Datasets:

---

Francesco-A

/

github-issues_huggingface-datasets

like 0

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "I think the only place where we need to add it is here https://github.com/huggingface/hub-docs/blob/main/js/src/lib/interfaces/Types.ts\r\n\r\nAnd I think we can remove tasks.json completely from this repo", "Isn't tasks.json used anymore in this repo?", "> I think the only place where we need to add it is here https://github.com/huggingface/hub-docs/blob/main/js/src/lib/interfaces/Types.ts\r\n> \r\n> And I think we can remove tasks.json completely from this repo\r\n\r\nWhat about the warning I mentioned in https://github.com/huggingface/datasets/issues/5255#issuecomment-1339013527? Also, the depth estimation entry is already present in https://github.com/huggingface/hub-docs/blob/main/js/src/lib/interfaces/Types.ts. ", "The update is based on what I received in the output of the export job (c.f. https://github.com/huggingface/datasets/issues/5255#issuecomment-1339107195). \r\n\r\nEdit: Oh, are you referring to the dataset card of NYU Depth V2?", "Yes, my suggestion was for the dataset card: you got the error message because you tried to set `depth-estimation` in `class_ids` instead of `class_categories`.", "> What about the warning I mentioned in https://github.com/huggingface/datasets/issues/5255#issuecomment-1339013527? Also, the depth estimation entry is already present in https://github.com/huggingface/hub-docs/blob/main/js/src/lib/interfaces/Types.ts.\r\n\r\nif you place it in `task_categories` you should be good :)", "yes i would suggest rm'ing tasks.json here for clarity", "Closing it. ", "It's not clear if we can remove it btw, since old versions of `evaluate` rely on it (see https://github.com/huggingface/evaluate/pull/309)\r\n\r\ncc @lvwerra ", "Actually it can be removed without incidence in old versions of evaluate since we kept an hardcoded `known_task_ids` that is marked \"DEPRECATED\"" ]

[ { "color": "e99695", "default": false, "description": "Requesting to add a new dataset", "id": 2067376369, "name": "dataset request", "node_id": "MDU6TGFiZWwyMDY3Mzc2MzY5", "url": "https://api.github.com/repos/huggingface/datasets/labels/dataset%20request" } ]

[]

[]

[ "Update `datasets` or downgrade `huggingface-hub` ;)\r\n\r\nThe `huggingface-hub` lib did a breaking change a few months ago, and you're using an old version of `datasets` that does't support it", "Awesome thanks a lot! Everything works just fine with `datasets==2.9.0` :-) ", "Getting same error with latest versions.\r\n\r\n\r\n```shell\r\n---------------------------------------------------------------------------\r\nTypeError Traceback (most recent call last)\r\nCell In[99], line 1\r\n----> 1 dataset.push_to_hub(\"mirfan899/kids_phoneme_asr\")\r\n\r\nFile /opt/conda/lib/python3.10/site-packages/datasets/arrow_dataset.py:3538, in Dataset.push_to_hub(self, repo_id, split, private, token, branch, shard_size, embed_external_files)\r\n 3493 def push_to_hub(\r\n 3494 self,\r\n 3495 repo_id: str,\r\n (...)\r\n 3501 embed_external_files: bool = True,\r\n 3502 ):\r\n 3503 \"\"\"Pushes the dataset to the hub.\r\n 3504 The dataset is pushed using HTTP requests and does not need to have neither git or git-lfs installed.\r\n 3505 \r\n (...)\r\n 3536 ```\r\n 3537 \"\"\"\r\n-> 3538 repo_id, split, uploaded_size, dataset_nbytes = self._push_parquet_shards_to_hub(\r\n 3539 repo_id=repo_id,\r\n 3540 split=split,\r\n 3541 private=private,\r\n 3542 token=token,\r\n 3543 branch=branch,\r\n 3544 shard_size=shard_size,\r\n 3545 embed_external_files=embed_external_files,\r\n 3546 )\r\n 3547 organization, dataset_name = repo_id.split(\"/\")\r\n 3548 info_to_dump = self.info.copy()\r\n\r\nFile /opt/conda/lib/python3.10/site-packages/datasets/arrow_dataset.py:3474, in Dataset._push_parquet_shards_to_hub(self, repo_id, split, private, token, branch, shard_size, embed_external_files)\r\n 3472 shard.to_parquet(buffer)\r\n 3473 uploaded_size += buffer.tell()\r\n-> 3474 _retry(\r\n 3475 api.upload_file,\r\n 3476 func_kwargs=dict(\r\n 3477 path_or_fileobj=buffer.getvalue(),\r\n 3478 path_in_repo=path_in_repo(index),\r\n 3479 repo_id=repo_id,\r\n 3480 token=token,\r\n 3481 repo_type=\"dataset\",\r\n 3482 revision=branch,\r\n 3483 identical_ok=True,\r\n 3484 ),\r\n 3485 exceptions=HTTPError,\r\n 3486 status_codes=[504],\r\n 3487 base_wait_time=2.0,\r\n 3488 max_retries=5,\r\n 3489 max_wait_time=20.0,\r\n 3490 )\r\n 3491 return repo_id, split, uploaded_size, dataset_nbytes\r\n\r\nFile /opt/conda/lib/python3.10/site-packages/datasets/utils/file_utils.py:330, in _retry(func, func_args, func_kwargs, exceptions, status_codes, max_retries, base_wait_time, max_wait_time)\r\n 328 while True:\r\n 329 try:\r\n--> 330 return func(*func_args, **func_kwargs)\r\n 331 except exceptions as err:\r\n 332 if retry >= max_retries or (status_codes and err.response.status_code not in status_codes):\r\n\r\nFile /opt/conda/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py:120, in validate_hf_hub_args.<locals>._inner_fn(*args, **kwargs)\r\n 117 if check_use_auth_token:\r\n 118 kwargs = smoothly_deprecate_use_auth_token(fn_name=fn.__name__, has_token=has_token, kwargs=kwargs)\r\n--> 120 return fn(*args, **kwargs)\r\n\r\nTypeError: HfApi.upload_file() got an unexpected keyword argument 'identical_ok'\r\n```", "Feel free to update `datasets` and `huggingface-hub`, it should fix it :)", "I went ahead and upgraded both datasets and hub and still getting the same error\r\n", "Which version do you have ? It's been a while since it has been fixed" ]

