we still need to hash the args (e.g., generator func) of the \"dataset creation methods\" (`from_generator`, `from_csv`, etc.) to compute the cache directory (to store the initial version and transformed dataset versions)\r\n\r\nI agree it should be easier to bypass the hashing mechanism in this instance, too. However, we should probably first address https://github.com/huggingface/datasets/issues/5080 before solving this (e.g., maybe exposing `hash` in `load_dataset`/`load_dataset_builder`.", "Adding +1 here:\r\n\r\nIf the generator needs to access some external resources or state, then it's not always straightforward to make it pickle-able. So I'd like to be able to override how the default cache key derivation needs to pickle the generator (and of course, I'd accept responsibility for that part of cache consistency).\r\n\r\nAppears to be a recurrent roadbump: #6118 #5963 #5819 #5750 #4983 ", "Silly hack incoming:\r\n\r\n```python\r\nimport uuid\r\n\r\nclass _DatasetGeneratorPickleHack:\r\n def __init__(self, generator, generator_id=None):\r\n self.generator = generator\r\n self.generator_id = (\r\n generator_id if generator_id is not None else str(uuid.uuid4())\r\n )\r\n\r\n def __call__(self, *args, **kwargs):\r\n return self.generator(*kwargs, **kwargs)\r\n\r\n def __reduce__(self):\r\n return (_DatasetGeneratorPickleHack_raise, (self.generator_id,))\r\n\r\n\r\ndef _DatasetGeneratorPickleHack_raise(*args, **kwargs):\r\n raise AssertionError(\"cannot actually unpickle _DatasetGeneratorPickleHack!\")\r\n```\r\n\r\nNow `Dataset.from_generator(_DatasetGeneratorPickleHack(gen))` works even if `gen` is unpicklable, because Dataset just pickles the shim object that avoids actually traversing `gen`. Then, one can work out how to set `generator_id` meaningfully to allow cache reuse.", "I'd like some way to do this too. I find that sometimes the hash doesn't cover enough, and that the dataset is not regenerated even when underlying data has changed, and by supplying a custom fingerprint I could do a better job of controlling when my dataset is regenerated.", "This is what I did and it works: \r\n\r\nhttps://github.com/stevemadere/s3-datasets/blob/e475a566a16d3051656a66f8ff4d3baa4c55a66c/src/tokengenerators/text_ds_2_tokens_generator.py#L200\r\n" ]

[ "Before dynamically loading `.py` scripts with `importlib.import_module`, we also parse their contents to check imports, which is tricky to implement for binary `.pyc` files (requires parsing bytecode), so I don't think this is something we want to support (unless more users request it ofc) as this use case is a bit too specific.\r\n\r\n@lhoestq What's your opinion on this?", "> Before dynamically loading .py scripts with importlib.import_module, we also parse their contents to check imports, which is tricky to implement for binary .pyc files (requires parsing bytecode), so I don't think this is something we want to support (unless more users request it ofc) as this use case is a bit too specific.\r\n\r\nYes indeed. Though you can use a .py that imports a package that contains your .pyc code and that you previously installed", "Hi @lhoestq ,\r\nCould you share some example code related to the approach that you are suggesting? " ]

[ "This is because `download_config.use_auth_token` is deprecated - you should use `download_config.token` instead", "Works! Thanks for the quick fix! <3" ]

[ "I think this error means you filter all examples within an (input) batch by deleting its columns. In that case, to avoid the error, you can set the column value to an empty list (`input_batch[\"col\"] = []`) instead." ]

[ "Hi! You can load this dataset with:\r\n```python\r\ndata_files = {\r\n \"train\": \"/content/PUBHEALTH/train.tsv\",\r\n \"validation\": \"/content/PUBHEALTH/dev.tsv\",\r\n \"test\": \"/content/PUBHEALTH/test.tsv\",\r\n}\r\n\r\ntsv_datasets_reloaded = load_dataset(\"csv\", data_files=data_files, sep=\"\\t\")\r\n```\r\n\r\nTo support your `load_dataset` call, defining aliases for the packaged builders, as suggested in https://github.com/huggingface/datasets/issues/5625, must be implemented. We can consider adding this feature if more people request it.\r\n \r\n(Also answered on the Discord [here](https://discord.com/channels/879548962464493619/1145956791134470224/1146071491260186744))" ]

[ "That'd be a great idea! @mariosasko or @lhoestq, would it be possible to fix the code snippet or do you have another suggested way for doing this?", "Indeed `if __name__ == \"__main__\"` is important in this case.\r\n\r\nNot sure about the imbalanced GPU usage though, but maybe you can try using the `torch.cuda.device` context manager ?\r\n\r\n> also, should I do it like this or use nn.DataParallel?\r\n\r\nIn this case you wouldn't need a multiprocessed map no ? Since nn.DataParallel would take care of parallelism", "Adding this Tweet for reference: https://twitter.com/jxmnop/status/1716834517909119019.", "I think the issue is that we set `CUDA_VISIBLE_DEVICES` after pytorch is imported ?\r\n\r\nWe should use `torch.cuda.set_device(...)` instead", "@lhoestq \r\n> In this case you wouldn't need a multiprocessed map no ?\r\n\r\nYes. But how to load a model to 2 GPU simultaneously without something like accelerate?", "> @lhoestq\r\n> \r\n> > In this case you wouldn't need a multiprocessed map no ?\r\n> \r\n> Yes. But how to load a model to 2 GPU simultaneously without something like accelerate?\r\n\r\nTake a look at this fix #6550 . Basically, you move the model to each GPU inside of the function to be mapped. \r\n\r\n", "In case someone also runs into this issue, I wrote a [blog post](https://forrestbao.github.io/2024/01/30/datasets_map_with_rank_multiple_GPUs.html) with a complete working example by compiling information from several PRs and issues here. Hope it can help. This issue cost me a few hours. I hope my blog post can save you time before the official document gets fixed. ", "Thanks ! I updated the docs in https://github.com/huggingface/datasets/pull/6550", "hey @forrestbao , i was too struggling with the same issue for weeks hence i checked out your blog. great work on the blog. \r\nhowever i wanted to ask you could we scale up the process by reinitializing the same model on the same GPU multiple times for even more speedups ? \r\n\r\ni mean to say given that on a multi GPU setup where GPU vram is above 40GB each, after intializing the translation model which is barely 1-2GB in VRAM size, the rest of VRAM sits idle, how could i keep creating multiple instances of the same model on the same GPU for all GPUs to maxmize flops ? ", "You can use one single instance on your GPU and increase the batch size until you fill the VRAM", "@lhoestq i tried that, but i noticed that after a certain number of batch_size, using a larger batch_size makes the overall process really slow than using a lower batch_size.", "Hi @lhoestq , could you help with my two questions: \r\n1. You mentioned `if __name__ == \"__main__\"`, why is that? I tried with a toy dataset and didn't put this line, my two GPU usage looks balanced. \r\n2. Is there any difference between \r\n`from multiprocess import set_start_method` and `from multiprocessing import set_start_method`? The latter is Python's built-in library. In [the official doc](https://huggingface.co/docs/datasets/en/process), it uses `from multiprocess import set_start_method`, but it gives me error like \r\n```\r\n[jobuser@f6e2419a0a63d45638da-n0-0 ~]$ python test.py\r\nTraceback (most recent call last):\r\n File \"/home/jobuser/test.py\", line 33, in <module>\r\n updated_dataset = dataset.map(\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 593, in wrapper\r\n out: Union[\"Dataset\", \"DatasetDict\"] = func(self, *args, **kwargs)\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 558, in wrapper\r\n out: Union[\"Dataset\", \"DatasetDict\"] = func(self, *args, **kwargs)\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 3189, in map\r\n with Pool(len(kwargs_per_job)) as pool:\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/multiprocess/context.py\", line 119, in Pool\r\n return Pool(processes, initializer, initargs, maxtasksperchild,\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/multiprocess/pool.py\", line 191, in __init__\r\n self._setup_queues()\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/multiprocess/pool.py\", line 343, in _setup_queues\r\n self._inqueue = self._ctx.SimpleQueue()\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/multiprocess/context.py\", line 113, in SimpleQueue\r\n return SimpleQueue(ctx=self.get_context())\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/multiprocess/queues.py\", line 339, in __init__\r\n self._rlock = ctx.Lock()\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/multiprocess/context.py\", line 68, in Lock\r\n return Lock(ctx=self.get_context())\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/multiprocess/synchronize.py\", line 168, in __init__\r\n SemLock.__init__(self, SEMAPHORE, 1, 1, ctx=ctx)\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/multiprocess/synchronize.py\", line 86, in __init__\r\n register(self._semlock.name, \"semaphore\")\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/multiprocess/resource_tracker.py\", line 150, in register\r\n self._send('REGISTER', name, rtype)\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/multiprocess/resource_tracker.py\", line 157, in _send\r\n self.ensure_running()\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/multiprocess/resource_tracker.py\", line 124, in ensure_running\r\n pid = util.spawnv_passfds(exe, args, fds_to_pass)\r\n File \"/home/jobuser/.local/lib/python3.10/site-packages/multiprocess/util.py\", line 452, in spawnv_passfds\r\n return _posixsubprocess.fork_exec(\r\nTypeError: fork_exec() takes exactly 21 arguments (17 given)\r\n```\r\nwhich seems caused by python version. I am using Python 3.10.2. ", "Hi ! \r\n\r\n> You mentioned if __name__ == \"__main__\", why is that? I tried with a toy dataset and didn't put this line, my two GPU usage looks balanced.\r\n\r\nIt's a good practice when doing multiprocessing in python. Depending on the multiprocessing method and your python version, python could re-run the code in your main.py in subprocesses that you don't want to re-run (e.g. recursively spawning processes and failing). Though some multiprocessing methods don't re-run main.py and it appears to be your case ;)\r\n\r\n> Is there any difference between\r\nfrom multiprocess import set_start_method and from multiprocessing import set_start_method? The latter is Python's built-in library. In [the official doc](https://huggingface.co/docs/datasets/en/process), it uses from multiprocess import set_start_method, but it gives me error like\r\n\r\nYes, `datasets` uses `multiprocess` which is a separate library from the built-in `multiprocessing`.\r\n\r\n`multiprocess` is an extended version of `multiprocessing` which allows e.g. to pass `lambda` functions to subprocesses", "Thanks @lhoestq for explanation. Is it okay we use `multiprocessing` for set_start_method given the above-mentioned issue for multiprocess? From my run with toy example, it's fine. Just want to check if you foresee any problems. ", "Not sure whether `multiprocessing.set_start_method` has any effect actually since we use `dill` for multiprocessed `map()`", "I'm running the [code example of multi-GPU processing](https://huggingface.co/docs/datasets/en/process#multiprocessing) on a Linux 8x A100 instance. The entire python code run time is 30 seconds faster if I add one line to set torch number of threads immediately after the `import torch` statement. It loads faster to the eight GPUs (however the map() progress bars take similar amount of time without/with this additional line).\r\n```\r\nimport torch\r\ntorch.set_num_threads(1) # I added this line.\r\n\r\nfrom multiprocess import set_start_method\r\n```\r\nFWIW: my instance has these versions.\r\n```\r\nCUDA 12.2 driver 535.161.08\r\nPython 3.10.12\r\ntorch '2.2.2'\r\nmultiprocess '0.70.16'\r\ntransformers '4.39.2'\r\ndatasets '2.18.0'\r\n```" ]

[ "You can cast the `input_image` column to the `Image` type to fix the issue:\r\n```python\r\nds.cast_column(\"input_image\", datasets.Image())\r\n```" ]

[ "This issue is a duplicate of https://github.com/huggingface/datasets/issues/3297. This is a limitation of `dill`, a package we use for caching (non-`__main__` module objects are serialized by reference). You can find more info about it here: https://github.com/uqfoundation/dill/issues/424.\r\n\r\nIn your case, moving \r\n```\r\ndata = datasets.load_dataset('json', data_files=['/tmp/test.json'], split='train')\r\ndata = data.map(transform)\r\n``` \r\nto `test.py` and setting `transform.__module__ = None` at the end of `dataset.py` should fix the issue.", "I understand this may be a limitation of an upstream tool, but for a user for datasets this is very annoying, as when you have dozens of different datasets with different preprocessing functions you can't really move them all into the same file. It may be worth seeing if there is a way to specialize the dependency (eg. subclass it) and enforce behaviors that makes sense for your product.\r\n\r\nI was able to work around this for now by setting `__module__ = None`. If such workarounds are required for now it may be better to document it somewhere than a single obscure issue from a long time ago.\r\n\r\nAs this is a duplicate issue I'm closing it.\r\n\r\nI have another issue with the cache https://github.com/huggingface/datasets/issues/6179 can you take a look?" ]

[ "Same problem", "This was fixed in https://github.com/huggingface/datasets/pull/6155, which will be included in the next release (or you can install `datasets` from source to use it immediately)." ]

[ "Our minimal Python version requirement is 3.8, so we dropped `importlib_metadata`. \r\n\r\nFeel free to open a PR in the `evaluate` repo to replace the problematic import with\r\n```python\r\nif PY_VERSION < version.parse(\"3.8\"):\r\n import importlib_metadata\r\nelse:\r\n import importlib.metadata as importlib_metadata\r\n```", "Any idea when you guys will release the next version which deals with this problem?\r\nI'm still having the same issue with py 3.10 when I install the lib with pip.\r\nI'm assuming that it has not yet been updated since the merge was 3 days ago.", "Yes, this requires a new `evaluate` release (cc @lvwerra for this). \r\n\r\nIn the meantime, you can get the fixed version by installing `evaluate` from `main`: `pip install git+https://github.com/huggingface/evaluate.git`", "I'll aim for a release this week!" ]

[ "https://github.com/huggingface/datasets/issues/5147 may be a solution, by passing in the tokenizer in a fn_kwargs and ignoring it in the fingerprint calculations", "I have a similar issue. I was using a Jupyter Notebook and every time I call the map function it performs tokenization from scratch again although the cache files of last run still exists. \r\n\r\nI ran with 20 processes and now in the cache folder there are two groups of cached results of tokenized dataset:\r\n\r\n```\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Sat Aug 26 12:56:46 2023 cache-1982fea76aa54a13_00001_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Sat Aug 26 13:02:08 2023 cache-1982fea76aa54a13_00004_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Sat Aug 26 12:56:40 2023 cache-1982fea76aa54a13_00005_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 241 MB Sat Aug 26 12:50:59 2023 cache-1982fea76aa54a13_00006_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Sat Aug 26 12:57:37 2023 cache-1982fea76aa54a13_00007_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Sat Aug 26 12:57:31 2023 cache-1982fea76aa54a13_00008_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Sat Aug 26 12:59:47 2023 cache-1982fea76aa54a13_00010_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 241 MB Sat Aug 26 12:59:44 2023 cache-1982fea76aa54a13_00011_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 241 MB Sat Aug 26 12:55:24 2023 cache-1982fea76aa54a13_00012_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 241 MB Sat Aug 26 12:56:21 2023 cache-1982fea76aa54a13_00013_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Sat Aug 26 12:57:24 2023 cache-1982fea76aa54a13_00014_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Sat Aug 26 13:00:48 2023 cache-1982fea76aa54a13_00015_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Sat Aug 26 12:56:56 2023 cache-1982fea76aa54a13_00017_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Sat Aug 26 12:56:54 2023 cache-1982fea76aa54a13_00018_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Sat Aug 26 12:57:27 2023 cache-1982fea76aa54a13_00019_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:15:40 2023 cache-454431f643cdc5e8_00000_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:16:46 2023 cache-454431f643cdc5e8_00001_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:14:53 2023 cache-454431f643cdc5e8_00002_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:13:10 2023 cache-454431f643cdc5e8_00003_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:13:04 2023 cache-454431f643cdc5e8_00004_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:16:42 2023 cache-454431f643cdc5e8_00005_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 241 MB Wed Aug 23 19:01:29 2023 cache-454431f643cdc5e8_00006_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:16:41 2023 cache-454431f643cdc5e8_00007_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:14:04 2023 cache-454431f643cdc5e8_00008_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:17:41 2023 cache-454431f643cdc5e8_00009_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:17:06 2023 cache-454431f643cdc5e8_00010_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 241 MB Wed Aug 23 19:17:16 2023 cache-454431f643cdc5e8_00011_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 241 MB Wed Aug 23 19:15:13 2023 cache-454431f643cdc5e8_00012_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 241 MB Wed Aug 23 19:16:01 2023 cache-454431f643cdc5e8_00013_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:16:35 2023 cache-454431f643cdc5e8_00014_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:16:20 2023 cache-454431f643cdc5e8_00015_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:14:48 2023 cache-454431f643cdc5e8_00016_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 18:59:32 2023 cache-454431f643cdc5e8_00017_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:17:58 2023 cache-454431f643cdc5e8_00018_of_00020.arrow\r\n.rw-r--r-- fad3ew bii_dsc_community 240 MB Wed Aug 23 19:15:25 2023 cache-454431f643cdc5e8_00019_of_00020.arrow\r\n```\r\n\r\ncan we specify the cache file for map so that it won't redo everything again?", "@Luosuu [map](https://huggingface.co/docs/datasets/v2.14.4/en/package_reference/main_classes#datasets.Dataset.map) has cache_file_name parameter\r\n\r\nIn my case, I do want the cache to detect when the map function changes, so I can't pass a constant cache file name.", "Implementing a proper hashing function for the (fast) tokenizers is currently impossible for the reasons mentioned in the referenced issues. So the only alternative to the `cache_file_name` (or `new_fingerprint`) parameter is a custom serializer (e.g., that deserializes the tokenizer from a local save path) defined using `copyreg` or a class that wraps the tokenizer object and has `__reduce__`(`__setstate__`/`__getstate__`) " ]