[]

[]

[]

[ "Subsequent [PR #1426 ](https://github.com/huggingface/datasets/pull/1426) since this PR has uploaded other files along with the MultiReQA dataset.", "closing this one since a new PR has been created" ]

[ { "color": "a2eeef", "default": true, "description": "New feature or request", "id": 1935892871, "name": "enhancement", "node_id": "MDU6TGFiZWwxOTM1ODkyODcx", "url": "https://api.github.com/repos/huggingface/datasets/labels/enhancement" } ]

[]

[]

[ "I'm closing the PR since we're not going to merge it" ]

[ { "color": "a2eeef", "default": true, "description": "New feature or request", "id": 1935892871, "name": "enhancement", "node_id": "MDU6TGFiZWwxOTM1ODkyODcx", "url": "https://api.github.com/repos/huggingface/datasets/labels/enhancement" } ]

[]

[]

[ "@yjernite I have updated the PR to be language modeling task specific. Please review!\r\n", "Yes a line corresponds to a sentence in this data." ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi @fdtomasi, thanks for reporting.\r\n\r\nYou are right: the original host data URL does no longer exist.\r\n\r\nI've contacted the authors of the dataset to ask them if they host this dataset in another URL.", "Hi, @fdtomasi, the URL is fixed.\r\n\r\nThe fix is already in our master branch and it will be accessible in our next release.\r\n\r\nIn the meantime, you can include the fix if you install the `datasets` library from the master branch:\r\n```\r\npip install -U git+ssh://[email protected]/huggingface/datasets.git@master#egg=datasest\r\n```\r\nor\r\n```\r\npip install -U git+https://github.com/huggingface/datasets.git@master#egg=datasets\r\n```", "Awesome thank you so much for the quick fix!" ]

[]

[]

[]

[ "Thank you for working on this, @bhavitvyamalik \r\n\r\n10% is not solving the issue, we want 5-10x faster on a machine that has lots of resources, but limited processing time.\r\n\r\nSo let's benchmark it on an instance with many more cores, I can test with 12 on my dev box and 40 on JZ. \r\n\r\nCould you please share the test I could run with both versions?\r\n\r\nShould we also test the sharded version I shared in https://github.com/huggingface/datasets/issues/2663#issue-946552273 so optionally 3 versions to test.", "Since I was facing `OSError: [Errno 12] Cannot allocate memory` in CircleCI tests, I've added `num_proc` option instead of always using full `cpu_count`. You can test both v1 and v2 through this branch (some redundancy needs to be removed). \r\n\r\nUpdate: I was able to convert into json which took 50% less time as compared to v1 on `ascent_kb` dataset. Will post the benchmarking script with results here.", "Here are the benchmarks with the current branch for both v1 and v2 (dataset: `ascent_kb`, 8.9M samples):\r\n| batch_size | time (in sec) | time (in sec) |\r\n|------------|---------------|---------------|\r\n| | num_proc = 1 | num_proc = 4 |\r\n| 10k | 185.56 | 170.11 |\r\n| 50k | 175.79 | 86.84 |\r\n| **100k** | 191.09 | **78.35** |\r\n| 125k | 198.28 | 90.89 |\r\n\r\nIncreasing the batch size on my machine helped in making v2 around 50% faster as compared to v1. Timings may vary depending on the machine. I'm including the benchmarking script as well. CircleCI errors are unrelated (something related to `bertscore`)\r\n```\r\nimport time\r\nfrom datasets import load_dataset\r\nimport pathlib\r\nimport os\r\nfrom pathlib import Path\r\nimport shutil\r\nimport gc\r\n\r\nbatch_sizes = [10_000, 50_000, 100_000, 125_000]\r\nnum_procs = [1, 4] # change this according to your machine\r\n\r\nSAVE_LOC = \"./new_dataset.json\"\r\n\r\nfor batch in batch_sizes:\r\n for num in num_procs:\r\n dataset = load_dataset(\"ascent_kb\")\r\n\r\n local_start = time.time()\r\n ans = dataset['train'].to_json(SAVE_LOC, batch_size=batch, num_proc=num)\r\n local_end = time.time() - local_start\r\n\r\n print(f\"Time taken on {num} num_proc and {batch} batch_size: \", local_end)\r\n\r\n # remove that dataset and its contents from cache and newly generated json\r\n new_json = pathlib.Path(SAVE_LOC)\r\n new_json.unlink()\r\n\r\n try:\r\n shutil.rmtree(os.path.join(str(Path.home()), \".cache\", \"huggingface\"))\r\n except OSError as e:\r\n print(\"Error: %s - %s.\" % (e.filename, e.strerror))\r\n\r\n gc.collect()\r\n```\r\nThis will download the dataset in every iteration and run `to_json`. I didn't do multiple iterations here for `to_json` (for a specific batch_size and num_proc) and took average time as I found that v1 got faster after 1st iteration (maybe it's caching somewhere). Since you'll be doing this operation only once, I thought it'll be better to report how both v1 and v2 performed in single iteration only. \r\n\r\nImportant: Benchmarking script will delete the newly generated json and `~/.cache/huggingface/` after every iteration so that it doesn't end up using any cached data (just to be on a safe side)", "Thank you for sharing the benchmark, @bhavitvyamalik. Your results look promising.\r\n\r\nBut if I remember correctly the sharded version at https://github.com/huggingface/datasets/issues/2663#issue-946552273 was much faster. So we probably should compare to it as well? And if it's faster than at least document that manual sharding version?\r\n\r\n-------\r\n\r\nThat's a dangerous benchmark as it'd wipe out many other HF things. Why not wipe out:\r\n```\r\n~/.cache/huggingface/datasets/ascent_kb/\r\n```\r\n\r\nRunning the benchmark now.", "Weird, I tried to adapt your benchmark to using shards and the program no longer works. It instead quickly uses up all available RAM and hangs. Has something changed recently in `datasets`? You can try:\r\n\r\n```\r\nimport time\r\nfrom datasets import load_dataset\r\nimport pathlib\r\nimport os\r\nfrom pathlib import Path\r\nimport shutil\r\nimport gc\r\nfrom multiprocessing import cpu_count, Process, Queue\r\n\r\nbatch_sizes = [10_000, 50_000, 100_000, 125_000]\r\nnum_procs = [1, 8] # change this according to your machine\r\n\r\nDATASET_NAME = (\"ascent_kb\")\r\nnum_shards = [1, 8]\r\nfor batch in batch_sizes:\r\n for shards in num_shards:\r\n dataset = load_dataset(DATASET_NAME)[\"train\"]\r\n #print(dataset)\r\n\r\n def process_shard(idx):\r\n print(f\"Sharding {idx}\")\r\n ds_shard = dataset.shard(shards, idx, contiguous=True)\r\n # ds_shard = ds_shard.shuffle() # remove contiguous=True above if shuffling\r\n print(f\"Saving {DATASET_NAME}-{idx}.jsonl\")\r\n ds_shard.to_json(f\"{DATASET_NAME}-{idx}.jsonl\", orient=\"records\", lines=True, force_ascii=False)\r\n\r\n local_start = time.time()\r\n queue = Queue()\r\n processes = [Process(target=process_shard, args=(idx,)) for idx in range(shards)]\r\n for p in processes:\r\n p.start()\r\n\r\n for p in processes:\r\n p.join()\r\n local_end = time.time() - local_start\r\n\r\n print(f\"Time taken on {shards} shards and {batch} batch_size: \", local_end)\r\n```\r\n\r\nJust careful, so that it won't crash your compute environment. As it almost crashed mine.", "So this part seems to no longer work:\r\n```\r\n dataset = load_dataset(\"ascent_kb\")[\"train\"]\r\n ds_shard = dataset.shard(1, 0, contiguous=True)\r\n ds_shard.to_json(\"ascent_kb-0.jsonl\", orient=\"records\", lines=True, force_ascii=False)\r\n```", "If you are using `to_json` without any `num_proc`or `num_proc=1` then essentially it'll fall back to v1 only and I've kept it as it is (the tests were passing as well)\r\n\r\n> That's a dangerous benchmark as it'd wipe out many other HF things. Why not wipe out:\r\n\r\nThat's because some dataset related files were still left inside `~/.cache/huggingface/datasets` folder. You can wipe off datasets folder inside your cache maybe\r\n\r\n> dataset = load_dataset(\"ascent_kb\")[\"train\"]\r\n> ds_shard = dataset.shard(1, 0, contiguous=True)\r\n> ds_shard.to_json(\"ascent_kb-0.jsonl\", orient=\"records\", lines=True, force_ascii=False)\r\n\r\nI tried this `lama` dataset (1.3M) and it worked fine. Trying it with `ascent_kb` currently, will update it here.", "I don't think the issue has anything to do with your work, @bhavitvyamalik. I forgot to mention I tested to see the same problem with the latest datasets release.\r\n\r\nInteresting, I tried your suggestion. This:\r\n```\r\npython -c 'import datasets; ds=\"lama\"; dataset = datasets.load_dataset(ds)[\"train\"]; \\\r\ndataset.shard(1, 0, contiguous=True).to_json(f\"{ds}-0.jsonl\", orient=\"records\", lines=True, force_ascii=False)'\r\n```\r\nworks fine and takes just a few GBs to complete.