[ "This seems to be related to your environment and not the `datasets` code (e.g., this could happen when exposing the Python 3.9 site packages to a lower Python version (interpreter))" ]

[ "Hi! Can you share the error this reproducer throws in your environment? `streaming=True` streams the dataset as it's iterated over without creating a memory-map file.", "The trace of the error. Streaming works but is slower.\r\n```\r\nRoot Cause (first observed failure):\r\n[0]:\r\n time : 2023-08-24_06:06:01\r\n host : compute-126.cm.cluster\r\n rank : 0 (local_rank: 0)\r\n exitcode : 1 (pid: 48442)\r\n error_file: /tmp/torchelastic_4fqzcuuz/none_rx2470jl/attempt_0/0/error.json\r\n traceback : Traceback (most recent call last):\r\n File \"/users/yli7/.conda/envs/pytorch2.0/lib/python3.8/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py\", line 346, in wrapper\r\n return f(*args, **kwargs)\r\n File \"Pretrain.py\", line 214, in main\r\n pair_dataset, c4_dataset = create_dataset('pretrain', config)\r\n File \"/dcs05/qiao/data/william/project/DaVinci/dataset/__init__.py\", line 109, in create_dataset\r\n c4_dataset = load_dataset(\"c4\", \"en\", split=\"train\").to_iterable_dataset(num_shards=1024).map(pre_caption_huggingface)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/load.py\", line 1810, in load_dataset\r\n ds = builder_instance.as_dataset(split=split, verification_mode=verification_mode, in_memory=keep_in_memory)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/builder.py\", line 1145, in as_dataset\r\n datasets = map_nested(\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/utils/py_utils.py\", line 436, in map_nested\r\n return function(data_struct)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/builder.py\", line 1175, in _build_single_dataset\r\n ds = self._as_dataset(\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/builder.py\", line 1246, in _as_dataset\r\n dataset_kwargs = ArrowReader(cache_dir, self.info).read(\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 244, in read\r\n return self.read_files(files=files, original_instructions=instructions, in_memory=in_memory)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 265, in read_files\r\n pa_table = self._read_files(files, in_memory=in_memory)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 200, in _read_files\r\n pa_table: Table = self._get_table_from_filename(f_dict, in_memory=in_memory)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 336, in _get_table_from_filename\r\n table = ArrowReader.read_table(filename, in_memory=in_memory)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 357, in read_table\r\n return table_cls.from_file(filename)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/table.py\", line 1059, in from_file\r\n table = _memory_mapped_arrow_table_from_file(filename)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/table.py\", line 65, in _memory_mapped_arrow_table_from_file\r\n opened_stream = _memory_mapped_record_batch_reader_from_file(filename)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/table.py\", line 50, in _memory_mapped_record_batch_reader_from_file\r\n memory_mapped_stream = pa.memory_map(filename)\r\n File \"pyarrow/io.pxi\", line 1009, in pyarrow.lib.memory_map\r\n File \"pyarrow/io.pxi\", line 956, in pyarrow.lib.MemoryMappedFile._open\r\n File \"pyarrow/error.pxi\", line 144, in pyarrow.lib.pyarrow_internal_check_status\r\n File \"pyarrow/error.pxi\", line 115, in pyarrow.lib.check_status\r\n OSError: Memory mapping file failed: Cannot allocate memory\r\n```", "This issue has previously been reported here: https://github.com/huggingface/datasets/issues/5710. Reporting it in the Arrow repo makes more sense as they have control over memory mapping.\r\n\r\nPS: this is the API to reduce the size of the generated Arrow file:\r\n```python\r\nfrom datasets import load_dataset\r\nbuilder = load_dataset_builder(\"c4\", \"en\")\r\nbuilder.download_and_prepare(max_shard_size=\"5GB\")\r\ndataset = builder.as_dataset()\r\n```\r\n\r\nIf this resolves the issue, we can consider exposing `max_shard_size` in `load_dataset`.", "Thanks for the response. The problem seems not resolved. The memory I allocated to the environment is 64G and the following error still occurs\r\n`Python 3.8.16 (default, Jun 12 2023, 18:09:05) \r\n[GCC 11.2.0] :: Anaconda, Inc. on linux\r\nType \"help\", \"copyright\", \"credits\" or \"license\" for more information.\r\n>>> from datasets import load_dataset_builder\r\n>>> builder = load_dataset_builder(\"c4\", \"en\")\r\n>>> builder.download_and_prepare(max_shard_size=\"5GB\")\r\nFound cached dataset c4 (/users/yli7/.cache/huggingface/datasets/c4/en/0.0.0/df532b158939272d032cc63ef19cd5b83e9b4d00c922b833e4cb18b2e9869b01)\r\n>>> dataset = builder.as_dataset()\r\n 0%| | 0/2 [00:00<?, ?it/s]Traceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/builder.py\", line 1145, in as_dataset\r\n datasets = map_nested(\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/utils/py_utils.py\", line 444, in map_nested\r\n mapped = [\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/utils/py_utils.py\", line 445, in <listcomp>\r\n _single_map_nested((function, obj, types, None, True, None))\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/utils/py_utils.py\", line 347, in _single_map_nested\r\n return function(data_struct)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/builder.py\", line 1175, in _build_single_dataset\r\n ds = self._as_dataset(\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/builder.py\", line 1246, in _as_dataset\r\n dataset_kwargs = ArrowReader(cache_dir, self.info).read(\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 244, in read\r\n return self.read_files(files=files, original_instructions=instructions, in_memory=in_memory)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 265, in read_files\r\n pa_table = self._read_files(files, in_memory=in_memory)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 200, in _read_files\r\n pa_table: Table = self._get_table_from_filename(f_dict, in_memory=in_memory)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 336, in _get_table_from_filename\r\n table = ArrowReader.read_table(filename, in_memory=in_memory)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 357, in read_table\r\n return table_cls.from_file(filename)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/table.py\", line 1059, in from_file\r\n table = _memory_mapped_arrow_table_from_file(filename)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/table.py\", line 65, in _memory_mapped_arrow_table_from_file\r\n opened_stream = _memory_mapped_record_batch_reader_from_file(filename)\r\n File \"/users/yli7/.local/lib/python3.8/site-packages/datasets/table.py\", line 50, in _memory_mapped_record_batch_reader_from_file\r\n memory_mapped_stream = pa.memory_map(filename)\r\n File \"pyarrow/io.pxi\", line 1009, in pyarrow.lib.memory_map\r\n File \"pyarrow/io.pxi\", line 956, in pyarrow.lib.MemoryMappedFile._open\r\n File \"pyarrow/error.pxi\", line 144, in pyarrow.lib.pyarrow_internal_check_status\r\n File \"pyarrow/error.pxi\", line 115, in pyarrow.lib.check_status\r\nOSError: Memory mapping file failed: Cannot allocate memory`", "Have you solved the problem?", "Nope. Streaming works but is slower." ]

[]

[ "Hi! The streaming mode also retries requests - `datasets.config.STREAMING_READ_MAX_RETRIES` (20 sec by default) controls the number of retries and `datasets.config.STREAMING_READ_RETRY_INTERVAL` (5 sec) the sleep time between retries.\r\n\r\n> At step 1800 I got a 504 HTTP status code error from Huggingface hub for my pytorch dataloader\r\n\r\nA minor Hub outage that we experienced yesterday could be the cause.", "I wanted something similar. I have a huge dataset I want to process (laion-2b), but after processing several batches, it sometimes fails with this error: `HTTP 502 Bad Gateway for url`. I had the following code to handle it but this way I believe it restarts processing the data from the first batch? How can I set the attribute values you mention above?\r\n\r\n```\r\niterable_dataset = load_dataset(\"laion/laion2B-multi\", streaming=True, split='train')\r\ndataloader = DataLoader(iterable_dataset, batch_size=131072, collate_fn=custom_collate_fn, num_workers=8)\r\n\r\nMAX_RETRIES = 5\r\nRETRY_WAIT = 10 # wait 10 seconds before retry\r\n\r\n for retry in range(MAX_RETRIES):\r\n try:\r\n for j, batch in enumerate(dataloader):\r\n < process batch>\r\n\r\n except HfHubHTTPError as e:\r\n if \"502\" in str(e) and retry < MAX_RETRIES - 1:\r\n logging.warning(f\"Encountered a 502 error on batch {j}. Waiting for {RETRY_WAIT} seconds before retrying.\")\r\n time.sleep(RETRY_WAIT)\r\n continue\r\n else:\r\n raise", "Hey all! Wondering if there's a way of making Datasets streaming mode somewhat robust to Hub outages? Over the weekend, I got two quite cryptic errors, which I reckon were probably from Hub issues:\r\n\r\n<details>\r\n<summary> Stack Trace 1 </summary>\r\n\r\n```\r\n File \"/home/sanchitgandhi/small-12-4-tpu-timestamped-prob-0.2/run_distillation.py\", line 2119, in <module>\r\n main()\r\n File \"/home/sanchitgandhi/small-12-4-tpu-timestamped-prob-0.2/run_distillation.py\", line 1954, in main\r\n for batch in train_loader:\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/torch/utils/data/dataloader.py\", line 630, in __next__\r\n data = self._next_data()\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/torch/utils/data/dataloader.py\", line 1325, in _next_data\r\n return self._process_data(data)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/torch/utils/data/dataloader.py\", line 1371, in _process_data\r\n data.reraise()\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/torch/_utils.py\", line 694, in reraise\r\n raise exception\r\nConnectionError: Caught ConnectionError in DataLoader worker process 8.\r\nOriginal Traceback (most recent call last):\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/aiohttp/connector.py\", line 1155, in _create_direct_connection\r\n hosts = await asyncio.shield(host_resolved)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/aiohttp/connector.py\", line 874, in _resolve_host\r\n addrs = await self._resolver.resolve(host, port, family=self._family)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/aiohttp/resolver.py\", line 33, in resolve\r\n infos = await self._loop.getaddrinfo(\r\n File \"/usr/lib/python3.10/asyncio/base_events.py\", line 863, in getaddrinfo\r\n return await self.run_in_executor(\r\n File \"/usr/lib/python3.10/concurrent/futures/thread.py\", line 58, in run\r\n result = self.fn(*self.args, **self.kwargs)\r\n File \"/usr/lib/python3.10/socket.py\", line 955, in getaddrinfo\r\n for res in _socket.getaddrinfo(host, port, family, type, proto, flags):\r\nsocket.gaierror: [Errno -3] Temporary failure in name resolution\r\nThe above exception was the direct cause of the following exception:\r\nTraceback (most recent call last):\r\n File \"/home/sanchitgandhi/datasets/src/datasets/download/streaming_download_manager.py\", line 333, in read_with_retries\r\n out = read(*args, **kwargs)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/fsspec/implementations/http.py\", line 612, in read\r\n return super().read(length)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/fsspec/spec.py\", line 1856, in read\r\n out = self.cache._fetch(self.loc, self.loc + length)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/fsspec/caching.py\", line 439, in _fetch\r\n new = self.fetcher(self.end, bend)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/fsspec/asyn.py\", line 118, in wrapper\r\n return sync(self.loop, func, *args, **kwargs)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/fsspec/asyn.py\", line 103, in sync\r\n raise return_result\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/fsspec/asyn.py\", line 56, in _runner\r\n result[0] = await coro\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/fsspec/implementations/http.py\", line 660, in async_fetch_range\r\n r = await self.session.get(\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/aiohttp/client.py\", line 562, in _request\r\n conn = await self._connector.connect(\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/aiohttp/connector.py\", line 540, in connect\r\n proto = await self._create_connection(req, traces, timeout)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/aiohttp/connector.py\", line 901, in _create_connection\r\n _, proto = await self._create_direct_connection(req, traces, timeout)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/aiohttp/connector.py\", line 1169, in _create_direct_connection\r\n raise ClientConnectorError(req.connection_key, exc) from exc\r\naiohttp.client_exceptions.ClientConnectorError: Cannot connect to host huggingface.co:443 ssl:default [Temporary failure in name resolution]\r\nThe above exception was the direct cause of the following exception:\r\nTraceback (most recent call last):\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/torch/utils/data/_utils/worker.py\", line 308, in _worker_loop\r\n data = fetcher.fetch(index)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py\", line 32, in fetch\r\n data.append(next(self.dataset_iter))\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 1358, in __iter__\r\n yield from self._iter_pytorch()\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 1293, in _iter_pytorch\r\n for key, example in ex_iterable:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 982, in __iter__\r\n for x in self.ex_iterable:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 862, in __iter__\r\n yield from self._iter()\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 899, in _iter\r\n for key, example in iterator:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 862, in __iter__\r\n yield from self._iter()\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 899, in _iter\r\n for key, example in iterator:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 678, in __iter__\r\n yield from self._iter()\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 740, in _iter\r\n for key, example in iterator:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 862, in __iter__\r\n yield from self._iter()\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 899, in _iter\r\n for key, example in iterator:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 1114, in __iter__\r\n for key, example in self.ex_iterable:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 429, in __iter__\r\n if not iterators[i].hasnext():\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 106, in hasnext\r\n self._thenext = next(self.it)\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 678, in __iter__\r\n yield from self._iter()\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 740, in _iter\r\n for key, example in iterator:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 1114, in __iter__\r\n for key, example in self.ex_iterable:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 281, in __iter__\r\n for key, pa_table in self.generate_tables_fn(**self.kwargs):\r\n File \"/home/sanchitgandhi/.cache/huggingface/modules/datasets_modules/datasets/distil-whisper--switchboard-data/9472ee64cca0e1a7e11909c7033c2354511fa62805f81a2e07616980c765abfe/switchboard-data.py\", line 247, in _generate_tables\r\n for record_batch in pf.iter_batches():\r\n File \"pyarrow/_parquet.pyx\", line 1327, in iter_batches\r\n File \"/home/sanchitgandhi/datasets/src/datasets/download/streaming_download_manager.py\", line 342, in read_with_retries\r\n raise ConnectionError(\"Server Disconnected\") from disconnect_err\r\nConnectionError: Server Disconnected\r\n```\r\n\r\n</details>\r\n\r\n<details>\r\n<summary> Stack Trace 2 </summary>\r\n\r\n```\r\n File \"/home/sanchitgandhi/small-12-2-tpu-v3-timestamped-prob-0.2-bs-512/run_distillation.py\", line 2119, in <module>\r\n main()\r\n File \"/home/sanchitgandhi/small-12-2-tpu-v3-timestamped-prob-0.2-bs-512/run_distillation.py\", line 1954, in main\r\n for batch in train_loader:\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/torch/utils/data/dataloader.py\", line 630, in __next__\r\n data = self._next_data()\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/torch/utils/data/dataloader.py\", line 1325, in _next_data\r\n return self._process_data(data)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/torch/utils/data/dataloader.py\", line 1371, in _process_data\r\n data.reraise()\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/torch/_utils.py\", line 694, in reraise\r\n raise exception\r\nrequests.exceptions.ConnectionError: Caught ConnectionError in DataLoader worker process 13.\r\nOriginal Traceback (most recent call last):\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/urllib3/connectionpool.py\", line 791, in urlopen\r\n response = self._make_request(\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/urllib3/connectionpool.py\", line 537, in _make_request\r\n response = conn.getresponse()\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/urllib3/connection.py\", line 461, in getresponse\r\n httplib_response = super().getresponse()\r\n File \"/usr/lib/python3.10/http/client.py\", line 1375, in getresponse\r\n response.begin()\r\n File \"/usr/lib/python3.10/http/client.py\", line 318, in begin\r\n version, status, reason = self._read_status()\r\n File \"/usr/lib/python3.10/http/client.py\", line 287, in _read_status\r\n raise RemoteDisconnected(\"Remote end closed connection without\"\r\nhttp.client.RemoteDisconnected: Remote end closed connection without response\r\nDuring handling of the above exception, another exception occurred:\r\nTraceback (most recent call last):\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/requests/adapters.py\", line 486, in send\r\n resp = conn.urlopen(\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/urllib3/connectionpool.py\", line 845, in urlopen\r\n retries = retries.increment(\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/urllib3/util/retry.py\", line 470, in increment\r\n raise reraise(type(error), error, _stacktrace)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/urllib3/util/util.py\", line 38, in reraise\r\n raise value.with_traceback(tb)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/urllib3/connectionpool.py\", line 791, in urlopen\r\n response = self._make_request(\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/urllib3/connectionpool.py\", line 537, in _make_request\r\n response = conn.getresponse()\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/urllib3/connection.py\", line 461, in getresponse\r\n httplib_response = super().getresponse()\r\n File \"/usr/lib/python3.10/http/client.py\", line 1375, in getresponse\r\n response.begin()\r\n File \"/usr/lib/python3.10/http/client.py\", line 318, in begin\r\n version, status, reason = self._read_status()\r\n File \"/usr/lib/python3.10/http/client.py\", line 287, in _read_status\r\n raise RemoteDisconnected(\"Remote end closed connection without\"\r\nurllib3.exceptions.ProtocolError: ('Connection aborted.', RemoteDisconnected('Remote end closed connection without response'))\r\nDuring handling of the above exception, another exception occurred:\r\nTraceback (most recent call last):\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/torch/utils/data/_utils/worker.py\", line 308, in _worker_loop\r\n data = fetcher.fetch(index)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py\", line 32, in fetch\r\n data.append(next(self.dataset_iter))\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 1358, in __iter__\r\n yield from self._iter_pytorch()\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 1293, in _iter_pytorch\r\n for key, example in ex_iterable:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 982, in __iter__\r\n for x in self.ex_iterable:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 862, in __iter__\r\n yield from self._iter()\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 899, in _iter\r\n for key, example in iterator:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 862, in __iter__\r\n yield from self._iter()\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 899, in _iter\r\n for key, example in iterator:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 678, in __iter__\r\n yield from self._iter()\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 740, in _iter\r\n for key, example in iterator:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 862, in __iter__\r\n yield from self._iter()\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 899, in _iter\r\n for key, example in iterator:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 1114, in __iter__\r\n for key, example in self.ex_iterable:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 429, in __iter__\r\n if not iterators[i].hasnext():\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 106, in hasnext\r\n self._thenext = next(self.it)\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 678, in __iter__\r\n yield from self._iter()\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 740, in _iter\r\n for key, example in iterator:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 1114, in __iter__\r\n for key, example in self.ex_iterable:\r\n File \"/home/sanchitgandhi/datasets/src/datasets/iterable_dataset.py\", line 281, in __iter__\r\n for key, pa_table in self.generate_tables_fn(**self.kwargs):\r\n File \"/home/sanchitgandhi/datasets/src/datasets/packaged_modules/parquet/parquet.py\", line 87, in _generate_tables\r\n for batch_idx, record_batch in enumerate(\r\n File \"pyarrow/_parquet.pyx\", line 1327, in iter_batches\r\n File \"/home/sanchitgandhi/datasets/src/datasets/download/streaming_download_manager.py\", line 333, in read_with_retries\r\n out = read(*args, **kwargs)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/fsspec/spec.py\", line 1856, in read\r\n out = self.cache._fetch(self.loc, self.loc + length)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/fsspec/caching.py\", line 189, in _fetch\r\n self.cache = self.fetcher(start, end) # new block replaces old\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/huggingface_hub/hf_file_system.py\", line 410, in _fetch_range\r\n r = http_backoff(\"GET\", url, headers=headers)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/huggingface_hub/utils/_http.py\", line 258, in http_backoff\r\n response = session.request(method=method, url=url, **kwargs)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/requests/sessions.py\", line 589, in request\r\n resp = self.send(prep, **send_kwargs)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/requests/sessions.py\", line 703, in send\r\n r = adapter.send(request, **kwargs)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/huggingface_hub/utils/_http.py\", line 63, in send\r\n return super().send(request, *args, **kwargs)\r\n File \"/home/sanchitgandhi/hf/lib/python3.10/site-packages/requests/adapters.py\", line 501, in send\r\n raise ConnectionError(err, request=request)\r\nrequests.exceptions.ConnectionError: (ProtocolError('Connection aborted.', RemoteDisconnected('Remote end closed connection without response')), '(Request ID: 5fce9fc2-e22f-41c2-91af-529f13f1d611)')\r\n```\r\n\r\n</details>\r\n\r\nHaving streaming mode fail when the Hub goes down makes using it problematic for long training runs where large amounts of data is involved. However, this is the precise situation for which streaming mode is so appealing!\r\n\r\nWondering if there were a 'common' set of Hub errors that we could catch in `iterable_datasets` and prevent from crashing the script?\r\n\r\ncc @lhoestq @mariosasko ", "Errors are already caught and requests are already retried.\r\n\r\nWhat you can do is increase the number of retries before an error is raised.\r\n\r\n```python\r\nimport datasets\r\n\r\ndatasets.config.STREAMING_READ_MAX_RETRIES = 20 # default\r\ndatasets.config.STREAMING_READ_RETRY_INTERVAL = 5 # default\r\n```" ]

[ "Unfortunately, I cannot reproduce this behavior on my machine or Colab - the reproducer returns `['main_data', 'additional_data']` as expected.", "Thank you for looking into this, Mario. Is this on [my repository](https://huggingface.co/datasets/tsor13/test), or on another one that you have reproduced? Would you mind pointing me to it if so?", "Whoa, in colab I received the correct behavior using my dataset. It must have something to do with my local copy of `datasets` (which again just failed).\r\n\r\nI've tried uninstalling/reinstnalling to no avail", "hi @tsor13 , I haven't been able to reproduce your issue on `tsor13/test` dataset locally either. reinstalling doesn't help?" ]

[ "Answered on the forum [here](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Ferror-type-arrowinvalid-details-failed-to-parse-string-254-254-as-a-scalar-of-type-int32%2F51323)." ]

[ "Hi ! Feel free to open a discussion at https://huggingface.co/datasets/togethercomputer/RedPajama-Data-1T/discussions to ask the file to be fixed (or directly open a PR with the fixed file)\r\n\r\n`datasets` expects all the examples to have the same fields", "@lhoestq I think the problem is caused by the fact that hugging face datasets writes a copy of data to the local cache using pyarrow. And the data scheme is inferred from the first few data blocks as can be seen [here](https://github.com/huggingface/datasets/blob/main/src/datasets/arrow_writer.py#L570). Maybe setting `streaming=True` can workaround this problem. Would you agree with my statement? ", "> @lhoestq I think the problem is caused by the fact that hugging face datasets writes a copy of data to the local cache using pyarrow. And the data scheme is inferred from the first few data blocks as can be seen [here](https://github.com/huggingface/datasets/blob/main/src/datasets/arrow_writer.py#L570).\r\n\r\nCorrect. Therefore any example that doesn't follow the inferred schema will make the code fail.\r\n\r\n> Maybe setting streaming=True can workaround this problem. Would you agree with my statement?\r\n\r\nYou'll meet the same problem but later - when streaming and arriving at the problematic example", "@lhoestq I just run below test with streaming=True and is not failing at the problematic example\r\n```python\r\nds = load_dataset('json', data_files='/path_to_local_RedPajamaData/filtered_27f05c041a1c401783f90b9415e40e4b.sampled.jsonl', streaming=True)\r\ncount = 0\r\nfor i in ds['train']:\r\n count += 1\r\n print(count)\r\n```\r\n\r\nand completes the 262241 samples successfully. It does error our when streaming is not used " ]

[]

[ "Thanks for reporting, but we can only fix this issue if you can provide a reproducer that consistently reproduces it.", "@mariosasko Ok. What exactly does it mean to provide a reproducer", "To provide a code that reproduces the issue :)", "@mariosasko I complete the above code, is it enough?", "@mariosasko That's all the code, I'm using locally stored data", "Does this error occur even if you change the cache directory (the `cache_dir` parameter in `load_dataset`)?", "@mariosasko I didn't add any parameters for catch. Nor did any cache configuration change.", "@mariosasko And I changed the data file, but executing load_dataset is always the previous result. I had to change something in images.py to use the new results. Using 'cleanup_cache_files' is invalid! Help me.", "@mariosasko I added a comprehensive error message. Check that _column_requires_decoding is being passed where it shouldn't be. DatasetInfo.__init__() Whether this parameter is required", "I can see the issue now... \r\n\r\nYou can fix it by returning a `DatasetInfo` object in the `_info` method as follows:\r\n```python\r\n def _info(self):\r\n if self.config.name == \"similar_pairs\":\r\n features = datasets.Features(\r\n {\r\n \"image1\": datasets.features.Image(),\r\n \"prompt1\": datasets.Value(\"string\"),\r\n \"image2\": datasets.features.Image(),\r\n \"prompt2\": datasets.Value(\"string\"),\r\n \"similarity\": datasets.Value(\"float32\"),\r\n }\r\n )\r\n elif self.config.name == \"image_prompt_pairs\":\r\n features = datasets.Features(\r\n {\"image\": datasets.features.Image(), \"prompt\": datasets.Value(\"string\")}\r\n )\r\n return datasets.DatasetInfo(features=features)\r\n```", "@mariosasko Oh, that's the problem. Thank you very much. Returned the wrong object and it actually works? I've been training with it for a long time", "@mariosasko The original code can still see progress. emmm, I can't see how many examples is generated so far, so I don't know if we should wait", "The original issue has been addressed, so I'm closing it. \r\n\r\nPlease open a new issue if you encounter more errors." ]

[ "@lhoestq ", "`_effective_generator` returns a RNG that takes into account `self._epoch` and the current dataset's base shuffling RNG (which can be set by specifying `seed=` in `.shuffle() for example`).\r\n\r\nTo fix your error you can pass `seed=` to `.shuffle()`. And the shuffling will depend on both this seed and `self._epoch`", "Thanks for the reply" ]

[ "@lhoestq ", "Makes sense, I guess this can be fixed in the load_dataset_builder method.\r\nIt concerns every packaged builder I think (see values in `_PACKAGED_DATASETS_MODULES`)", "I think the behavior is related to these lines, which short circuited the error handling.\r\nhttps://github.com/huggingface/datasets/blob/664a1cb72ea1e6ef7c47e671e2686ca4a35e8d63/src/datasets/load.py#L946-L952\r\n\r\nSo should data_dir be checked here or still delegating to actual `DatasetModule`? In that case, how to properly set `data_files` here.", "This is location in PackagedDatasetModuleFactory.get_module seems the be the right place to check if at least data_dir or data_files are passed", "@mariosasko can you please assign this issue to me,I want to work on this", "#self-assign", "@mariosasko is this issue still open? i would love to kickstart my journey to open source with this issue!\r\nRegards\r\nzutarich", "@zutarich It is unless @debrupf2946 is working on it.", "#self-assign", "I am working and will open a pull request soon @Etelis \r\n", "@mariosasko can i take this up? ", "#self-assign", "Yes, feel free to work on this :)", "i think its working as expected . Heres the log i get for the same line -\r\n\r\n\r\n" ]

[ "`Dataset.sort` essentially does the same thing except it uses `pyarrow.compute.sort_indices` which doesn't involve copying the data into python objects (saving memory)\r\n\r\n```python\r\nsort_keys = [(col, \"ascending\") for col in column_names]\r\nindices = pc.sort_indices(self.data, sort_keys=sort_keys)\r\nreturn self.select(indices)\r\n```", "Ok interesting, I'll continue debugging to see what is going wrong on my end." ]

[ "```\r\n dataset = IterableDataset(dataset) if type(dataset) != IterableDataset else dataset # to force dataset.take(batch_size) to work in non-streaming mode\r\n ```\r\n", "hf discuss: /static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fhow-does-one-make-dataset-take-512-work-with-streaming-false-with-hugging-face-data-set%2F50770", "so: https://stackoverflow.com/questions/76902824/how-does-one-make-dataset-take512-work-with-streaming-false-with-hugging-fac", "Feel free to work on this. In addition, `IterableDataset` supports `skip`, so we should also add this method to `Dataset`." ]

[ "Looks like this regression was introduced in `datasets==2.13.0` (`2.12.0` could load a subset of columns)\r\n\r\nThis does not appear to be fixed by https://github.com/huggingface/datasets/pull/6045 (bug still exists on `main`)" ]

[ "The cause of the error seems to be that `datasets` adds \"gcs://\" as a schema, while `beam` checks only \"gs://\".\r\n\r\ndatasets: https://github.com/huggingface/datasets/blob/c02a44715c036b5261686669727394b1308a3a4b/src/datasets/builder.py#L822\r\n\r\nbeam: [link](https://github.com/apache/beam/blob/25e1a64641b1c8a3c0a6c75c6e86031b87307f22/sdks/python/apache_beam/io/filesystems.py#L98-L101)\r\n```\r\n systems = [\r\n fs for fs in FileSystem.get_all_subclasses()\r\n if fs.scheme() == path_scheme\r\n ]\r\n```", "Hi! We've deprecated the Beam API, as we don't have the bandwidth to support it properly..." ]

[ "I've tried clear the .cache file, doesn't work.", "This issue happens on AWS sagemaker", "This issue can happen if there is a directory named \"glue\" relative to the Python script with the `load_dataset` call (similar issue to this one: https://github.com/huggingface/datasets/issues/5228). Is this the case?", "> This issue can happen if there is a directory named \"glue\" relative to the Python script with the `load_dataset` call (similar issue to this one: #5228). Is this the case?\r\n\r\nThats correct!\r\nSorry for my late response." ]

[ "This is an issue for the [Datasets repo](https://github.com/huggingface/datasets).", "> This is an issue for the [Datasets repo](https://github.com/huggingface/datasets).\r\n\r\nThanks @sgugger , I guess I will wait for them to address the issue . Looking forward to hearing from them ", "You can use `.push_to_hub(\"<username>/<repo>\")` to push a `Dataset` to the Hub.", "> You can use `.push_to_hub(\"<username>/<repo>\")` to push a `Dataset` to the Hub.\r\n\r\nhow about subset? like `.load_dataset(\"<username>/<repo>\", \"<subset>\")`, how can I upload multi-dataset in one repo? thanks a lot ! ", "> > You can use `.push_to_hub(\"<username>/<repo>\")` to push a `Dataset` to the Hub.\r\n> \r\n> how about subset? like `.load_dataset(\"<username>/<repo>\", \"<subset>\")`, how can I upload multi-dataset in one repo? thanks a lot !\r\n\r\nI solved it by upgrading `Datasets` version to 2.15.0" ]