\r\n\r\nthis on the other hand blows up memory-wise:\r\n```\r\npython -c 'import datasets; ds=\"ascent_kb\"; dataset = datasets.load_dataset(ds)[\"train\"]; \\\r\ndataset.shard(1, 0, contiguous=True).to_json(f\"{ds}-0.jsonl\", orient=\"records\", lines=True, force_ascii=False)'\r\n```\r\nand I have to kill it before it uses up all RAM. (I have 128GB of it, so it should be more than enough)", "> That's because some dataset related files were still left inside ~/.cache/huggingface/datasets folder. You can wipe off datasets folder inside your cache maybe\r\n\r\nI think recent datasets added a method that will print out the path for all the different components for a given dataset, I can't recall the name though. It was when we were discussing a janitor program to clear up space selectively.", "> and I have to kill it before it uses up all RAM. (I have 128GB of it, so it should be more than enough)\r\n\r\nSame thing just happened on my machine too. Memory leak somewhere maybe? Even if you were to load this dataset in your memory it shouldn't take more than 4GB. You were earlier doing this for `oscar` dataset. Is it working fine for that?", "Hmm, looks like `datasets` has changed and won't accept my currently cached oscar-en (crashes), so I'd rather not download 0.5TB again. \r\n\r\nWere you able to reproduce the memory blow up with `ascent_kb`? It's should be a much quicker task to verify.\r\n\r\nBut yes, oscar worked just fine with `.shard()` which is what I used to process it fast.", "What I tried is:\r\n```\r\nHF_DATASETS_OFFLINE=1 HF_DATASETS_CACHE=cache python -c 'import datasets; ds=\"oscar\"; \\\r\ndataset = datasets.load_dataset(ds, \"unshuffled_deduplicated_en\")[\"train\"]; \\\r\ndataset.shard(1000000, 0, contiguous=True).to_json(f\"{ds}-0.jsonl\", orient=\"records\", lines=True, force_ascii=False)'\r\n```\r\nand got:\r\n```\r\nUsing the latest cached version of the module from /gpfswork/rech/six/commun/modules/datasets_modules/datasets/oscar/e4f06cecc7ae02f7adf85640b4019bf476d44453f251a1d84aebae28b0f8d51d (last modified on Fri Aug 6 01:52:35 2021) since it couldn't be found locally at oscar/oscar.py or remotely (OfflineModeIsEnabled).\r\nReusing dataset oscar (cache/oscar/unshuffled_deduplicated_en/1.0.0/e4f06cecc7ae02f7adf85640b4019bf476d44453f251a1d84aebae28b0f8d51d)\r\nTraceback (most recent call last):\r\n File \"<string>\", line 1, in <module>\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/load.py\", line 755, in load_dataset\r\n ds = builder_instance.as_dataset(split=split, ignore_verifications=ignore_verifications, in_memory=keep_in_memory)\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/builder.py\", line 737, in as_dataset\r\n datasets = utils.map_nested(\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/utils/py_utils.py\", line 203, in map_nested\r\n mapped = [\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/utils/py_utils.py\", line 204, in <listcomp>\r\n _single_map_nested((function, obj, types, None, True)) for obj in tqdm(iterable, disable=disable_tqdm)\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/utils/py_utils.py\", line 142, in _single_map_nested\r\n return function(data_struct)\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/builder.py\", line 764, in _build_single_dataset\r\n ds = self._as_dataset(\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/builder.py\", line 834, in _as_dataset\r\n dataset_kwargs = ArrowReader(self._cache_dir, self.info).read(\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 217, in read\r\n return self.read_files(files=files, original_instructions=instructions, in_memory=in_memory)\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 238, in read_files\r\n pa_table = self._read_files(files, in_memory=in_memory)\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 173, in _read_files\r\n pa_table: Table = self._get_table_from_filename(f_dict, in_memory=in_memory)\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 308, in _get_table_from_filename\r\n table = ArrowReader.read_table(filename, in_memory=in_memory)\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/arrow_reader.py\", line 327, in read_table\r\n return table_cls.from_file(filename)\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/table.py\", line 450, in from_file\r\n table = _memory_mapped_arrow_table_from_file(filename)\r\n File \"/gpfswork/rech/six/commun/conda/stas/lib/python3.8/site-packages/datasets/table.py\", line 43, in _memory_mapped_arrow_table_from_file\r\n memory_mapped_stream = pa.memory_map(filename)\r\n File \"pyarrow/io.pxi\", line 782, in pyarrow.lib.memory_map\r\n File \"pyarrow/io.pxi\", line 743, in pyarrow.lib.MemoryMappedFile._open\r\n File \"pyarrow/error.pxi\", line 122, in pyarrow.lib.pyarrow_internal_check_status\r\n File \"pyarrow/error.pxi\", line 99, in pyarrow.lib.check_status\r\nOSError: Memory mapping file failed: Cannot allocate memory\r\n```", "> Were you able to reproduce the memory blow up with ascent_kb? It's should be a much quicker task to verify.\r\n\r\nYes, this blows up memory-wise on my machine too. \r\n\r\nI found that a [similar error](https://discuss.huggingface.co/t/saving-memory-with-run-mlm-py-with-wikipedia-datasets/4160) was posted on the forum on 5th March. Since you already knew how much time [#2663 comment](https://github.com/huggingface/datasets/issues/2663#issue-946552273) took, can you try benchmarking v1 and v2 for now maybe until we have a fix for this memory blow up?", "OK, so I benchmarked using \"lama\" though it's too small for this kind of test, since the sharding is much slower than one thread here.\r\n\r\nResults: https://gist.github.com/stas00/dc1597a1e245c5915cfeefa0eee6902c\r\n\r\nSo sharding does really bad there, and your json over procs is doing great!\r\n\r\nAny suggestions to a somewhat bigger dataset, but not too big? say 10 times of lama?", "Looks great! I had a few questions/suggestions related to `benchmark-datasets-to_json.py`:\r\n \r\n1. You have used only 10_000 and 100_000 batch size. Including more batch sizes may help you find the perfect batch size for your machine and even give you some extra speed-up. \r\nFor eg, I found `load_dataset(\"cc100\", lang=\"eu\")` with batch size 125_000 took less time as compared to batch size 100_000 (71.16 sec v/s 67.26 sec) since this dataset has 2 fields only `['id', 'text']`, so that's why we can go for higher batch size here. \r\n \r\n2. Why have you used `num_procs` 1 and 4 only? \r\n\r\nYou can use:\r\n1. `dataset = load_dataset(\"cc100\", lang=\"af\")`. Even though it has only 2 fields but there are around 9.9 mil samples. (lama had around 1.3 mil samples)\r\n2. `dataset = load_dataset(\"cc100\", lang=\"eu\")` -> 16 mil samples. (if you want something more than 9.9 mil)\r\n3. `dataset = load_dataset(\"neural_code_search\", 'search_corpus')` -> 4.7 mil samples", "Thank you, @bhavitvyamalik \r\n\r\nMy apologies, at the moment I have not found time to do more benchmark with the proposed other datasets. I will try to do it later, but I don't want it to hold your PR, it's definitely a great improvement based on the benchmarks I did run! And the comparison to sharded is really just of interest to me to see if it's on par or slower.\r\n\r\nSo if other reviewers are happy, this definitely looks like a great improvement to me and addresses the request I made in the first place.\r\n\r\n> Why have you used num_procs 1 and 4 only?\r\n\r\nOh, no particular reason, I was just comparing to 4 shards on my desktop. Typically it's sufficient to go from 1 to 2-4 to see whether the distributed approach is faster or not. Once hit larger numbers you often run into bottlenecks like IO, and then numbers can be less representative. I hope it makes sense.", "Tested it with a larger dataset (`srwac`) and memory utilisation remained constant with no swap memory used. @lhoestq should I also add test for the same? Last time I tried this, I got `OSError: [Errno 12] Cannot allocate memory` in CircleCI tests" ]