[ "related: https://github.com/huggingface/datasets/issues/3504", "another file not found:\r\n```\r\nTraceback (most recent call last):\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/fsspec/implementations/http.py\", line 417, in _info\r\n await _file_info(\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/fsspec/implementations/http.py\", line 837, in _file_info\r\n r.raise_for_status()\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/aiohttp/client_reqrep.py\", line 1005, in raise_for_status\r\n raise ClientResponseError(\r\naiohttp.client_exceptions.ClientResponseError: 404, message='Not Found', url=URL('https://the-eye.eu/public/AI/pile_preliminary_components/pile_uspto.tar')\r\n\r\nThe above exception was the direct cause of the following exception:\r\n\r\nTraceback (most recent call last):\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/runpy.py\", line 196, in _run_module_as_main\r\n return _run_code(code, main_globals, None,\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/runpy.py\", line 86, in _run_code\r\n exec(code, run_globals)\r\n File \"/lfs/ampere1/0/brando9/.vscode-server-insiders/extensions/ms-python.python-2023.14.0/pythonFiles/lib/python/debugpy/adapter/../../debugpy/launcher/../../debugpy/__main__.py\", line 39, in <module>\r\n cli.main()\r\n File \"/lfs/ampere1/0/brando9/.vscode-server-insiders/extensions/ms-python.python-2023.14.0/pythonFiles/lib/python/debugpy/adapter/../../debugpy/launcher/../../debugpy/../debugpy/server/cli.py\", line 430, in main\r\n run()\r\n File \"/lfs/ampere1/0/brando9/.vscode-server-insiders/extensions/ms-python.python-2023.14.0/pythonFiles/lib/python/debugpy/adapter/../../debugpy/launcher/../../debugpy/../debugpy/server/cli.py\", line 284, in run_file\r\n runpy.run_path(target, run_name=\"__main__\")\r\n File \"/lfs/ampere1/0/brando9/.vscode-server-insiders/extensions/ms-python.python-2023.14.0/pythonFiles/lib/python/debugpy/_vendored/pydevd/_pydevd_bundle/pydevd_runpy.py\", line 321, in run_path\r\n return _run_module_code(code, init_globals, run_name,\r\n File \"/lfs/ampere1/0/brando9/.vscode-server-insiders/extensions/ms-python.python-2023.14.0/pythonFiles/lib/python/debugpy/_vendored/pydevd/_pydevd_bundle/pydevd_runpy.py\", line 135, in _run_module_code\r\n _run_code(code, mod_globals, init_globals,\r\n File \"/lfs/ampere1/0/brando9/.vscode-server-insiders/extensions/ms-python.python-2023.14.0/pythonFiles/lib/python/debugpy/_vendored/pydevd/_pydevd_bundle/pydevd_runpy.py\", line 124, in _run_code\r\n exec(code, run_globals)\r\n File \"/lfs/ampere1/0/brando9/beyond-scale-language-data-diversity/src/diversity/div_coeff.py\", line 526, in <module>\r\n experiment_compute_diveristy_coeff_single_dataset_then_combined_datasets_with_domain_weights()\r\n File \"/lfs/ampere1/0/brando9/beyond-scale-language-data-diversity/src/diversity/div_coeff.py\", line 475, in experiment_compute_diveristy_coeff_single_dataset_then_combined_datasets_with_domain_weights\r\n column_names = next(iter(dataset)).keys()\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/datasets/iterable_dataset.py\", line 1353, in __iter__\r\n for key, example in ex_iterable:\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/datasets/iterable_dataset.py\", line 207, in __iter__\r\n yield from self.generate_examples_fn(**self.kwargs)\r\n File \"/lfs/ampere1/0/brando9/.cache/huggingface/modules/datasets_modules/datasets/EleutherAI--pile/ebea56d358e91cf4d37b0fde361d563bed1472fbd8221a21b38fc8bb4ba554fb/pile.py\", line 257, in _generate_examples\r\n for path, file in files[subset]:\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/datasets/download/streaming_download_manager.py\", line 840, in __iter__\r\n yield from self.generator(*self.args, **self.kwargs)\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/datasets/download/streaming_download_manager.py\", line 891, in _iter_from_urlpath\r\n with xopen(urlpath, \"rb\", download_config=download_config) as f:\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/datasets/download/streaming_download_manager.py\", line 496, in xopen\r\n file_obj = fsspec.open(file, mode=mode, *args, **kwargs).open()\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/fsspec/core.py\", line 134, in open\r\n return self.__enter__()\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/fsspec/core.py\", line 102, in __enter__\r\n f = self.fs.open(self.path, mode=mode)\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/fsspec/spec.py\", line 1241, in open\r\n f = self._open(\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/fsspec/implementations/http.py\", line 356, in _open\r\n size = size or self.info(path, **kwargs)[\"size\"]\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/fsspec/asyn.py\", line 121, in wrapper\r\n return sync(self.loop, func, *args, **kwargs)\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/fsspec/asyn.py\", line 106, in sync\r\n raise return_result\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/fsspec/asyn.py\", line 61, in _runner\r\n result[0] = await coro\r\n File \"/lfs/ampere1/0/brando9/miniconda/envs/beyond_scale/lib/python3.10/site-packages/fsspec/implementations/http.py\", line 430, in _info\r\n raise FileNotFoundError(url) from exc\r\nFileNotFoundError: https://the-eye.eu/public/AI/pile_preliminary_components/pile_uspto.tar\r\n```", "```\r\nFileNotFoundError: https://the-eye.eu/public/AI/pile_preliminary_components/pile_uspto.tar\r\n```\r\nmost relevant line I think.", "link to tweet: https://twitter.com/BrandoHablando/status/1690081313519489024?s=20 about issue", "so: https://stackoverflow.com/questions/76891189/how-to-download-data-from-hugging-face-that-is-visible-on-the-data-viewer-but-th", "this seems to work but it's rather annoying.\r\n\r\nSummary of how to make it work:\r\n1. get urls to parquet files into a list\r\n2. load list to load_dataset via `load_dataset('parquet', data_files=urls)` (note api names to hf are really confusing sometimes)\r\n3. then it should work, print a batch of text.\r\n\r\npresudo code\r\n```python\r\nurls_hacker_news = [\r\n \"https://huggingface.co/datasets/EleutherAI/pile/resolve/refs%2Fconvert%2Fparquet/hacker_news/pile-train-00000-of-00004.parquet\",\r\n \"https://huggingface.co/datasets/EleutherAI/pile/resolve/refs%2Fconvert%2Fparquet/hacker_news/pile-train-00001-of-00004.parquet\",\r\n \"https://huggingface.co/datasets/EleutherAI/pile/resolve/refs%2Fconvert%2Fparquet/hacker_news/pile-train-00002-of-00004.parquet\",\r\n \"https://huggingface.co/datasets/EleutherAI/pile/resolve/refs%2Fconvert%2Fparquet/hacker_news/pile-train-00003-of-00004.parquet\"\r\n]\r\n\r\n...\r\n\r\n\r\n # streaming = False\r\n from diversity.pile_subset_urls import urls_hacker_news\r\n path, name, data_files = 'parquet', 'hacker_news', urls_hacker_news\r\n # not changing\r\n batch_size = 512\r\n today = datetime.datetime.now().strftime('%Y-m%m-d%d-t%Hh_%Mm_%Ss')\r\n run_name = f'{path} div_coeff_{num_batches=} ({today=} ({name=}) {data_mixture_name=} {probabilities=})'\r\n print(f'{run_name=}')\r\n\r\n # - Init wandb\r\n debug: bool = mode == 'dryrun'\r\n run = wandb.init(mode=mode, project=\"beyond-scale\", name=run_name, save_code=True)\r\n wandb.config.update({\"num_batches\": num_batches, \"path\": path, \"name\": name, \"today\": today, 'probabilities': probabilities, 'batch_size': batch_size, 'debug': debug, 'data_mixture_name': data_mixture_name, 'streaming': streaming, 'data_files': data_files})\r\n # run.notify_on_failure() # https://community.wandb.ai/t/how-do-i-set-the-wandb-alert-programatically-for-my-current-run/4891\r\n print(f'{debug=}')\r\n print(f'{wandb.config=}')\r\n\r\n # -- Get probe network\r\n from datasets import load_dataset\r\n import torch\r\n from transformers import GPT2Tokenizer, GPT2LMHeadModel\r\n\r\n tokenizer = GPT2Tokenizer.from_pretrained(\"gpt2\")\r\n if tokenizer.pad_token_id is None:\r\n tokenizer.pad_token = tokenizer.eos_token\r\n probe_network = GPT2LMHeadModel.from_pretrained(\"gpt2\")\r\n device = torch.device(f\"cuda:{0}\" if torch.cuda.is_available() else \"cpu\")\r\n probe_network = probe_network.to(device)\r\n\r\n # -- Get data set\r\n def my_load_dataset(path, name):\r\n print(f'{path=} {name=} {streaming=}')\r\n if path == 'json' or path == 'bin' or path == 'csv':\r\n print(f'{data_files_prefix+name=}')\r\n return load_dataset(path, data_files=data_files_prefix+name, streaming=streaming, split=\"train\").with_format(\"torch\")\r\n elif path == 'parquet':\r\n print(f'{data_files=}')\r\n return load_dataset(path, data_files=data_files, streaming=streaming, split=\"train\").with_format(\"torch\")\r\n else:\r\n return load_dataset(path, name, streaming=streaming, split=\"train\").with_format(\"torch\")\r\n # - get data set for real now\r\n if isinstance(path, str):\r\n dataset = my_load_dataset(path, name)\r\n else:\r\n print('-- interleaving datasets')\r\n datasets = [my_load_dataset(path, name).with_format(\"torch\") for path, name in zip(path, name)]\r\n [print(f'{dataset.description=}') for dataset in datasets]\r\n dataset = interleave_datasets(datasets, probabilities)\r\n print(f'{dataset=}')\r\n batch = dataset.take(batch_size)\r\n print(f'{next(iter(batch))=}')\r\n column_names = next(iter(batch)).keys()\r\n print(f'{column_names=}')\r\n\r\n # - Prepare functions to tokenize batch\r\n def preprocess(examples):\r\n return tokenizer(examples[\"text\"], padding=\"max_length\", max_length=128, truncation=True, return_tensors=\"pt\")\r\n remove_columns = column_names # remove all keys that are not tensors to avoid bugs in collate function in task2vec's pytorch data loader\r\n def map(batch):\r\n return batch.map(preprocess, batched=True, remove_columns=remove_columns)\r\n tokenized_batch = map(batch)\r\n print(f'{next(iter(tokenized_batch))=}')\r\n```\r\n\r\nhttps://stackoverflow.com/questions/76891189/how-to-download-data-from-hugging-face-that-is-visible-on-the-data-viewer-but-th/76902681#76902681\r\n\r\n/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fhow-to-download-data-from-hugging-face-that-is-visible-on-the-data-viewer-but-the-files-are-not-available%2F50555%2F5%3Fu%3Dsevero" ]

[ "@severo ", "It seems that some parquet files have additional columns.\r\n\r\nI ran a scan and found that two files have the additional `__id__` column:\r\n\r\n1. `hf://datasets/bigcode/the-stack-dedup/data/numpy/data-00000-of-00001.parquet`\r\n2. `hf://datasets/bigcode/the-stack-dedup/data/omgrofl/data-00000-of-00001.parquet`\r\n\r\nWe should open a PR to fix those two files", "I opened https://huggingface.co/datasets/bigcode/the-stack-dedup/discussions/31", "The files have been fixed ! I'm closing this one but feel free to re-open if you still have the issue" ]

[ "Hi! I cannot reproduce this error on my machine or in Colab. Which version of `fsspec` do you have installed?" ]

[]

[ "Hi, thanks for the suggestion. It's not possible at the moment. The viewer is part of the Hub codebase and only works on public datasets. Also, it relies on [Datasets Server](https://github.com/huggingface/datasets-server/), which prepares the data and provides an API to access the rows, size, etc.\r\n\r\nIf you're interested in hosting your data as a private dataset on the Hub, you might want to look at https://github.com/huggingface/datasets-server/issues/39.", "Hi, we are building an offline dataset viewer: https://github.com/Renumics/spotlight\r\nIt supports many HF datasets, but currently you have to use it via Pandas:\r\ndf=ds.to_pandas()\r\nspotlight.show(df)\r\n\r\nWould love to hear from you if that works for your use case. If not, feel free to open an issue on the repo: https://github.com/Renumics/spotlight/issues", "@ssuwelack thank you! I will definitely try it out.", "Related issues:\r\n- https://github.com/huggingface/datasets-server/issues/213\r\n- https://github.com/huggingface/datasets-server/issues/441\r\n- https://github.com/huggingface/datasets/issues/6014", "Closing for now, as developing and maintaining an offline viewer is not planned.", "@yuvalkirstain the dataset viewer is now available on private datasets for [PRO users](https://huggingface.co/pricing#pro) and [Enterprise Hub orgs](https://huggingface.co/enterprise). Would it fit your needs?" ]

[]

[]

[ "I noticed that this is actually covered by issue #5613, which for some reason I didn't see when I searched the issues in this repo the first time." ]

[ "@lhoestq ", "It's roughly the same code between the two so we can expected roughly the same speed, could you share a benchmark ?" ]

[ "Fixed with PR" ]

[ "@lhoestq ", "What should be the behavior in this case ? Should it override the default config with the added parameter ?", "I know why it should be treated as a new config if overriding parameters are passed. But in some case, I just pass in some common fields like `data_dir`.\r\n\r\nFor example, I want to extend the FolderBasedBuilder as a multi-config version, the `data_dir` or `data_files` are always passed by user and should not be considered as overriding the default config. In current state, I cannot leverage the feature of default config since passing `data_dir` will disable the default config.", "Thinking more about it I think the current behavior is the right one.\r\n\r\nProvided parameters should be passed to instantiate a new BuilderConfig.\r\n\r\nWhat's the error you're getting ?", "For example, this works to use default config with name '_all_':\r\n```python\r\ndatasets.load_dataset(\"indonesian-nlp/librivox-indonesia\", split=\"train\")\r\n```\r\nwhile this failed to use default config\r\n```python\r\ndatasets.load_dataset(\"indonesian-nlp/librivox-indonesia\", split=\"train\", data_dir='.')\r\n```\r\nAfter manually specifying it, it works again.\r\n```python\r\ndatasets.load_dataset(\"indonesian-nlp/librivox-indonesia\", \"_all_\", split=\"train\", data_dir='.')\r\n```", "@lhoestq ", "It should work if you explicitly ask for the config you want to override\r\n\r\n```python\r\nload_dataset('/dataset/with/multiple/config', 'name_of_the_default_config', some_field_in_config='some')\r\n```\r\n\r\nAlternatively you can have a BuilderConfig class that when instantiated returns a config with the right default values. In this case this code would instantiate this config with the default values except for the parameter to override:\r\n\r\n```python\r\nload_dataset('/dataset/with/multiple/config', some_field_in_config='some')\r\n```", "@lhoestq Yes. But it doesn't work for me.\r\n\r\nHere's my dataset for example.\r\n```\r\nlass MyDatasetConfig(datasets.BuilderConfig):\r\n def __init__(self, name: str, version: str, **kwargs):\r\n self.option1 = kwargs.pop(\"option1\", False)\r\n self.option2 = kwargs.pop(\"option2\", 5)\r\n\r\n super().__init__(\r\n name=name,\r\n version=datasets.Version(version),\r\n **kwargs)\r\n\r\n\r\nclass MyDataset(datasets.GeneratorBasedBuilder):\r\n DEFAULT_CONFIG_NAME = \"v1\"\r\n\r\n BUILDER_CONFIGS = [\r\n UnifiedTtsDatasetConfig(\r\n name=\"v1\",\r\n version=\"1.0.0\",\r\n description=\"Initial version of the dataset\"\r\n ),\r\n ]\r\n\r\n def _info(self) -> DatasetInfo:\r\n _ = self.option1\r\n ....\r\n```\r\n\r\nHere it's okay to use `load_dataset('my_dataset.py')` for loading the default config `v1`.\r\n\r\nBut if I want to override the default values in config with `load_dataset('my_dataset.py', option2=3)`, it failed to find my default config `v1.\r\n\r\nUnless I use `load_dataset('my_dataset.py', 'v1', option2=3)`\r\n\r\nSo according to your advice, how can I modify my dataset to be able to override default config without manually specifying it.", "What's the error ? It should try to instantiate `MyDatasetConfig` with `option2=3`", "@lhoestq The error is\r\n```\r\ndef _info(self) -> DatasetInfo:\r\n _ = self.option1 <-\r\n ....\r\nAttributeError: 'BuilderConfig' object has no attribute 'option1'\r\n```\r\nwhich seems to find another unknown config.\r\n\r\nYou can try this line `datasets.load_dataset(\"indonesian-nlp/librivox-indonesia\", split=\"train\", data_dir='.')`, it's a multi-config dataset on HF hub and the error is the same.\r\n\r\nMy insights:\r\nhttps://github.com/huggingface/datasets/blob/12cfc1196e62847e2e8239fbd727a02cbc86ddec/src/datasets/builder.py#L518\r\nif `config_kwargs` is provided here, the if branch is skipped.", "I see, you just have to set this class attribute to your builder class :)\r\n\r\n```python\r\nBUILDER_CONFIG_CLASS = MyDatasetConfig\r\n```", "So what does this attribute do? In most cases it's not used and the [documents for multi-config dataset](https://huggingface.co/docs/datasets/main/en/image_dataset#multiple-configurations) never mentioned that.", "It tells which builder config class to instantiate if additional config parameters are passed to load_dataset", "@lhoestq maybe we can enhance the document to say something about the common attributes of `DatasetBuilder`", "Ah indeed it's missing in the docs, thanks for reporting. I'm opening a PR" ]

[ "Hi @TomasAndersonFang,\r\n\r\nHave you tried instead to use `torch_compile` in `transformers.TrainingArguments`? https://huggingface.co/docs/transformers/v4.31.0/en/main_classes/trainer#transformers.TrainingArguments.torch_compile", "> \r\n\r\nI tried this and got the following error:\r\n\r\n```\r\nTraceback (most recent call last):\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py\", line 324, in _compile\r\n out_code = transform_code_object(code, transform)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/bytecode_transformation.py\", line 445, in transform_code_object\r\n transformations(instructions, code_options)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py\", line 311, in transform\r\n tracer.run()\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py\", line 1726, in run\r\n super().run()\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py\", line 576, in run\r\n and self.step()\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py\", line 540, in step\r\n getattr(self, inst.opname)(inst)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py\", line 1030, in LOAD_ATTR\r\n result = BuiltinVariable(getattr).call_function(\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/variables/builtin.py\", line 566, in call_function\r\n result = handler(tx, *args, **kwargs)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/variables/builtin.py\", line 931, in call_getattr\r\n return obj.var_getattr(tx, name).add_options(options)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/variables/nn_module.py\", line 124, in var_getattr\r\n subobj = inspect.getattr_static(base, name)\r\n File \"/apps/Arch/software/Python/3.10.8-GCCcore-12.2.0/lib/python3.10/inspect.py\", line 1777, in getattr_static\r\n raise AttributeError(attr)\r\nAttributeError: config\r\n\r\nfrom user code:\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/peft/peft_model.py\", line 909, in forward\r\n if self.base_model.config.model_type == \"mpt\":\r\n\r\nSet torch._dynamo.config.verbose=True for more information\r\n\r\n\r\nYou can suppress this exception and fall back to eager by setting:\r\n torch._dynamo.config.suppress_errors = True\r\n\r\n\r\nThe above exception was the direct cause of the following exception:\r\n\r\nTraceback (most recent call last):\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/llm-copt/fine-tune/falcon/falcon_sft.py\", line 228, in <module>\r\n main()\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/llm-copt/fine-tune/falcon/falcon_sft.py\", line 221, in main\r\n trainer.train()\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/transformers/trainer.py\", line 1539, in train\r\n return inner_training_loop(\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/transformers/trainer.py\", line 1809, in _inner_training_loop\r\n tr_loss_step = self.training_step(model, inputs)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/transformers/trainer.py\", line 2654, in training_step\r\n loss = self.compute_loss(model, inputs)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/transformers/trainer.py\", line 2679, in compute_loss\r\n outputs = model(**inputs)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/nn/modules/module.py\", line 1501, in _call_impl\r\n return forward_call(*args, **kwargs)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py\", line 82, in forward\r\n return self.dynamo_ctx(self._orig_mod.forward)(*args, **kwargs)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py\", line 209, in _fn\r\n return fn(*args, **kwargs)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/accelerate/utils/operations.py\", line 581, in forward\r\n return model_forward(*args, **kwargs)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/accelerate/utils/operations.py\", line 569, in __call__\r\n return convert_to_fp32(self.model_forward(*args, **kwargs))\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/amp/autocast_mode.py\", line 14, in decorate_autocast\r\n return func(*args, **kwargs)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py\", line 337, in catch_errors\r\n return callback(frame, cache_size, hooks)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py\", line 404, in _convert_frame\r\n result = inner_convert(frame, cache_size, hooks)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py\", line 104, in _fn\r\n return fn(*args, **kwargs)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py\", line 262, in _convert_frame_assert\r\n return _compile(\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/utils.py\", line 163, in time_wrapper\r\n r = func(*args, **kwargs)\r\n File \"/cephyr/NOBACKUP/groups/snic2021-23-24/LLM4-CodeOpt/env/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py\", line 394, in _compile\r\n raise InternalTorchDynamoError() from e\r\ntorch._dynamo.exc.InternalTorchDynamoError\r\n```", "Hi @TomasAndersonFang,\r\n\r\nI guess in this case it may be an issue with `transformers` (or `PyTorch`). I would recommend you open an issue on their repo.", "@albertvillanova Thanks for your recommendation. I'll do it", "@TomasAndersonFang were you able to find a solution to this issue? I would highly appreciate any help. \r\n\r\nThanks!" ]

[ "Our CI did not catch this issue because with current implementation, stored token in `HfFolder` (which always exists) is used by default.", "I can confirm this and have the same problem (and just went almost crazy because I couldn't figure out the source of this problem because on another computer everything worked well even with `DownloadMode.FORCE_REDOWNLOAD`).", "We are planning to do a patch release today, after the merge of the fix:\r\n- #6127\r\n\r\nIn the meantime, the problem can be circumvented by passing `download_config` instead:\r\n```python\r\nfrom datasets import DownloadConfig, load_dataset\r\n\r\nload_dataset(\"<DATASET-NAME>\", split=\"train\", download_config=DownloadConfig(token=\"<TOKEN>\"))\r\n``` ", "> We are planning to do a patch release today, after the merge of the fix:\r\n> \r\n> * [Fix authentication issues #6127](https://github.com/huggingface/datasets/pull/6127)\r\n> \r\n> \r\n> In the meantime, the problem can be circumvented by passing `download_config` instead:\r\n> \r\n> ```python\r\n> from datasets import DownloadConfig, load_dataset\r\n> \r\n> load_dataset(\"<DATASET-NAME>\", split=\"train\", download_config=DownloadConfig(token=\"<TOKEN>\"))\r\n> ```\r\n\r\nThis did not work for me (there was some other error with the split being an unexpected size 0). Downgrading to 2.13 fixed it...." ]