[]

[]

[ { "color": "E3165C", "default": false, "description": "", "id": 4190228726, "name": "transfer-to-evaluate", "node_id": "LA_kwDODunzps75wdD2", "url": "https://api.github.com/repos/huggingface/datasets/labels/transfer-to-evaluate" } ]

[ "_The documentation is not available anymore as the PR was closed or merged._", "Metrics are deprecated in `datasets` and `evaluate` should be used instead: https://github.com/huggingface/evaluate" ]

[ { "color": "e99695", "default": false, "description": "Requesting to add a new dataset", "id": 2067376369, "name": "dataset request", "node_id": "MDU6TGFiZWwyMDY3Mzc2MzY5", "url": "https://api.github.com/repos/huggingface/datasets/labels/dataset%20request" } ]

[ "I started working on that." ]

[]

[ "Hi ! you can speed it up using multiprocessing by passing `num_proc=` to `load_dataset()`", "already did，but not useful for step Generating train split，it works in step \"Resolving data files\" & \"Downloading data files\" ", "@mariosasko some advice ， thanks！", "I met the same problem, terrible experience", "@mariosasko ", "We need more info about the issue to provide help. \r\n\r\nCan you interrupt the process (with `num_proc=None`) after the `load_dataset` call when the slowdown occurs? So we can know what part of the code is causing it.\r\n\r\nThe `audiofolder` \\ `imagefolder` with metadata is not performant for large datasets. Luckily, we can make them much faster if drop the nested metadata files feature (not that useful). I plan to work on this soon.\r\n\r\nIn the meantime, it's better to use `Dataset.from_generator` (requires replacing the `load_dataset` calls in the transformers script with `Dataset.from_generator`) or write a dataset loading script for large datasets.", "Can you interrupt the process (with num_proc=None) after the load_dataset call when the slowdown occurs? So we can know what part of the code is causing it.\r\n（I'll try this operation）\r\nThe audiofolder \\ imagefolder with metadata is not performant for large datasets. Luckily, we can make them much faster if drop the nested metadata files feature (not that useful). I plan to work on this soon.\r\n(My data is indeed a bit large, exceeding 10000 hours of audio data. Looking forward to your improvement work very much)\r\n\r\nIn the meantime, it's better to use Dataset.from_generator (requires replacing the load_dataset calls in the transformers script with Dataset.from_generator) or write a dataset loading script for large datasets.\r\n（I want to use Dataset.from_generator instead of load_dataset ，where can i found sample code to load audio&label dataset， I was to do asr task）", "Can you interrupt the process (with num_proc=None) after the load_dataset call when the slowdown occurs? So we can know what part of the code is causing it.\r\n================================================================================\r\nHere is the log：\r\n[load_dataset.log](https://github.com/huggingface/datasets/files/12169362/load_dataset.log)\r\n（The larger my training data, the slower it loads）\r\n\r\n\r\n", "In the meantime, it's better to use Dataset.from_generator (requires replacing the load_dataset calls in the transformers script with Dataset.from_generator) or write a dataset loading script for large datasets.\r\n================================================================================\r\nI tried ‘Dataset. from_generator’ implements data loading, but the testing results show no improvement" ]

[ { "color": "e99695", "default": false, "description": "Requesting to add a new dataset", "id": 2067376369, "name": "dataset request", "node_id": "MDU6TGFiZWwyMDY3Mzc2MzY5", "url": "https://api.github.com/repos/huggingface/datasets/labels/dataset%20request" } ]

[ "Hey @yjernite. Was going to start working on this but found taskmaster 1,2 & 3 in the datasets library already so think this can be closed now?", "You are absolutely right :) \r\n\r\nClosed by https://github.com/huggingface/datasets/pull/1193 https://github.com/huggingface/datasets/pull/1197 https://github.com/huggingface/datasets/pull/1213" ]

[]

[]

[ { "color": "0075ca", "default": true, "description": "Improvements or additions to documentation", "id": 1935892861, "name": "documentation", "node_id": "MDU6TGFiZWwxOTM1ODkyODYx", "url": "https://api.github.com/repos/huggingface/datasets/labels/documentation" } ]

[ "The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_4778). All of your documentation changes will be reflected on that endpoint.", "I would rather have a section in the docs that explains how to modify the script of an existing dataset (`inspect_dataset` + modification + `load_dataset`) instead of focusing on the GH datasets bundled with the source (only applicable for devs).", "Good idea! I went with @mariosasko's suggestion to use `inspect_dataset` instead of cloning a dataset repository since it's a good opportunity to show off more of the library's lesser-known functions if that's ok with everyone :)", "One advantage of cloning the repo is that it fetches potential data files referenced inside a script using relative paths, so if we decide to use `inspect_dataset`, we should at least add a tip to explain this limitation and how to circumvent it.", "Oh you're right. Calling `load_dataset` on the modified script without having the files that come with it is not ideal. I agree it should be `git clone` instead - and inspect is for inspection only ^^'" ]

[]

[ "CI seems to fail due to languages incorrectly being flagged as invalid, I guess that's related to the currently-broken bcp47 validation (see #4304)", "_The documentation is not available anymore as the PR was closed or merged._" ]

[ { "color": "0e8a16", "default": false, "description": "Contribution to a dataset script", "id": 4564477500, "name": "dataset contribution", "node_id": "LA_kwDODunzps8AAAABEBBmPA", "url": "https://api.github.com/repos/huggingface/datasets/labels/dataset%20contribution" } ]

[ "_The documentation is not available anymore as the PR was closed or merged._" ]

[]

[]

[]