[]

[ "i once had the same error and I could fix that by pushing a fake or a dummy commit on my hugging face dataset repo", "Hi! We need a reproducer to fix this. Can you provide a link to the dataset (if it's public)?", "> Hi! We need a reproducer to fix this. Can you provide a link to the dataset (if it's public)?\r\n\r\nHi Mario,\r\n\r\nUnfortunately, the dataset in question is currently private until the model is trained and released.\r\n\r\nThis is not happening with one dataset but numerous hosted private datasets.\r\n\r\nI am only loading the dataset and doing nothing else currently. It seems to happen completely sporadically.\r\n\r\nThank you,\r\n\r\nEnrico", "Hi,\r\n\r\nI have the same error in the dataset viewer with my dataset\r\nhttps://huggingface.co/datasets/elsaEU/ELSA10M_track1\r\n\r\nHas anyone solved this issue?\r\n\r\nEdit: After a dummy commit the error changed in ConfigNamesError", "@rs9000 The problem seems to be the (large) number of commits, as explained in https://huggingface.co/docs/hub/repositories-recommendations. This can be fixed by running:\r\n```python\r\nimport huggingface_hub\r\nhuggingface_hub.super_squash_history(repo_id=\"elsaEU/ELSA10M_track1\")\r\n``` \r\n\r\nThe issue stems from `push_to_hub` creating one commit per shard - https://github.com/huggingface/datasets/pull/6269 should fix this issue (will create one commit per 50 uploaded shards by default). The linked PR will be included in the next `datasets` release.\r\n\r\n\r\ncc @lhoestq @severo for visibility", "Thank you @mariosasko it works.", "#6269 has been merged, so I'm closing this issue" ]

[ "Hi! Thanks for the investigation, but we are not the authors of these datasets, so please report this on the Hub instead so that the actual authors can fix it." ]

[ "You can use `huggingface_hub`'s [Card API](https://huggingface.co/docs/huggingface_hub/package_reference/cards) to programmatically push a dataset card to the Hub." ]

[ "In which format is your dataset? We could expose the `pre_buffer` flag for Parquet to use PyArrow's background thread pool to speed up loading. " ]

[ "Hi! `IterableDataset.from_generator` expects a generator function, not the object (to be consistent with `Dataset.from_generator`).\r\n\r\nYou can fix the above snippet as follows:\r\n```python\r\ntrain_dataset = IterableDataset.from_generator(line_generator, fn_kwargs={\"files\": model_training_files})\r\n```", "to anyone reaching this issue, the argument is `gen_kwargs`:\r\n```py\r\ntrain_dataset = IterableDataset.from_generator(line_generator, gen_kwargs={\"files\": model_training_files})\r\n```" ]

[ "Great idea, feel free to open a PR! :)" ]

[ "You can avoid this by using the `revision` parameter in `load_dataset` to always force downloading a specific commit (if not specified it defaults to HEAD, hence the redownload).", "Thanks @mariosasko this works well, looks like I should have read the documentation a bit more carefully. \r\n\r\nIt is still a bit confusing which hash I should provide: passing `revision = c8fd66e85f086e3abb11eeee55b1737a3d1e8487` from https://huggingface.co/datasets/mozilla-foundation/common_voice_8_0/commits/main caused the cached version at `~/.cache/huggingface/datasets/mozilla-foundation___common_voice_8_0/en/8.0.0/b2f8b72f8f30b2e98c41ccf855954d9e35a5fa498c43332df198534ff9797a4a` to be loaded, so I had to know that it was the previous commit unless I've missed something else." ]

[ "Hi! This should be fixed in the latest (patch) release (run `pip install -U datasets` to install it). This behavior was due to a bug in our authentication logic." ]

[ "Thanks for reporting! This is a bug in converting the `ArrayXD` types to YAML. It will be fixed soon." ]

[ "any idea?", "This should work: `load_dataset(\"path/to/downloaded_repo\")`\r\n\r\n`load_from_disk` is intended to be used on directories created with `Dataset.save_to_disk` or `DatasetDict.save_to_disk`", "> This should work: `load_dataset(\"path/to/downloaded_repo\")`\r\n> \r\n> `load_from_disk` is intended to be used on directories created with `Dataset.save_to_disk` or `DatasetDict.save_to_disk`\r\n\r\nThanks for your help. This works." ]

[ "This is expected behavior. You must provide `cache_file_name` when performing `.map` on an in-memory dataset for the result to be cached." ]

[ "Thanks for reporting, @610v4nn1.\r\n\r\nIndeed, the source data files are no longer available. We have contacted the authors of the dataset and they report that Amazon has decided to stop distributing the multilingual reviews dataset.\r\n\r\nWe are adding a notification about this issue to the dataset card.\r\n\r\nSee: https://huggingface.co/datasets/amazon_reviews_multi/discussions/4#64c3898db63057f1fd3ce1a0 ", "The dataset can be accessed from https://www.kaggle.com/datasets/mexwell/amazon-reviews-multi.", "For those willing to transform the csv files from Kaggle into Huggingface datasets for their NLP course (exercise on summarisation), you can use this code on Google Collab:\r\n\r\n`from datasets import load_dataset\r\n\r\nimport pandas as pd\r\nfrom datasets import Dataset, DatasetDict\r\n\r\n# Load your CSV previously downloaded files from Kaggle on Google Collab\r\ntrain_csv_path = \"/content/train.csv\"\r\nvalidation_csv_path = \"/content/validation.csv\"\r\ntest_csv_path = '/content/test.csv'\r\n\r\n# Read CSV files into pandas DataFrames\r\ntrain_df = pd.read_csv(train_csv_path, engine='python')\r\nvalidation_df = pd.read_csv(validation_csv_path, engine='python')\r\ntest_df = pd.read_csv(test_csv_path, engine='python')\r\n\r\n# Filter by language ('es' for Spanish and 'en' for English)\r\nspanish_train_df = train_df[train_df['language'] == 'es']\r\nspanish_validation_df = validation_df[validation_df['language'] == 'es']\r\nspanish_test_df = test_df[test_df['language'] == 'es']\r\n\r\nenglish_train_df = train_df[train_df['language'] == 'en']\r\nenglish_validation_df = validation_df[validation_df['language'] == 'en']\r\nenglish_test_df = test_df[test_df['language'] == 'en']\r\n\r\n# Create Hugging Face datasets\r\nspanish_dataset = DatasetDict({\r\n 'train': Dataset.from_pandas(spanish_train_df),\r\n 'validation': Dataset.from_pandas(spanish_validation_df),\r\n 'test': Dataset.from_pandas(spanish_test_df)\r\n})\r\n\r\nenglish_dataset = DatasetDict({\r\n 'train': Dataset.from_pandas(english_train_df),\r\n 'validation': Dataset.from_pandas(english_validation_df),\r\n 'test': Dataset.from_pandas(english_test_df)\r\n})\r\nenglish_dataset = english_dataset.remove_columns(['Unnamed: 0', '__index_level_0__'])\r\nspanish_dataset = spanish_dataset.remove_columns(['Unnamed: 0', '__index_level_0__'])`" ]

[ "Yesterday I waited for more than 12 hours to make sure it was really **stuck** instead of proceeding too slow.", "I've had similar weird issues with `load_dataset` as well. Not multiple files, but dataset is quite big, about 50G.", "We use a generic multiprocessing code, so there is little we can do about this - unfortunately, turning off multiprocessing seems to be the only solution. Multithreading would make our code easier to maintain and (most likely) avoid issues such as this one, but we cannot use it until the GIL is dropped (no-GIL Python should be released in 2024, so we can start exploring this then)", "The problem seems to be the `Generating train split`. Is it possible to avoid that? I have a dataset saved, just want to load it but somehow running into issues with that again.", "Hey guys, recently I ran into this problem again and I spent one whole day trying to locate the problem. Finally I found the problem seems to be with `pyarrow`'s json parser, and it seems a long-existing problem. Similar issue can be found in #2181. Anyway, my solution is to adjust the `load_dataset`'s parameter `chunksize`. You can inspect the parameter set in `datasets/packaged_modules/json/json.py`, now the actual chunksize should be very small, and you can increase the value. For me, `chunksize=10<<23` could solve the stuck problem. But I also find that too big `chunksize`, like `10 << 30`, would also cause a stuck, which is rather weird. I think I may explore this when I am free. And hope this can help those who also encounter the same problem. ", "Experiencing the same issue with the `kaist-ai/Feedback-Collection` dataset, which is comparatively small i.e. 100k rows.\r\nCode to reproduce\r\n\r\n```\r\nfrom datasets import load_dataset\r\ndataset = load_dataset(\"kaist-ai/Feedback-Collection\")\r\n```\r\n\r\nI have tried setting `num_proc=1` as well as `chunksize=1024, 64` but problem persists. Any pointers?" ]

[ "Hi! We use PyArrow to read JSON files, and PyArrow doesn't allow different value types in the same column. #5776 should address this.\r\n\r\nIn the meantime, you can combine `Dataset.from_generator` with the above code to cast the values to the same type. ", "Thanks for your help!" ]

[ "Possibly related:\r\n- https://github.com/pytorch/pytorch/issues/22462" ]

[ "Thanks for reporting, @bilelomrani1.\r\n\r\nWe are fixing it. ", "We have fixed it. We are planning to do a patch release today." ]

[ "Seems like the problem isn't with the library, but the dataset itself hosted on AWS S3.\r\n\r\nIts [homepage](https://s3.amazonaws.com/amazon-reviews-pds/readme.html) returns an `AccessDenied` XML response, which is the same thing you get if you try to log the `record` that triggers the exception\r\n\r\n```python\r\ntry:\r\n example = self.info.features.encode_example(record) if self.info.features is not None else record\r\nexcept Exception as e:\r\n print(record)\r\n```\r\n\r\n⬇️\r\n\r\n```\r\n{'<?xml version=\"1.0\" encoding=\"UTF-8\"?>': '<Error><Code>AccessDenied</Code><Message>Access Denied</Message><RequestId>N2HFJ82ZV8SZW9BV</RequestId><HostId>Zw2DQ0V2GdRmvH5qWEpumK4uj5+W8YPcilQbN9fLBr3VqQOcKPHOhUZLG3LcM9X5fkOetxp48Os=</HostId></Error>'}\r\n```", "I'm getting same errors when loading this dataset", "I have figured it out. there was an option of **parquet formated files** i downloaded some from there. ", "this dataset is unfortunately no longer public", "Thanks for reporting, @IqraBaluch.\r\n\r\nWe contacted the authors and unfortunately they reported that Amazon has decided to stop distributing this dataset.", "If anyone still needs this dataset, you could find it on kaggle here : https://www.kaggle.com/datasets/cynthiarempel/amazon-us-customer-reviews-dataset", "Thanks @Maryam-Mostafa ", "@albertvillanova don't tell 'em, we have figured it out. XD", "I noticed that some book data is missing, we can only get Books_v1_02 data. \r\nIs there any way we can get the Books_v1_00 and Books_v1_01? \r\nReally appreciate !!!", "@albertvillanova will this dataset be retired given the data are no longer hosted on S3? What is done in cases such as these?" ]

[ "Actually, my bad -- specifying\r\n```python\r\nfoo.set_format('numpy', ['vectors'], output_all_columns=True)\r\n```\r\nfixes it." ]

[ "Thanks for reporting. We've merged a PR with a fix." ]

[ "We could add a \"threads\" parallel backend to `datasets.parallel.parallel_backend` to support downloading with threads but note that `download_and_extract` also decompresses archives, and this is a CPU-intensive task, which is not ideal for (Python) threads (good for IO-intensive tasks).", "> We could add a \"threads\" parallel backend to `datasets.parallel.parallel_backend` to support downloading with threads but note that `download_and_extract` also decompresses archives, and this is a CPU-intensive task, which is not ideal for (Python) threads (good for IO-intensive tasks).\r\n\r\nGreat! Download takes more time than extract, multiple threads can download in parallel, which can speed up a lot." ]

[]

[ "Thanks for reporting! A PR with a fix has just been merged." ]

[ "Hi @mariosasko unless someone's already working on it, I guess I can tackle it!", "Hi! Sure, feel free to tackle this.", "#self-assign", "I'm assuming this should just cover `to_csv`, `to_parquet`, and `to_json`, right? As `to_list` and `to_dict` just return Python objects, `to_pandas` returns a `pandas.DataFrame` and `to_sql` just inserts into a SQL DB, is that right?", "Fixed by #6096. " ]

[]

[ "When the process starts to hang, can you interrupt it with CTRL + C and paste the error stack trace here? ", "Thanks @mariosasko for your prompt response, here's the stack trace:\r\n\r\n```\r\nKeyboardInterrupt Traceback (most recent call last)\r\nCell In[12], line 4\r\n 2 t = time.time()\r\n 3 iter_ = 0\r\n----> 4 for batch in train_dataloader:\r\n 5 #batch_proc = streaming_obj.collect_streaming_data_batch(batch)\r\n 6 iter_ += 1\r\n 8 if iter_ == 1:\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/torch/utils/data/dataloader.py:634, in _BaseDataLoaderIter.__next__(self)\r\n 631 if self._sampler_iter is None:\r\n 632 # TODO(https://github.com/pytorch/pytorch/issues/76750)\r\n 633 self._reset() # type: ignore[call-arg]\r\n--> 634 data = self._next_data()\r\n 635 self._num_yielded += 1\r\n 636 if self._dataset_kind == _DatasetKind.Iterable and \\\r\n 637 self._IterableDataset_len_called is not None and \\\r\n 638 self._num_yielded > self._IterableDataset_len_called:\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/torch/utils/data/dataloader.py:678, in _SingleProcessDataLoaderIter._next_data(self)\r\n 676 def _next_data(self):\r\n 677 index = self._next_index() # may raise StopIteration\r\n--> 678 data = self._dataset_fetcher.fetch(index) # may raise StopIteration\r\n 679 if self._pin_memory:\r\n 680 data = _utils.pin_memory.pin_memory(data, self._pin_memory_device)\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py:32, in _IterableDatasetFetcher.fetch(self, possibly_batched_index)\r\n 30 for _ in possibly_batched_index:\r\n 31 try:\r\n---> 32 data.append(next(self.dataset_iter))\r\n 33 except StopIteration:\r\n 34 self.ended = True\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/datasets/iterable_dataset.py:1353, in IterableDataset.__iter__(self)\r\n 1350 yield formatter.format_row(pa_table)\r\n 1351 return\r\n-> 1353 for key, example in ex_iterable:\r\n 1354 if self.features:\r\n 1355 # `IterableDataset` automatically fills missing columns with None.\r\n 1356 # This is done with `_apply_feature_types_on_example`.\r\n 1357 example = _apply_feature_types_on_example(\r\n 1358 example, self.features, token_per_repo_id=self._token_per_repo_id\r\n 1359 )\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/datasets/iterable_dataset.py:956, in BufferShuffledExamplesIterable.__iter__(self)\r\n 954 # this is the shuffle buffer that we keep in memory\r\n 955 mem_buffer = []\r\n--> 956 for x in self.ex_iterable:\r\n 957 if len(mem_buffer) == buffer_size: # if the buffer is full, pick and example from it\r\n 958 i = next(indices_iterator)\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/datasets/iterable_dataset.py:296, in ShuffledDataSourcesArrowExamplesIterable.__iter__(self)\r\n 294 for key, pa_table in self.generate_tables_fn(**kwargs_with_shuffled_shards):\r\n 295 for pa_subtable in pa_table.to_reader(max_chunksize=config.ARROW_READER_BATCH_SIZE_IN_DATASET_ITER):\r\n--> 296 formatted_batch = formatter.format_batch(pa_subtable)\r\n 297 for example in _batch_to_examples(formatted_batch):\r\n 298 yield key, example\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/datasets/formatting/formatting.py:448, in PythonFormatter.format_batch(self, pa_table)\r\n 446 if self.lazy:\r\n 447 return LazyBatch(pa_table, self)\r\n--> 448 batch = self.python_arrow_extractor().extract_batch(pa_table)\r\n 449 batch = self.python_features_decoder.decode_batch(batch)\r\n 450 return batch\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/datasets/formatting/formatting.py:150, in PythonArrowExtractor.extract_batch(self, pa_table)\r\n 149 def extract_batch(self, pa_table: pa.Table) -> dict:\r\n--> 150 return pa_table.to_pydict()\r\n\r\nKeyboardInterrupt: \r\n```\r\n", "Update: If i let it run, it eventually fails with:\r\n\r\n```\r\nRuntimeError Traceback (most recent call last)\r\nCell In[16], line 4\r\n 2 t = time.time()\r\n 3 iter_ = 0\r\n----> 4 for batch in train_dataloader:\r\n 5 #batch_proc = streaming_obj.collect_streaming_data_batch(batch)\r\n 6 iter_ += 1\r\n 8 if iter_ == 1:\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/torch/utils/data/dataloader.py:634, in _BaseDataLoaderIter.__next__(self)\r\n 631 if self._sampler_iter is None:\r\n 632 # TODO(https://github.com/pytorch/pytorch/issues/76750)\r\n 633 self._reset() # type: ignore[call-arg]\r\n--> 634 data = self._next_data()\r\n 635 self._num_yielded += 1\r\n 636 if self._dataset_kind == _DatasetKind.Iterable and \\\r\n 637 self._IterableDataset_len_called is not None and \\\r\n 638 self._num_yielded > self._IterableDataset_len_called:\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/torch/utils/data/dataloader.py:678, in _SingleProcessDataLoaderIter._next_data(self)\r\n 676 def _next_data(self):\r\n 677 index = self._next_index() # may raise StopIteration\r\n--> 678 data = self._dataset_fetcher.fetch(index) # may raise StopIteration\r\n 679 if self._pin_memory:\r\n 680 data = _utils.pin_memory.pin_memory(data, self._pin_memory_device)\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py:32, in _IterableDatasetFetcher.fetch(self, possibly_batched_index)\r\n 30 for _ in possibly_batched_index:\r\n 31 try:\r\n---> 32 data.append(next(self.dataset_iter))\r\n 33 except StopIteration:\r\n 34 self.ended = True\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/datasets/iterable_dataset.py:1360, in IterableDataset.__iter__(self)\r\n 1354 if self.features:\r\n 1355 # `IterableDataset` automatically fills missing columns with None.\r\n 1356 # This is done with `_apply_feature_types_on_example`.\r\n 1357 example = _apply_feature_types_on_example(\r\n 1358 example, self.features, token_per_repo_id=self._token_per_repo_id\r\n 1359 )\r\n-> 1360 yield format_dict(example) if format_dict else example\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/datasets/formatting/torch_formatter.py:85, in TorchFormatter.recursive_tensorize(self, data_struct)\r\n 84 def recursive_tensorize(self, data_struct: dict):\r\n---> 85 return map_nested(self._recursive_tensorize, data_struct, map_list=False)\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/datasets/utils/py_utils.py:463, in map_nested(function, data_struct, dict_only, map_list, map_tuple, map_numpy, num_proc, parallel_min_length, types, disable_tqdm, desc)\r\n 461 num_proc = 1\r\n 462 if num_proc != -1 and num_proc <= 1 or len(iterable) < parallel_min_length:\r\n--> 463 mapped = [\r\n 464 _single_map_nested((function, obj, types, None, True, None))\r\n 465 for obj in logging.tqdm(iterable, disable=disable_tqdm, desc=desc)\r\n 466 ]\r\n 467 else:\r\n 468 mapped = parallel_map(function, iterable, num_proc, types, disable_tqdm, desc, _single_map_nested)\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/datasets/utils/py_utils.py:464, in <listcomp>(.0)\r\n 461 num_proc = 1\r\n 462 if num_proc != -1 and num_proc <= 1 or len(iterable) < parallel_min_length:\r\n 463 mapped = [\r\n--> 464 _single_map_nested((function, obj, types, None, True, None))\r\n 465 for obj in logging.tqdm(iterable, disable=disable_tqdm, desc=desc)\r\n 466 ]\r\n 467 else:\r\n 468 mapped = parallel_map(function, iterable, num_proc, types, disable_tqdm, desc, _single_map_nested)\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/datasets/utils/py_utils.py:366, in _single_map_nested(args)\r\n 364 # Singleton first to spare some computation\r\n 365 if not isinstance(data_struct, dict) and not isinstance(data_struct, types):\r\n--> 366 return function(data_struct)\r\n 368 # Reduce logging to keep things readable in multiprocessing with tqdm\r\n 369 if rank is not None and logging.get_verbosity() < logging.WARNING:\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/datasets/formatting/torch_formatter.py:82, in TorchFormatter._recursive_tensorize(self, data_struct)\r\n 80 elif isinstance(data_struct, (list, tuple)):\r\n 81 return self._consolidate([self.recursive_tensorize(substruct) for substruct in data_struct])\r\n---> 82 return self._tensorize(data_struct)\r\n\r\nFile ~/anaconda3/envs/pytorch_p310/lib/python3.10/site-packages/datasets/formatting/torch_formatter.py:68, in TorchFormatter._tensorize(self, value)\r\n 66 if isinstance(value, PIL.Image.Image):\r\n 67 value = np.asarray(value)\r\n---> 68 return torch.tensor(value, **{**default_dtype, **self.torch_tensor_kwargs})\r\n\r\nRuntimeError: Could not infer dtype of decimal.Decimal\r\n```", "PyTorch tensors cannot store `Decimal` objects. Casting the column with decimals to `float` should fix the issue.", "I already have cast in collate_fn, in which I perform .astype(float) for each numerical field.\r\nOn the same instance, I installed a conda env with python 3.6, and this works well.\r\n\r\nSample:\r\n\r\n```\r\ndef streaming_data_collate_fn(batch):\r\n df = pd.DataFrame.from_dict(batch)\r\n feat_vals = torch.FloatTensor(np.nan_to_num(np.array(df[feats].astype(float))))\r\n\r\n```", "`collate_fn` is applied after the `torch` formatting step, so I think the only option when working with an `IterableDataset` is to remove the `with_format` call and perform the conversion from Python values to PyTorch tensors in `collate_fn`. The standard `Dataset` supports `with_format(\"numpy\")`, which should make this conversion faster.", "Thanks! \r\nPython 3.10 conda-env: After replacing with_format(\"torch\") with with_format(\"numpy\"), the error went away. However, it was still taking over 2 minutes to load a very small batch of 64 samples with num_workers set to 32. Once I removed with_format call altogether, it is finishing in 11 seconds.\r\n\r\nPython 3.6 based conda-env: When I switch the kernel , neither of the above work, and with_format(\"torch\") is the only thing that works, and executes in 1.6 seconds.\r\n\r\nI feel something else is also amiss here.", "Can you share the `datasets` and `torch` versions installed in these conda envs?\r\n\r\n> Once I removed with_format call altogether, it is finishing in 11 seconds.\r\n\r\nHmm, that's surprising. What are your dataset's `.features`?", "Python 3.6: \r\ndatasets.__version__ 2.4.0\r\ntorch.__version__ 1.10.1+cu102\r\n\r\nPython 3.10:\r\ndatasets.__version__ 2.14.0\r\ntorch.__version__ 2.0.0\r\n\r\nAnonymized features are of the form (subset shown here):\r\n{\r\n'string_feature_i': Value(dtype='string', id=None),\r\n'numerical_feature_i': Value(dtype='decimal128(38, 0)', id=None),\r\n'numerical_feature_series_i': Sequence(feature=Value(dtype='float64', id=None), length=-1, id=None),\r\n}\r\n\r\n\r\nThere is no output from .features in python 3.6 kernel BTW.", "One more thing, in python 3.10 based kernel, interestingly increasing num_workers seem to be increasing the runtime of iterating I was trying out. In python 3.10 kernel execution, I do not even see multiple CPU cores spiking unlike in 3.6.\r\n\r\n512 batch size on 32 workers executes in 2.4 seconds on python 3.6 kernel, while it takes ~118 seconds on 3.10!", "**Update**: It seems the latency part is more of a multiprocessing issue with torch and some host specific issue, and I had to scourge through relevant pytorch issues, when I stumbled across these threads:\r\n1. https://github.com/pytorch/pytorch/issues/102494\r\n2. https://github.com/pytorch/pytorch/issues/102269\r\n3. https://github.com/pytorch/pytorch/issues/99625\r\n\r\nOut of the suggested solutions, the one that worked in my case was:\r\n```\r\nos.environ['KMP_AFFINITY'] = \"disabled\"\r\n```\r\nIt is working for now, though I have no clue why, just I hope it does not get stuck when I do actual model training, will update by tomorrow.\r\n\r\n\r\n", "I'm facing a similar situation in the local VS Code. \r\n\r\nDatasets version 2.14.4\r\nTorch 2.0.1+cu118\r\n\r\nSame code runs without issues in Colab\r\n\r\n```\r\nfrom datasets import load_dataset\r\n\r\ndataset = load_dataset(\"Supermaxman/esa-hubble\", streaming=True)\r\nsample = next(iter(dataset[\"train\"]))\r\n```\r\n\r\nis stuck for minutes. If I interrupt, I get\r\n\r\n```\r\n---------------------------------------------------------------------------\r\nKeyboardInterrupt Traceback (most recent call last)\r\nCell In[5], line 5\r\n 1 from datasets import load_dataset\r\n 3 dataset = load_dataset(\"Supermaxman/esa-hubble\", streaming=True)\r\n----> 5 sample = next(iter(dataset[\"train\"]))\r\n 6 print(sample[\"text\"])\r\n 7 sample[\"image\"]\r\n\r\nFile [~/miniconda3/envs/book/lib/python3.10/site-packages/datasets/iterable_dataset.py:1353](https://file+.vscode-resource.vscode-cdn.net/home/osanseviero/Desktop/workspace/genai/nbs/~/miniconda3/envs/book/lib/python3.10/site-packages/datasets/iterable_dataset.py:1353), in IterableDataset.__iter__(self)\r\n 1350 yield formatter.format_row(pa_table)\r\n 1351 return\r\n-> 1353 for key, example in ex_iterable:\r\n 1354 if self.features:\r\n 1355 # `IterableDataset` automatically fills missing columns with None.\r\n 1356 # This is done with `_apply_feature_types_on_example`.\r\n 1357 example = _apply_feature_types_on_example(\r\n 1358 example, self.features, token_per_repo_id=self._token_per_repo_id\r\n 1359 )\r\n\r\nFile [~/miniconda3/envs/book/lib/python3.10/site-packages/datasets/iterable_dataset.py:255](https://file+.vscode-resource.vscode-cdn.net/home/osanseviero/Desktop/workspace/genai/nbs/~/miniconda3/envs/book/lib/python3.10/site-packages/datasets/iterable_dataset.py:255), in ArrowExamplesIterable.__iter__(self)\r\n 253 def __iter__(self):\r\n 254 formatter = PythonFormatter()\r\n--> 255 for key, pa_table in self.generate_tables_fn(**self.kwargs):\r\n 256 for pa_subtable in pa_table.to_reader(max_chunksize=config.ARROW_READER_BATCH_SIZE_IN_DATASET_ITER):\r\n...\r\n-> 1130 return self._sslobj.read(len, buffer)\r\n 1131 else:\r\n 1132 return self._sslobj.read(len)\r\n```", "@osanseviero I assume the `self._sslobj.read(len, buffer)` line comes from the built-in `ssl` module, so this probably has something to do with your network. Please open a new issue with the full stack trace in case you haven't resolved this yet.", "Thank you reporting this and sharing the solution, I ran into this as well!", "Ran into same issue after upgrading to pytorch-2.0. Disabling KMP_AFFINITY as mentioned above worked for me. Thanks!\r\n" ]

[ "Currently, it's not possible to efficiently resume streaming after an error. Eventually, we plan to support this for Parquet (see https://github.com/huggingface/datasets/issues/5380). ", "Ok thank you for your answer", "I'm closing this as a duplicate of #5380" ]