[ "Fix #3089.", "@albertvillanova all tests are passing now. Either you or @lhoestq can review it!" ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi, @Chenfei-Kang.\r\n\r\nI'm sorry, but I'm not able to reproduce your bug:\r\n```python\r\nfrom datasets import load_dataset\r\n\r\nds = load_dataset(\"glue\", 'cola')\r\nds\r\n```\r\n```\r\nDatasetDict({\r\n train: Dataset({\r\n features: ['sentence', 'label', 'idx'],\r\n num_rows: 8551\r\n })\r\n validation: Dataset({\r\n features: ['sentence', 'label', 'idx'],\r\n num_rows: 1043\r\n })\r\n test: Dataset({\r\n features: ['sentence', 'label', 'idx'],\r\n num_rows: 1063\r\n })\r\n})\r\n```\r\n\r\nCould you please give more details and environment info (platform, PyArrow version)?", "> Hi, @Chenfei-Kang.\r\n> \r\n> I'm sorry, but I'm not able to reproduce your bug:\r\n> \r\n> ```python\r\n> from datasets import load_dataset\r\n> \r\n> ds = load_dataset(\"glue\", 'cola')\r\n> ds\r\n> ```\r\n> \r\n> ```\r\n> DatasetDict({\r\n> train: Dataset({\r\n> features: ['sentence', 'label', 'idx'],\r\n> num_rows: 8551\r\n> })\r\n> validation: Dataset({\r\n> features: ['sentence', 'label', 'idx'],\r\n> num_rows: 1043\r\n> })\r\n> test: Dataset({\r\n> features: ['sentence', 'label', 'idx'],\r\n> num_rows: 1063\r\n> })\r\n> })\r\n> ```\r\n> \r\n> Could you please give more details and environment info (platform, PyArrow version)?\r\n\r\nSorry to reply you so late.\r\nplatform: pycharm 2021 + anaconda with python 3.7\r\nPyArrow version: 5.0.0\r\nhuggingface-hub: 0.0.16\r\ndatasets: 1.9.0\r\n", "- For the platform, we need to know the operating system of your machine. Could you please run the command `datasets-cli env` and copy-and-paste its output below?\r\n- In relation with the error, you just gave us the error type and message (`TypeError: 'NoneType' object is not callable`). Could you please copy-paste the complete stack trace, so that we know exactly which part of the code threw the error?", "> * For the platform, we need to know the operating system of your machine. Could you please run the command `datasets-cli env` and copy-and-paste its output below?\r\n> * In relation with the error, you just gave us the error type and message (`TypeError: 'NoneType' object is not callable`). Could you please copy-paste the complete stack trace, so that we know exactly which part of the code threw the error?\r\n\r\n1. For the platform, here are the output:\r\n - datasets` version: 1.11.0\r\n - Platform: Windows-10-10.0.19041-SP0\r\n - Python version: 3.7.10\r\n - PyArrow version: 5.0.0\r\n2. For the code and error：\r\n ```python\r\n from datasets import load_dataset, load_metric\r\n dataset = load_dataset(\"glue\", \"cola\")\r\n ```\r\n ```python\r\n Traceback (most recent call last):\r\n ....\r\n ....\r\n File \"my_file.py\", line 2, in <module>\r\n dataset = load_dataset(\"glue\", \"cola\")\r\n File \"My environments\\lib\\site-packages\\datasets\\load.py\", line 830, in load_dataset\r\n **config_kwargs,\r\n File \"My environments\\lib\\site-packages\\datasets\\load.py\", line 710, in load_dataset_builder\r\n **config_kwargs,\r\n TypeError: 'NoneType' object is not callable\r\n ```\r\n Thank you!", "For that environment, I am sorry but I can't reproduce the bug: I can load the dataset without any problem.", "One naive question: do you have internet access from the machine where you execute the code?", "> For that environment, I am sorry but I can't reproduce the bug: I can load the dataset without any problem.\r\n\r\nBut I can download other task dataset such as `dataset = load_dataset('squad')`. I don't know what went wrong. Thank you so much!", "Hi,friends. I meet the same problem. Do you have a way to fix this? Thanks!\r\n" ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" }, { "color": "cfd3d7", "default": true, "description": "This issue or pull request already exists", "id": 1935892865, "name": "duplicate", "node_id": "MDU6TGFiZWwxOTM1ODkyODY1", "url": "https://api.github.com/repos/huggingface/datasets/labels/duplicate" } ]

[ "Hi @willowdong! These issues were fixed on master. We will have a new release of `datasets` later today. In the meantime, you can avoid these issues by installing `datasets` from master as follows:\r\n```bash\r\npip install git+https://github.com/huggingface/datasets.git\r\n```", "You should force redownload:\r\n```python\r\ndataset = load_dataset(\"multi_news\", download_mode=\"force_redownload\")\r\ndataset_2 = load_dataset(\"reddit_tifu\", \"long\", download_mode=\"force_redownload\")", "Fixed by:\r\n- #3787 \r\n- #3843" ]

[]

[ "fixed @lhoestq " ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi! Yes, that makes sense! Would you like to submit a PR to fix this?", "Created PR https://github.com/huggingface/datasets/pull/3971" ]

[]

[]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi ! Maybe one of the objects in the function is not deterministic across sessions ? You can read more about it and how to investigate here: https://huggingface.co/docs/datasets/about_cache", "Hi @apsdehal! Can you verify that replacing\r\n```python\r\ndef fetch_single_image(image_url, timeout=None, retries=0):\r\n for _ in range(retries + 1):\r\n try:\r\n request = urllib.request.Request(\r\n image_url,\r\n data=None,\r\n headers={\"user-agent\": get_datasets_user_agent()},\r\n )\r\n with urllib.request.urlopen(request, timeout=timeout) as req:\r\n image = PIL.Image.open(io.BytesIO(req.read()))\r\n break\r\n except Exception:\r\n image = None\r\n return image\r\n```\r\nwith \r\n```python\r\nUSER_AGENT = get_datasets_user_agent()\r\n\r\ndef fetch_single_image(image_url, timeout=None, retries=0):\r\n for _ in range(retries + 1):\r\n try:\r\n request = urllib.request.Request(\r\n image_url,\r\n data=None,\r\n headers={\"user-agent\": USER_AGENT},\r\n )\r\n with urllib.request.urlopen(request, timeout=timeout) as req:\r\n image = PIL.Image.open(io.BytesIO(req.read()))\r\n break\r\n except Exception:\r\n image = None\r\n return image\r\n```\r\nfixes the issue?", "Thanks @mariosasko. That does fix the issue. In general, I think these image downloading utilities since they are being used by a lot of image dataset should be provided as a part of `datasets` library right to keep the logic consistent and READMEs smaller? If they already exists, that is also great, please point me to those. I saw that `http_get` does exist.", "You can find my rationale (and a proposed solution) for why these utilities are not a part of `datasets` here: https://github.com/huggingface/datasets/pull/4100#issuecomment-1097994003.", "Makes sense. But, I think as the number of image datasets as grow, more people are copying pasting original code from docs to work as it is while we make fixes to them later. I think we do need a central place for these to avoid that confusion as well as more easier access to image datasets. Should we restart that discussion, possible on slack?" ]

[]

[]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._" ]

[]

[ "@lhoestq I added the README files, and now the tests fail... (check commit history, only changed MD file)\r\nThe tests seem a bit unstable", "the `RemoteDatasetTest ` errors in the CI are fixed on master so it's fine", "merging since the CI is fixed on master" ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi! It seems like `flatten_indices` (called in `unique`) doesn't know how to handle empty indices mappings. I'm working on the fix.", "Thanks, that was fast!" ]

[]

[ "Currently this dataset is only available when the library is installed from source since it was added after the last release.\r\n\r\nWe pin the dataset version with the library version so that people can have a reproducible dataset and processing when pinning the library.\r\n\r\nWe'll see if we can provide access to newer datasets with a warning that they are newer than your library version, that would help in cases like yours.", "So for now, should i try and install the library from source and then try out the same piece of code? Will it work then, considering both the versions will match then?", "Yes", "Hey, so i tried installing the library from source using the commands : **git clone https://github.com/huggingface/datasets**, **cd datasets** and then **pip3 install -e .**. But i still am facing the same error that file is not found. Please advise.\r\n\r\nThe Datasets library version now is 1.1.3 by installing from source as compared to the earlier 1.0.3 that i had loaded using pip command but I am still getting same error\r\n\r\n\r\n", "Looks like the wikipedia dump for hindi at the date of 05/05/2020 is not available anymore.\r\nYou can try to load a more recent version of wikipedia\r\n```python\r\nfrom datasets import load_dataset\r\n\r\nd = load_dataset(\"wikipedia\", language=\"hi\", date=\"20210101\", split=\"train\", beam_runner=\"DirectRunner\")\r\n```", "Okay, thank you so much" ]

[]

[ "From my limited understanding, part of the issue seems related to the `prepare_module` and `download_and_prepare` functions each handling the case where no config is passed. For example, `prepare_module` does mutate the object passed and forces the flags `extract_compressed_file` and `force_extract` to `True`.\r\n\r\nSee:\r\n* https://github.com/huggingface/nlp/blob/5fb61e1012bda724a9b6b847307d90a1380abfa5/src/nlp/load.py#L227\r\n* https://github.com/huggingface/nlp/blob/5fb61e1012bda724a9b6b847307d90a1380abfa5/src/nlp/builder.py#L388\r\n\r\nMaybe a cleaner solution would be to always instantiate a default `DownloadConfig` object at the top-level, have it as non-optional for the lower-level functions and treat it as immutable. ", "Thanks for the report, I'll take a look.\r\n\r\nWhat is your specific use-case for providing a DownloadConfig object?\r\n", "Thanks. Our use case involves running a training job behind a corporate firewall with no access to any external resources (S3, GCP or other web resources).\r\n\r\nI was thinking about a 2-steps process:\r\n1) Download the resources / artifacts using some secure corporate channel, ie run `nlp.load_dataset()` without a specific `DownloadConfig`. After that, collect the files from the `$HF_HOME` folder\r\n2) Copy the `$HF_HOME` folder in the firewalled environment. Run `nlp.load_dataset()` with a custom config `DownloadConfig(local_files_only=True)`\r\n\r\nHowever this ends up a bit clunky in practice, even when solving the `DownloadConfig` issue above. For example, the `filename` hash computed in `get_from_cache()` differs in the `local_files_only=False` vs `local_files_only=True` case (local case defaults `etag` to `None`, which results in a different hash). So effectively step 2) above doesn't work because the hash computed differs from the hash in the cache folder. Some hacks / workaround are possible but this solution becomes very convoluted.\r\nhttps://github.com/huggingface/nlp/blob/c214aa5a4430c1df1bcd0619fd94d6abdf9d2da7/src/nlp/utils/file_utils.py#L417\r\n\r\nWould you recommend a different path?\r\n", "I see.\r\n\r\nProbably the easiest way for you would be that we add simple serialization/deserialization methods to the Dataset and DatasetDict objects once the data files have been downloaded and all the dataset is processed.\r\n\r\nWhat do you think @lhoestq ?", "This use-case will be solved with #571 ", "Thank you very much @thomwolf and @lhoestq we will give it a try" ]

[]

[]

[]

[]

[ { "color": "0075ca", "default": true, "description": "Improvements or additions to documentation", "id": 1935892861, "name": "documentation", "node_id": "MDU6TGFiZWwxOTM1ODkyODYx", "url": "https://api.github.com/repos/huggingface/datasets/labels/documentation" } ]

[ "I can work on this. The link to the tutorial seems to be broken though @polinaeterna. ", "@isunitha98selvan would be great, thank you! which link are you talking about? I think it should work: https://huggingface.co/docs/datasets/create_dataset" ]

[]

[ "This PR was already approved by @lhoestq in #456 . This one just make style to remove some typos" ]

[]

[]

[]

[ "Cool thank you ! Maybe we can just replace `os.rename` by `shutil.move` instead ?", "> Cool thank you ! Maybe we can just replace `os.rename` by `shutil.move` instead ?\r\n\r\nYes, I think that could be a better solution, but I didn't test it in Linux (e.g. Ubuntu) to guarantee that `os.rename()` could be completely replaced by `shutil.move()`.", "AFAIK `shutil.move` does call `os.rename` first before doing extra work to make it work on windows, so this is should be a safe safe change for linux ;)", "> AFAIK `shutil.move` does call `os.rename` first before doing extra work to make it work on windows, so this is should be a safe safe change for linux ;)\r\n\r\nalright, let me change the PR then.", "The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/datasets/pr_4825). All of your documentation changes will be reflected on that endpoint.", "Hi @lhoestq looks like one of the tests failed, but is not related to this change, do I need to do something from my side?" ]

[]