[ "The `MAX_MAP_BATCH_SIZE = 1_000_000_000` hack is bad as it loads the entire dataset into RAM when performing `.map`. Instead, it's best to use `.iter(batch_size)` to iterate over the data batches and compute `mean` for each column. (`stddev` can be computed in another pass).\r\n\r\nAlso, these arrays are big, so it makes sense to reduce `batch_size`/`writer_batch_size` to avoid RAM issues and slow IO.", "Hi @mariosasko !\r\n\r\nI agree, it's an ugly hack, but it was convenient since the resulting `mean_std` could be cached by the library. For my large dataset (which doesn't fit in RAM), I'm actually using something similar to what you suggested. I got rid of the first mapping in the above scripts and replaced it with an iterator, but the issue with the second mapping still persists.", "Have you tried to reduce `batch_size`/`writer_batch_size` in the 2nd `.map`? Also, can you interrupt the process when it gets stuck and share the error stack trace?", "I think `batch_size/writer_batch_size` is already at its lowest in the 2nd `.map` since `batched=False` implies `batch_size=1` and `len(ds) = 1000 = writer_batch_size`.\r\n\r\nHere is also a bunch of stack traces when I interrupted the process:\r\n\r\n<details>\r\n <summary>stack trace 1</summary>\r\n\r\n```python\r\n(pyg)[d623204@rosetta-bigviz01 stage-laurent-f]$ python src/random_scripts/uses_random_data.py \r\nFound cached dataset random_data (/local_scratch/lfainsin/.cache/huggingface/datasets/random_data/default/0.0.0/444e214e1d0e6298cfd3f2368323ec37073dc1439f618e19395b1f421c69b066)\r\nApplying mean/std: 97%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████ | 967/1000 [00:01<00:00, 534.87 examples/s]Traceback (most recent call last): \r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 179, in __arrow_array__\r\n storage = to_pyarrow_listarray(data, pa_type)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 1466, in to_pyarrow_listarray\r\n return pa.array(data, pa_type.storage_dtype)\r\n File \"pyarrow/array.pxi\", line 320, in pyarrow.lib.array\r\n File \"pyarrow/array.pxi\", line 39, in pyarrow.lib._sequence_to_array\r\n File \"pyarrow/error.pxi\", line 144, in pyarrow.lib.pyarrow_internal_check_status\r\n File \"pyarrow/error.pxi\", line 123, in pyarrow.lib.check_status\r\npyarrow.lib.ArrowTypeError: Could not convert tensor([[-1.0273, -0.8037, -0.6860],\r\n [-0.5034, -1.2685, -0.0558],\r\n [-1.0908, -1.1820, -0.3178],\r\n ...,\r\n [-0.8171, 0.1781, -0.5903],\r\n [ 0.4370, 1.9305, 0.5899],\r\n [-0.1426, 0.9053, -1.7559]]) with type Tensor: was not a sequence or recognized null for conversion to list type\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 3449, in _map_single\r\n writer.write(example)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 490, in write\r\n self.write_examples_on_file()\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 448, in write_examples_on_file\r\n self.write_batch(batch_examples=batch_examples)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 553, in write_batch\r\n arrays.append(pa.array(typed_sequence))\r\n File \"pyarrow/array.pxi\", line 236, in pyarrow.lib.array\r\n File \"pyarrow/array.pxi\", line 110, in pyarrow.lib._handle_arrow_array_protocol\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 223, in __arrow_array__\r\n return pa.array(cast_to_python_objects(data, only_1d_for_numpy=True))\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 446, in cast_to_python_objects\r\n return _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 407, in _cast_to_python_objects\r\n [\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 408, in <listcomp>\r\n _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 319, in _cast_to_python_objects\r\n [\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 320, in <listcomp>\r\n _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 263, in _cast_to_python_objects\r\n def _cast_to_python_objects(obj: Any, only_1d_for_numpy: bool, optimize_list_casting: bool) -> Tuple[Any, bool]:\r\nKeyboardInterrupt\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 179, in __arrow_array__\r\n storage = to_pyarrow_listarray(data, pa_type)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 1466, in to_pyarrow_listarray\r\n return pa.array(data, pa_type.storage_dtype)\r\n File \"pyarrow/array.pxi\", line 320, in pyarrow.lib.array\r\n File \"pyarrow/array.pxi\", line 39, in pyarrow.lib._sequence_to_array\r\n File \"pyarrow/error.pxi\", line 144, in pyarrow.lib.pyarrow_internal_check_status\r\n File \"pyarrow/error.pxi\", line 123, in pyarrow.lib.check_status\r\npyarrow.lib.ArrowTypeError: Could not convert tensor([[-1.0273, -0.8037, -0.6860],\r\n [-0.5034, -1.2685, -0.0558],\r\n [-1.0908, -1.1820, -0.3178],\r\n ...,\r\n [-0.8171, 0.1781, -0.5903],\r\n [ 0.4370, 1.9305, 0.5899],\r\n [-0.1426, 0.9053, -1.7559]]) with type Tensor: was not a sequence or recognized null for conversion to list type\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"/gpfs_new/data/users/lfainsin/stage-laurent-f/src/random_scripts/uses_random_data.py\", line 62, in <module>\r\n ds_normalized = ds.map(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 580, in wrapper\r\n out: Union[\"Dataset\", \"DatasetDict\"] = func(self, *args, **kwargs)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 545, in wrapper\r\n out: Union[\"Dataset\", \"DatasetDict\"] = func(self, *args, **kwargs)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 3087, in map\r\n for rank, done, content in Dataset._map_single(**dataset_kwargs):\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 3492, in _map_single\r\n writer.finalize()\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 584, in finalize\r\n self.write_examples_on_file()\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 448, in write_examples_on_file\r\n self.write_batch(batch_examples=batch_examples)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 553, in write_batch\r\n arrays.append(pa.array(typed_sequence))\r\n File \"pyarrow/array.pxi\", line 236, in pyarrow.lib.array\r\n File \"pyarrow/array.pxi\", line 110, in pyarrow.lib._handle_arrow_array_protocol\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 223, in __arrow_array__\r\n return pa.array(cast_to_python_objects(data, only_1d_for_numpy=True))\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 446, in cast_to_python_objects\r\n return _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 407, in _cast_to_python_objects\r\n [\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 408, in <listcomp>\r\n _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 319, in _cast_to_python_objects\r\n [\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 319, in <listcomp>\r\n [\r\nKeyboardInterrupt\r\n```\r\n\r\n</details>\r\n\r\n<details>\r\n <summary>stack trace 2</summary>\r\n\r\n```python\r\n(pyg)[d623204@rosetta-bigviz01 stage-laurent-f]$ python src/random_scripts/uses_random_data.py \r\nFound cached dataset random_data (/local_scratch/lfainsin/.cache/huggingface/datasets/random_data/default/0.0.0/444e214e1d0e6298cfd3f2368323ec37073dc1439f618e19395b1f421c69b066)\r\nApplying mean/std: 99%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▏ | 988/1000 [00:20<00:00, 526.19 examples/s]Applying mean/std: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▊| 999/1000 [00:21<00:00, 9.66 examples/s]Traceback (most recent call last): \r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 179, in __arrow_array__\r\n storage = to_pyarrow_listarray(data, pa_type)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 1466, in to_pyarrow_listarray\r\n return pa.array(data, pa_type.storage_dtype)\r\n File \"pyarrow/array.pxi\", line 320, in pyarrow.lib.array\r\n File \"pyarrow/array.pxi\", line 39, in pyarrow.lib._sequence_to_array\r\n File \"pyarrow/error.pxi\", line 144, in pyarrow.lib.pyarrow_internal_check_status\r\n File \"pyarrow/error.pxi\", line 123, in pyarrow.lib.check_status\r\npyarrow.lib.ArrowTypeError: Could not convert tensor([[-1.0273, -0.8037, -0.6860],\r\n [-0.5034, -1.2685, -0.0558],\r\n [-1.0908, -1.1820, -0.3178],\r\n ...,\r\n [-0.8171, 0.1781, -0.5903],\r\n [ 0.4370, 1.9305, 0.5899],\r\n [-0.1426, 0.9053, -1.7559]]) with type Tensor: was not a sequence or recognized null for conversion to list type\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 3449, in _map_single\r\n writer.write(example)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 490, in write\r\n self.write_examples_on_file()\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 448, in write_examples_on_file\r\n self.write_batch(batch_examples=batch_examples)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 553, in write_batch\r\n arrays.append(pa.array(typed_sequence))\r\n File \"pyarrow/array.pxi\", line 236, in pyarrow.lib.array\r\n File \"pyarrow/array.pxi\", line 110, in pyarrow.lib._handle_arrow_array_protocol\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 223, in __arrow_array__\r\n return pa.array(cast_to_python_objects(data, only_1d_for_numpy=True))\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 446, in cast_to_python_objects\r\n return _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 407, in _cast_to_python_objects\r\n [\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 408, in <listcomp>\r\n _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 319, in _cast_to_python_objects\r\n [\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 320, in <listcomp>\r\n _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 263, in _cast_to_python_objects\r\n def _cast_to_python_objects(obj: Any, only_1d_for_numpy: bool, optimize_list_casting: bool) -> Tuple[Any, bool]:\r\nKeyboardInterrupt\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 179, in __arrow_array__\r\n storage = to_pyarrow_listarray(data, pa_type)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 1466, in to_pyarrow_listarray\r\n return pa.array(data, pa_type.storage_dtype)\r\n File \"pyarrow/array.pxi\", line 320, in pyarrow.lib.array\r\n File \"pyarrow/array.pxi\", line 39, in pyarrow.lib._sequence_to_array\r\n File \"pyarrow/error.pxi\", line 144, in pyarrow.lib.pyarrow_internal_check_status\r\n File \"pyarrow/error.pxi\", line 123, in pyarrow.lib.check_status\r\npyarrow.lib.ArrowTypeError: Could not convert tensor([[-1.0273, -0.8037, -0.6860],\r\n [-0.5034, -1.2685, -0.0558],\r\n [-1.0908, -1.1820, -0.3178],\r\n ...,\r\n [-0.8171, 0.1781, -0.5903],\r\n [ 0.4370, 1.9305, 0.5899],\r\n [-0.1426, 0.9053, -1.7559]]) with type Tensor: was not a sequence or recognized null for conversion to list type\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"/gpfs_new/data/users/lfainsin/stage-laurent-f/src/random_scripts/uses_random_data.py\", line 62, in <module>\r\n ds_normalized = ds.map(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 580, in wrapper\r\n out: Union[\"Dataset\", \"DatasetDict\"] = func(self, *args, **kwargs)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 545, in wrapper\r\n out: Union[\"Dataset\", \"DatasetDict\"] = func(self, *args, **kwargs)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 3087, in map\r\n for rank, done, content in Dataset._map_single(**dataset_kwargs):\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 3492, in _map_single\r\n writer.finalize()\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 584, in finalize\r\n self.write_examples_on_file()\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 448, in write_examples_on_file\r\n self.write_batch(batch_examples=batch_examples)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 553, in write_batch\r\n arrays.append(pa.array(typed_sequence))\r\n File \"pyarrow/array.pxi\", line 236, in pyarrow.lib.array\r\n File \"pyarrow/array.pxi\", line 110, in pyarrow.lib._handle_arrow_array_protocol\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 223, in __arrow_array__\r\n return pa.array(cast_to_python_objects(data, only_1d_for_numpy=True))\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 446, in cast_to_python_objects\r\n return _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 407, in _cast_to_python_objects\r\n [\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 408, in <listcomp>\r\n _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 319, in _cast_to_python_objects\r\n [\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 320, in <listcomp>\r\n _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 291, in _cast_to_python_objects\r\n if config.JAX_AVAILABLE and \"jax\" in sys.modules:\r\nKeyboardInterrupt\r\n```\r\n\r\n</details>\r\n\r\n<details>\r\n <summary>stack trace 3</summary>\r\n\r\n```python\r\n(pyg)[d623204@rosetta-bigviz01 stage-laurent-f]$ python src/random_scripts/uses_random_data.py \r\nFound cached dataset random_data (/local_scratch/lfainsin/.cache/huggingface/datasets/random_data/default/0.0.0/444e214e1d0e6298cfd3f2368323ec37073dc1439f618e19395b1f421c69b066)\r\nApplying mean/std: 99%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▎ | 989/1000 [00:01<00:00, 504.80 examples/s]Traceback (most recent call last): \r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 179, in __arrow_array__\r\n storage = to_pyarrow_listarray(data, pa_type)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 1466, in to_pyarrow_listarray\r\n return pa.array(data, pa_type.storage_dtype)\r\n File \"pyarrow/array.pxi\", line 320, in pyarrow.lib.array\r\n File \"pyarrow/array.pxi\", line 39, in pyarrow.lib._sequence_to_array\r\n File \"pyarrow/error.pxi\", line 144, in pyarrow.lib.pyarrow_internal_check_status\r\n File \"pyarrow/error.pxi\", line 123, in pyarrow.lib.check_status\r\npyarrow.lib.ArrowTypeError: Could not convert tensor([[-1.0273, -0.8037, -0.6860],\r\n [-0.5034, -1.2685, -0.0558],\r\n [-1.0908, -1.1820, -0.3178],\r\n ...,\r\n [-0.8171, 0.1781, -0.5903],\r\n [ 0.4370, 1.9305, 0.5899],\r\n [-0.1426, 0.9053, -1.7559]]) with type Tensor: was not a sequence or recognized null for conversion to list type\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 3449, in _map_single\r\n writer.write(example)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 490, in write\r\n self.write_examples_on_file()\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 448, in write_examples_on_file\r\n self.write_batch(batch_examples=batch_examples)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 553, in write_batch\r\n arrays.append(pa.array(typed_sequence))\r\n File \"pyarrow/array.pxi\", line 236, in pyarrow.lib.array\r\n File \"pyarrow/array.pxi\", line 110, in pyarrow.lib._handle_arrow_array_protocol\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 223, in __arrow_array__\r\n return pa.array(cast_to_python_objects(data, only_1d_for_numpy=True))\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 446, in cast_to_python_objects\r\n return _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 407, in _cast_to_python_objects\r\n [\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 408, in <listcomp>\r\n _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 319, in _cast_to_python_objects\r\n [\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 320, in <listcomp>\r\n _cast_to_python_objects(\r\nKeyboardInterrupt\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 179, in __arrow_array__\r\n storage = to_pyarrow_listarray(data, pa_type)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 1466, in to_pyarrow_listarray\r\n return pa.array(data, pa_type.storage_dtype)\r\n File \"pyarrow/array.pxi\", line 320, in pyarrow.lib.array\r\n File \"pyarrow/array.pxi\", line 39, in pyarrow.lib._sequence_to_array\r\n File \"pyarrow/error.pxi\", line 144, in pyarrow.lib.pyarrow_internal_check_status\r\n File \"pyarrow/error.pxi\", line 123, in pyarrow.lib.check_status\r\npyarrow.lib.ArrowTypeError: Could not convert tensor([[-1.0273, -0.8037, -0.6860],\r\n [-0.5034, -1.2685, -0.0558],\r\n [-1.0908, -1.1820, -0.3178],\r\n ...,\r\n [-0.8171, 0.1781, -0.5903],\r\n [ 0.4370, 1.9305, 0.5899],\r\n [-0.1426, 0.9053, -1.7559]]) with type Tensor: was not a sequence or recognized null for conversion to list type\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"/gpfs_new/data/users/lfainsin/stage-laurent-f/src/random_scripts/uses_random_data.py\", line 62, in <module>\r\n ds_normalized = ds.map(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 580, in wrapper\r\n out: Union[\"Dataset\", \"DatasetDict\"] = func(self, *args, **kwargs)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 545, in wrapper\r\n out: Union[\"Dataset\", \"DatasetDict\"] = func(self, *args, **kwargs)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 3087, in map\r\n for rank, done, content in Dataset._map_single(**dataset_kwargs):\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_dataset.py\", line 3492, in _map_single\r\n writer.finalize()\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 584, in finalize\r\n self.write_examples_on_file()\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 448, in write_examples_on_file\r\n self.write_batch(batch_examples=batch_examples)\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 553, in write_batch\r\n arrays.append(pa.array(typed_sequence))\r\n File \"pyarrow/array.pxi\", line 236, in pyarrow.lib.array\r\n File \"pyarrow/array.pxi\", line 110, in pyarrow.lib._handle_arrow_array_protocol\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/arrow_writer.py\", line 223, in __arrow_array__\r\n return pa.array(cast_to_python_objects(data, only_1d_for_numpy=True))\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 446, in cast_to_python_objects\r\n return _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 407, in _cast_to_python_objects\r\n [\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 408, in <listcomp>\r\n _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 319, in _cast_to_python_objects\r\n [\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 320, in <listcomp>\r\n _cast_to_python_objects(\r\n File \"/local_scratch/lfainsin/.conda/envs/pyg/lib/python3.10/site-packages/datasets/features/features.py\", line 298, in _cast_to_python_objects\r\n if obj.ndim == 0:\r\nKeyboardInterrupt\r\n```\r\n\r\n</details>\r\n", "Same issue by following code:\r\n\r\n```python\r\nfrom datasets import load_dataset\r\nfrom torchvision.transforms import transforms\r\n\r\npath = \"~/dataset/diffusiondb50k\" # path maybe not necessary\r\ndataset = load_dataset(\"poloclub/diffusiondb\", \"2m_first_1k\", data_dir=path)\r\n\r\ntransform = transforms.Compose([transforms.ToTensor()])\r\ndataset = dataset.map(\r\n lambda x: {\r\n 'image': transform(x['image']),\r\n 'prompt': x['prompt'],\r\n 'width': x['width'],\r\n 'height': x['height'],\r\n }, \r\n # num_proc=4,\r\n)\r\ndataset\r\n```\r\n\r\nAnd the `dataset.map()` stucks at `Map:  99% 986/1000 [00:07<00:00, 145.72 examples/s]`.\r\n\r\nAlso, there is 1 process left in `htop` with 100% CPU usage. And if I add `num_proc=4,`, there will be 4 same processes left.\r\n\r\n### Environment Info\r\n\r\n- `datasets` version: 2.15.0\r\n- Python version: 3.12.2\r\n- Platform: Linux-6.8.0-36-generic-x86_64-with-glibc2.39", "Hi @zmoki688, I've noticed since that it's pretty common for disk writes to lag behind the operations performed by the `map` operator (especially when the data is large and the operations are cheap). Since the progress bar doesn't seem to account for the writes, it speeds up to 99% but wait until all writes are done. At least that's what I think happens when monitoring my disks I/O (with `iotop` and the likes)" ]

[ "This code behaves as expected on my local machine or in Colab. Which version of `soundfile` do you have installed? MP3 requires `soundfile>=0.12.1`.", "I upgraded the `soundfile` and it's working now! \r\nThanks @mariosasko for the help!" ]

[ "Version 2.3.2 is over one year old, so please use the latest release (2.14.0) to get the expected behavior. Version 2.3.2 does not contain some fixes we made to fix resolving hidden files/directories (starting with a dot)." ]

[ "Hi ! Thanks for reporting, I opened a PR to fix this\r\n\r\nWhat filesystem are you using ?", "Hi @lhoestq ! Thank you so much 🙌 \r\n\r\nIt's a bit of a custom setup, but in practice I am using a [pyarrow.fs.S3FileSystem](https://arrow.apache.org/docs/python/generated/pyarrow.fs.S3FileSystem.html) (wrapped in a `fsspec.implementations.arrow.ArrowFSWrapper` [to make it](https://arrow.apache.org/docs/python/filesystems.html#using-arrow-filesystems-with-fsspec) `fsspec` compatible). I also register it as an entrypoint with `fsspec` so that it's the one that gets automatically resolved when looking for filesystems for the `s3` protocol\r\n\r\nIn my case the `storage_option` that seemed not getting piped through was the filesystem's `endpoint_override` that I use in some tests to point at a mock S3 bucket" ]

[ "You can list the dataset's columns with `ds.column_names` before `.map` to check whether the dataset has an `image` column. If it doesn't, then this is a bug. Otherwise, please paste the line with the `.map` call.\r\n\r\n\r\n", "This is the piece of code I am running:\r\n```\r\ndata_transforms = utils.get_data_augmentation(args)\r\nimage_dataset = utils.load_image_dataset(args.dataset)\r\n\r\ndef resize(examples):\r\n examples[\"pixel_values\"] = [image.convert(\"RGB\").resize((300, 300)) for image in examples[\"image\"]]\r\n return examples\r\n\r\ndef preprocess_train(example_batch):\r\n print(f\"Example batch: \\n{example_batch}\")\r\n example_batch[\"pixel_values\"] = [\r\n data_transforms[\"train\"](image.convert(\"RGB\")) for image in example_batch[\"pixel_values\"]\r\n ]\r\n return example_batch\r\n\r\ndef preprocess_val(example_batch):\r\n example_batch[\"pixel_values\"] = [\r\n data_transforms[\"val\"](image.convert(\"RGB\")) for image in example_batch[\"pixel_values\"]\r\n ]\r\n return example_batch\r\n\r\nimage_dataset = image_dataset.map(resize, remove_columns=[\"image\"], batched=True)\r\n\r\nimage_dataset[\"train\"].set_transform(preprocess_train)\r\nimage_dataset[\"validation\"].set_transform(preprocess_val)\r\n```\r\n\r\nWhen I print ds.column_names I get the following\r\n`{'train': ['image', 'label'], 'validation': ['image', 'label'], 'test': ['image', 'label']}`\r\n\r\nThe `print(f\"Example batch: \\n{example_batch}\")` in the `preprocess_train` function outputs only labels without images:\r\n```\r\nExample batch: \r\n{'label': [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3]}\r\n```\r\n\r\nThe weird part of it all is that a sample code runs in a jupyter lab notebook without any bugs, but when I run my scripts from the terminal I get the bug. The same code.", "The `remove_columns=[\"image\"]` argument in the `.map` call removes the `image` column from the output, so drop this argument to preserve it.", "The problem is not with the removal of the image key. The bug is why only the labels are sent to be process, instead of all the featues or dictionary keys.\r\n\r\nP.S. I just dropped the removal argument as you've suggested, but that didn't solve the problem, because only the labels are being sent to be processed", "All the `image_dataset.column_names` after the `map` call should also be present in `preprocess_train `/`preprocess_val` unless (input) `columns` in `set_transform` are specified.\r\n\r\nIf that's not the case, we need a full reproducer (not snippets) with the environment info.", "I have resolved the error after including a collate function as indicated in the Quick Start session of the Datasets docs.:\r\n\r\nHere is what I did:\r\n```\r\ndata_transforms = utils.get_data_augmentation(args)\r\nimage_dataset = utils.load_image_dataset(args.dataset)\r\n\r\ndef preprocess_train(example_batch):\r\n example_batch[\"pixel_values\"] = [\r\n data_transforms[\"train\"](image.convert(\"RGB\")) for image in example_batch[\"image\"]\r\n ]\r\n return example_batch\r\n\r\ndef preprocess_val(example_batch):\r\n example_batch[\"pixel_values\"] = [\r\n data_transforms[\"val\"](image.convert(\"RGB\")) for image in example_batch[\"image\"]\r\n ]\r\n return example_batch\r\n\r\ndef collate_fn(examples):\r\n images = []\r\n labels = []\r\n for example in examples:\r\n images.append((example[\"pixel_values\"]))\r\n labels.append(example[\"label\"])\r\n\r\n pixel_values = torch.stack(images)\r\n labels = torch.tensor(labels)\r\n return {\"pixel_values\": pixel_values, \"label\": labels}\r\n\r\ntrain_dataset = image_dataset[\"train\"].with_transform(preprocess_train)\r\nval_dataset = image_dataset[\"validation\"].with_transform(preprocess_val)\r\n\r\nimage_datasets = {\r\n \"train\": train_dataset,\r\n \"val\": val_dataset\r\n}\r\n\r\nsamplers = {\r\n \"train\": data.RandomSampler(train_dataset),\r\n \"val\": data.SequentialSampler(val_dataset),\r\n}\r\n\r\ndataloaders = {\r\n x: data.DataLoader(\r\n image_datasets[x],\r\n collate_fn=collate_fn,\r\n batch_size=batch_size,\r\n sampler=samplers[x],\r\n num_workers=args.num_workers,\r\n worker_init_fn=utils.set_seed_for_worker,\r\n generator=g,\r\n pin_memory=True,\r\n )\r\n for x in [\"train\", \"val\"]\r\n}\r\n\r\ntrain_loader, val_loader = dataloaders[\"train\"], dataloaders[\"val\"]\r\n```\r\nEverything runs fine without any bug now. " ]

[ "Hi ! I opened https://github.com/huggingface/datasets/pull/6067 to add the missing `_lock`\r\n\r\nWe'll do a patch release soon, but feel free to install `datasets` from source in the meantime", "I have tested the latest main, it does not work.\r\n\r\nI add more logs to reproduce this issue, it looks like a multi threading bug:\r\n\r\n```python\r\n@contextmanager\r\ndef ensure_lock(tqdm_class, lock_name=\"\"):\r\n \"\"\"get (create if necessary) and then restore `tqdm_class`'s lock\"\"\"\r\n import os\r\n import threading\r\n print(os.getpid(), threading.get_ident(), \"ensure_lock\", tqdm_class, lock_name)\r\n old_lock = getattr(tqdm_class, '_lock', None) # don't create a new lock\r\n lock = old_lock or tqdm_class.get_lock() # maybe create a new lock\r\n lock = getattr(lock, lock_name, lock) # maybe subtype\r\n tqdm_class.set_lock(lock)\r\n print(os.getpid(), threading.get_ident(), \"set_lock\")\r\n yield lock\r\n if old_lock is None:\r\n print(os.getpid(), threading.get_ident(), \"del tqdm_class\")\r\n del tqdm_class._lock\r\n else:\r\n tqdm_class.set_lock(old_lock)\r\n```\r\noutput\r\n```\r\n64943 8424758784 ensure_lock <datasets.utils.logging._tqdm_cls object at 0x2aa7fb250> \r\n64943 8424758784 set_lock\r\n64943 8424758784 del tqdm_class\r\n64943 8424758784 ensure_lock <datasets.utils.logging._tqdm_cls object at 0x2aa7fb250> \r\n64943 8424758784 set_lock\r\n64943 8424758784 del tqdm_class\r\n64943 11638370304 ensure_lock <datasets.utils.logging._tqdm_cls object at 0x2aa7fb250> \r\n64943 11638370304 set_lock\r\n64943 11568967680 ensure_lock <datasets.utils.logging._tqdm_cls object at 0x2aa7fb250> \r\n64943 11568967680 set_lock\r\n64943 11638370304 del tqdm_class\r\n64943 11638370304 ensure_lock <datasets.utils.logging._tqdm_cls object at 0x2aa7fb250> \r\n64943 11638370304 set_lock\r\n64943 11638370304 del tqdm_class\r\n64943 11568967680 del tqdm_class\r\n```\r\n\r\nThread `11638370304` del the _lock from tqdm_class first, then thread `11568967680` del _lock failed.", "Maybe it is a bug of tqdm? I think simply use `try ... except AttributeError ...` wraps `del tqdm_class._lock` should work.", "Yes it looks like a bug on their end indeed, do you want to open a PR on tqdm ?\r\n\r\nLet me see if I can find a workaround in the meantime", "I opened https://github.com/huggingface/datasets/pull/6068 if you want to try it out", "> I opened #6068 if you want to try it out\r\n\r\nThis fix works! Thanks.", "Awesome ! closing this then :)\r\nWe'll do a patch release today or tomorrow" ]

[ "Sorry for asking a duplicate question about `num_proc`, I searched the forum and find the solution.\r\n\r\nBut I still can't make the trick with `torch.distributed.barrier()` to only map at the main process work. The [post on forum]( /static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fslow-processing-with-map-when-using-deepspeed-or-fairscale%2F7229%2F7) didn't help.", "If it does the `map` twice then it means the hash of your map function is not some same between your two processes.\r\n\r\nCan you make sure your map functions have the same hash in different processes ?\r\n\r\n```python\r\nfrom datasets.fingerprint import Hasher\r\n\r\nprint(Hasher.hash(lambda x: cut_reorder_keys(x, num_stations_list=args.num_stations_list, is_pad=True, is_train=True)))\r\nprint(Hasher.hash(lambda x: random_shift(x, shift_range=(-160, 0), feature_scale=16)))\r\n```\r\n\r\nYou can also set the fingerprint used to reload the resulting dataset by passing `new_finegrprint=` in `map`, see https://huggingface.co/docs/datasets/v2.13.1/en/about_cache#the-cache. This will force the different processes to use the same fingerprint used to locate the resulting dataset in the cache.", "Thanks for help! I find the fingerprint between processes don't have same hash:\r\n```\r\nRank 0: Gpu 0 cut_reorder_keys fingerprint c7f47f40e9a67657\r\nRank 0: Gpu 0 random_shift fingerprint 240a0ce79831e7d4\r\n\r\nRank 1: Gpu 1 cut_reorder_keys fingerprint 20edd3d9cf284001\r\nRank 1: Gpu 1 random_shift fingerprint 819f7c1c18e7733f\r\n```\r\nBut my functions only process the example one by one and don't need rank or other arguments. After all it can work in the test for dataset and dataloader.\r\nI'll try to set `new_fingerprint` to see if it works and figure out the reason of different hash.", "I finally figure it out. The fingerprint of the function will change if other key-value pairs change in `args` even the `args.num_stations_list` is not changed.\r\n\r\n```python\r\nlambda x: cut_reorder_keys(x, num_stations_list=args.num_stations_list, is_pad=True, is_train=True)\r\n```\r\n\r\nMy `args` contains the key `rank` which refers the rank of its GPU, so the fingerprints change among the GPUs.\r\nI use `partial` in `functools` to generate a partial function that fixs the argument `num_stations_list=args.num_stations_list`, and the fingerprint of this partial function keeps among the GPUs. Finally I can reuse the mapped cache." ]

[ "I would like this also as I have been working with a dataset with hierarchical classes. In fact, I encountered this very issue when trying to define the dataset with a script. I couldn't find a work around and reverted to hard coding the class names in the readme yaml.\r\n\r\n@david-waterworth do you envision also being able to define the hierarchical structure of the classes?", "@danielduckworth yes I did need to do that (but I ended up ditching datasets as it looks like this is a \"wont fix\"). ", "@david-waterworth Hmm, that's a shame. What are you using now? Also, I’m curious to know about the work you’re doing that involves hierarchical classes, if you don’t mind sharing." ]

[ "This can also mean one of the files was not downloaded correctly.\r\n\r\nWe log an erroneous file's name before raising the reader's error, so this is how you can find the problematic file. Then, you should delete it and call `load_dataset` again.\r\n\r\n(I checked all the uploaded files, and they seem to be valid Parquet files, so I don't think this is a bug on their side)\r\n" ]

[ "Hi!\r\n\r\nIt's hard to explain this behavior without more information. Can you profile the slower version with the following code\r\n```python\r\nimport cProfile, pstats\r\nfrom datasets import load_dataset\r\n\r\nwith cProfile.Profile() as profiler:\r\n ds = load_dataset(...)\r\n\r\nstats = pstats.Stats(profiler).sort_stats(\"cumtime\")\r\nstats.print_stats()\r\n```\r\nand share the output?" ]

[]

[ "A duplicate of https://github.com/huggingface/datasets/issues/5929" ]

[ "This would break a lot of existing code, so we can't really do this." ]

[ "Hi! `_select_contiguous` fetches a (zero-copy) slice of the dataset's Arrow table to build a shard, so I don't think this part is the problem. To me, the issue seems to be the step where we embed external image files' bytes (a lot of file reads). You can use `.map` with multiprocessing to perform this step before `push_to_hub` in a faster manner and cache it to disk:\r\n```python\r\nfrom datasets.table import embed_table_storage\r\n# load_dataset(...)\r\nformat = dataset.format\r\ndataset = dataset.with_format(\"arrow\")\r\ndataset = dataset.map(embed_table_storage, batched=True)\r\ndataset = dataset.with_format(**format)\r\n# push_to_hub(...)\r\n```\r\n\r\n(In Datasets 3.0, these external bytes will be written to an Arrow file when generating a dataset to avoid this \"embed\" step)", "Hi, thanks, this solution saves some time.\r\nBut can't we avoid embedding all external image files bytes with each push, skipping the images that have already been pushed into the repo?\r\n\r\nEdit: Ok I missed the part of cache it manually on the disk the first time, this solves the problem. Thank you" ]

[ "The `audiofolder` loader is not available in version `2.3.2`, hence the error. Please run the `pip install -U datasets` command to update the `datasets` installation to make `load_dataset(\"audiofolder\", ...)` work." ]

[ "hii @lhoestq can you assign this issue to me?\r\n", "You can reply \"#self-assign\" to this issue to automatically get assigned to it :)\r\nLet me know if you have any questions or if I can help", "#2289 ", "Actually i am quite new to figure it out how everything goes and done \r\n\r\n> You can reply \"#self-assign\" to this issue to automatically get assigned to it :)\r\n> Let me know if you have any questions or if I can help\r\n\r\nwhen i wrote #self-assign it automatically got converted to some number is it correct or i have done it some wrong way, I am quite new to open source thus wanna try to learn and explore it", "#2289 #self-assign ", "Ah yea github tries to replace the #self-assign with an issue link. I guess you can try to copy-paste instead to see if it works\r\n\r\nAnyway let me assign you manually", "thanks a lot @lhoestq ! though i have a very lil idea of the issue, i am new. as i said before, but gonna try my best shot to do it.\r\ncan you please suggest some tips or anything from your side, how basically we approach it will be really helpfull.\r\nWill try my best!", "The HfFileSystem from the `huggingface_hub` package can already read the HTTP_PROXY and HTTPS_PROXY environment variables. So the remaining thing missing is the `user_agent` that the user may include in a `DownloadConfig` object. The user agent can be used for regular http calls but also calls to the HfFileSystem.\r\n\r\n- for http, the `user_agent` isn't passed from `DownloadConfig` to `get_datasets_user_agent` in `_prepare_single_hop_path_and_storage_options` in `streaming_download_manager.py` so we need to include it\r\n- for HfFileSystem I think it requires a PR in https://github.com/huggingface/huggingface_hub to include it in the `HfFileSystem.__init__`" ]

[ "Hello @exs-avianello, I have reproduced the bug successfully and have understood the problem. But I am confused regarding this part of the statement, \"`pandas.DataFrame.to_csv` is always called with a buf-like `path_or_buf`\".\r\n\r\nCan you please elaborate on it?\r\n\r\nThanks!", "Hi @aryanxk02 ! Sure, what I actually meant is that when passing a path-like `path_or_buf` here\r\n\r\nhttps://github.com/huggingface/datasets/blob/14f6edd9222e577dccb962ed5338b79b73502fa5/src/datasets/arrow_dataset.py#L4708-L4714 \r\n\r\nit gets converted to a file object behind the scenes here\r\n\r\nhttps://github.com/huggingface/datasets/blob/14f6edd9222e577dccb962ed5338b79b73502fa5/src/datasets/io/csv.py#L92-L94\r\n\r\nand the eventual pandas `.to_csv()` calls that write to it always get `path_or_buf=None`, making pandas ignore the `compression` kwarg in the `to_csv_kwargs`\r\n\r\nhttps://github.com/huggingface/datasets/blob/14f6edd9222e577dccb962ed5338b79b73502fa5/src/datasets/io/csv.py#L107-L109", "@exs-avianello When `path_or_buf` is set to None, the `to_csv()` method will return the CSV data as a string instead of saving it to a file. Hence the compression doesn't take place. I think setting `path_or_buf=self.path_or_buf` should work. What you say?" ]

[]

[ "Instead of writing the loading script, you can use the built-in loader to [load JSON files](https://huggingface.co/docs/datasets/loading#json):\r\n```python\r\nfrom datasets import load_dataset\r\nds = load_dataset(\"json\", data_files={\"train\": os.path.join(data_dir[\"train\"]), \"dev\": os.path.join(data_dir[\"dev\"])})\r\n```" ]

[ "These docs are outdated (version 1.2.1 is over two years old). Please refer to [this](https://huggingface.co/docs/datasets/dataset_script) version instead.\r\n\r\nInitially, we hosted datasets in this repo, but now you can find them [on the HF Hub](https://huggingface.co/datasets) (e.g. the [`ag_news`](https://huggingface.co/datasets/ag_news/blob/main/ag_news.py) script)", "Sorry I thought I'd selected the latest version." ]

[ "Even if a dataset is cached, we still make requests to check whether the cache is up-to-date. [This](https://huggingface.co/docs/datasets/v2.13.1/en/loading#offline) section in the docs explains how to avoid them and directly load the cached version.", "Thanks - that works! However it doesn't resolve the original issue (but I am not sure if it is a WSL problem)", "We use `requests` to make HTTP requests (and `aiohttp` in the streaming mode), so I don't think we can provide much help regarding the socket issue (it probably has something to do with WSL). " ]

[]

[ "`HfApi` comes from the `huggingface_hub` package. You can use [this](https://huggingface.co/docs/huggingface_hub/v0.16.3/en/package_reference/utilities#huggingface_hub.configure_http_backend) utility to change the `huggingface_hub`'s `Session` proxies (see the example).\r\n\r\nWe plan to implement https://github.com/huggingface/datasets/issues/5080 and make this behavior more consistent eventually.", "> this\r\n\r\nThanks. I will try `huggingface_hub.configure_http_backend` to change session's config.", "@mariosasko are you saying if I do the following:\r\n\r\n```\r\ndef backend_factory() -> requests.Session:\r\n session = requests.Session()\r\n session.proxies = {\r\n \"https\": \"127.0.0.1:8887\",\r\n \"http\": \"127.0.0.1:8887\",\r\n }\r\n session.verify = \"/etc/ssl/certs/ca-certificates.crt\"\r\n return session\r\n\r\n# Set it as the default session factory\r\nconfigure_http_backend(backend_factory=backend_factory)\r\n```\r\n\r\nwhich works nicely with transformer library:\r\n\r\n```\r\ndef download_gpt_2_model():\r\n tokenizer = GPT2Tokenizer.from_pretrained(\r\n \"gpt2\", force_download=True, resume_download=False\r\n )\r\n text = \"Replace me by any text you'd like.\"\r\n encoded_input = tokenizer(text, return_tensors=\"pt\")\r\n print(encoded_input)\r\n\r\n model = GPT2Model.from_pretrained(\r\n \"gpt2\", force_download=True, resume_download=False\r\n )\r\n output = model(**encoded_input)\r\n```\r\n\r\nshould work for datasets library as well ?\r\n\r\nIn my case if I just do:\r\n\r\n```\r\ndef download_sts12_sts_dataset():\r\n dataset = load_dataset(\r\n \"mteb/sts12-sts\",\r\n download_mode=\"force_redownload\",\r\n verification_mode=\"basic_checks\",\r\n revision=\"main\",\r\n )\r\n\r\n```\r\nI am getting:\r\n`ConnectionError: Couldn't reach https://huggingface.co/datasets/mteb/sts12-sts/resolve/main/dataset_infos.json (ConnectTimeout(MaxRetryError(\"HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /datasets/mteb/sts12-sts/resolve/main/dataset_infos.json (Caused by ConnectTimeoutError(<urllib3.connection.HTTPSConnection object at 0x7f429e87a3a0>, 'Connection to huggingface.co timed out. (connect timeout=100)'))\")))`\r\n\r\nwhich is typical when the proxy server is not defined. Looks like what is set in configure_http_backend(backend_factory=backend_factory) is ignore.\r\n\r\nIf I use env variable instead, it is working \r\n```\r\ndef download_sts12_sts_dataset():\r\n\r\n os.environ[\"https_proxy\"] = \"127.0.0.1:8887\"\r\n os.environ[\"http_proxy\"] = \"127.0.0.1:8887\"\r\n os.environ[\"REQUESTS_CA_BUNDLE\"] = \"/etc/ssl/certs/ca-certificates.crt\"\r\n\r\n dataset = load_dataset(\r\n \"mteb/sts12-sts\",\r\n download_mode=\"force_redownload\",\r\n verification_mode=\"basic_checks\",\r\n revision=\"main\",\r\n )\r\n```\r\n\r\nShould I add something ?\r\n\r\nI am using `huggingface_hub 0.15.1`, `datasets 2.13.0`, `transformers 4.30.2`", "`huggingface_hub.configure_http_backend` works for `transformers` because they only use the `huggingface_hub` lib for downloads. Our download logic is a bit more complex (e.g., we also support downloading non-Hub files), so we are not aligned with them yet. In the meantime, it's best to use the env vars.", "@mariosasko I fully understand that the logic for dataset is different. I see 2 issues with the current implementation of the env variables:\r\n\r\n- having the same https_proxy/http_prox/no_proxy env variables for all tools is not good in some case. For example I have 2 differents proxy server. In 2019 we had discussion with the Tensorflow teams and they recommended to do the following: TFDS_HTTP_PROXY, TFDS_HTTPS_PROXY ...\r\n- with recent version of requests, it is not possible to deactivate TLS interception (verify=false) by using env variable. This is useful to debug things and in some case TLS is not working and you need to ignore verifying the SSL certificate (probably not recommended) \r\n\r\nOne of the best way is to able to pass our requests.Session() directly\r\n```\r\nimport openai\r\nsession = requests.Session()\r\nsession.cert = CERT\r\nsession.verify = False\r\nopenai.requestssession = session\r\n```\r\n\r\nMy 2 cents in this discussion" ]

[ "Yes, this is intended." ]

[ "I've opened a PR with a fix. In the meantime, you can avoid the error by deleting `task_templates` with `dataset.info.task_templates = None` before the `interleave_datasets` call.\r\n` " ]

[ "Thanks for reporting! I've opened a PR with a fix." ]

[ "This scenario currently requires explicitly passing the target features (to avoid the error): \r\n```python\r\nimport datasets\r\n\r\n...\r\n\r\nfeatures = dataset.features\r\nfeatures[\"output\"] = = [{\"test\": datasets.Value(\"int64\")}]\r\ntest2 = dataset.map(lambda row, idx: test_func(row, idx), with_indices=True, num_proc=32, features=features)\r\n```", "I just encountered the same error in the same situation (multiprocessing with variable length outputs).\r\n\r\nThe funny (or dangerous?) thing is, that this error only showed up when testing with a small test dataset (16 examples, ValueError with `num_proc` >1) but the same code works fine for the full dataset (~70k examples).\r\n\r\n@mariosasko Any idea on how to do that with a nested feature with lists of variable lengths containing dicts?\r\n\r\nEDIT: Was able to narrow it down: >200 Examples: no error, <150 Examples: Error. \r\nNow idea what to make of this but pretty obvious that this is a bug....", "This error also occurs while concatenating the datasets." ]

[]