[ "Normally, I'm expecting errors from the validation of the README file... 😅 ", "That is:\r\n```\r\n=========================== short test summary info ============================\r\nFAILED tests/test_dataset_cards.py::test_changed_dataset_card[blog_authorship_corpus]\r\n==== 1 failed, 3182 passed, 2763 skipped, 16 warnings in 201.23s (0:03:21) =====\r\n```", "@lhoestq, apart from the dataset card, everything is OK with this PR: I tested it locally." ]

[]

[]

[]

[ "For each dataset, If there exist a `dataset_info.json`, then the command `nlp-cli test path/to/my/dataset --al_configs` is successful only if the `dataset_infos.json` is correct. The infos are correct if the size and checksums of the downloaded file are correct, and if the number of examples in each split are correct.\r\n\r\nNote: the `test` command is supposed to test the script, that's why it runs the script even if the cached files already exist. Let me know if it's good to you.", "> For each dataset, If there exist a `dataset_info.json`, then the command `nlp-cli test path/to/my/dataset --al_configs` is successful only if the `dataset_infos.json` is correct. The infos are correct if the size and checksums of the downloaded file are correct, and if the number of examples in each split are correct.\r\n> \r\n> Note: the `test` command is supposed to test the script, that's why it runs the script even if the cached files already exist. Let me know if it's good to you.\r\n\r\nDoes it have to download the whole data to check if the checksums are correct? I guess so no? ", "> > For each dataset, If there exist a `dataset_info.json`, then the command `nlp-cli test path/to/my/dataset --al_configs` is successful only if the `dataset_infos.json` is correct. The infos are correct if the size and checksums of the downloaded file are correct, and if the number of examples in each split are correct.\r\n> > Note: the `test` command is supposed to test the script, that's why it runs the script even if the cached files already exist. Let me know if it's good to you.\r\n> \r\n> Does it have to download the whole data to check if the checksums are correct? I guess so no?\r\n\r\nYes it has to download them all (unless they were already downloaded in which case it just uses the cached downloaded files)." ]

[]

[]

[ { "color": "ffffff", "default": true, "description": "This will not be worked on", "id": 1935892913, "name": "wontfix", "node_id": "MDU6TGFiZWwxOTM1ODkyOTEz", "url": "https://api.github.com/repos/huggingface/datasets/labels/wontfix" } ]

[ "Hello @lhoestq, I would like just to ask you if it is OK that I include this feature 9f32ba1 in this PR or you would prefer to have it in a separate one.\r\n\r\nI was wondering whether include also a test, but I did not find any test for the other file formats...", "Hi ! Sure that would be valuable to support .xz files. Feel free to open a separate PR for this.\r\nAnd feel free to create the first test case for extracting compressed files if you have some inspiration (maybe create test_file_utils.py ?). We can still spend more time on tests next week when the sprint is over though so don't spend too much time on it.", "@lhoestq, DONE! ;) See PR #950.", "Thanks for adding support for `.xz` files :)\r\n\r\nFeel free to rebase from master to include it in your PR", "@lhoestq DONE; I have merged instead, to avoid changing the history of my public PR ;)", "Hi @lhoestq, I would need that you generate the dataset_infos.json and the dummy data for this dataset with a bigger computer. Sorry, but my laptop did not succeed...", "Thanks for your work @albertvillanova \r\nWe'll definitely look into it after this sprint :)", "Looks like #1456 added CC100 already.\r\nThe difference with your approach is that this implementation uses the `BuilderConfig` parameters to allow the creation of custom configs for all the languages, without having to specify them in the `BUILDER_CONFIGS` class attribute.\r\nFor example even if the dataset doesn't have a config for english already, you can still load the english CC100 with\r\n```python\r\nfrom datasets import load_dataset\r\n\r\nload_dataset(\"cc100\", lang=\"en\")\r\n```", "@lhoestq, oops!! I remember having assigned this dataset to me in the Google sheet, besides having mentioned the corresponding issue in the Pull Request... Nevermind! :)", "Yes indeed I can see that...\r\nSorry for noticing that only now \r\n\r\nThe code of the other PR ended up being pretty close to yours though\r\nIf you want to add more details to the cc100 dataset card or in the script feel to do so, any addition is welcome" ]

[]

[ "Hi ! The SplitInfo is not always available. By default you would get `split_info.num_examples == 0`\r\nSo unfortunately we can't use this assertion you suggested", "> Hi ! The SplitInfo is not always available. By default you would get `split_info.num_examples == 0`\r\n> So unfortunately we can't use this assertion you suggested\r\n\r\nThen it would be better to just remove the assertion, because the existing assertion does nothing." ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi! We've recently updated the Wikipedia script, so these changes are only available on master and can be fetched as follows:\r\n```python\r\ndataset_wikipedia = load_dataset(\"wikipedia\", \"20220301.en\", revision=\"master\")\r\n```", "Hi, how can I load the previous \"20200501.en\" version of wikipedia which had been downloaded to the default path? Thanks!", "@JiaQiSJTU just reinstall the previous verision of the package, e.g. `!pip install -q datasets==1.0.0`" ]

[]

[ "Thanks for reporting, @myluki2000.\r\n\r\nI am working on a fix." ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._" ]

[]

[ "The CI failure is unrelated to your PR and fixed on master, merging :)" ]

[]

[ "@lhoestq would you be able to review this please and also approve the workflow?", "Sounds good to me :) feel free to run `make style` to apply code formatting", "_The documentation is not available anymore as the PR was closed or merged._", "cool ! I think we can merge once all comments have been addressed", "@lhoestq I just addressed the comments and I think we can move ahead with this! \r\n", "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==8.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.007734 / 0.011353 (-0.003619) | 0.004608 / 0.011008 (-0.006400) | 0.094466 / 0.038508 (0.055958) | 0.086477 / 0.023109 (0.063368) | 0.410311 / 0.275898 (0.134413) | 0.455560 / 0.323480 (0.132080) | 0.006112 / 0.007986 (-0.001874) | 0.003845 / 0.004328 (-0.000483) | 0.072506 / 0.004250 (0.068256) | 0.066721 / 0.037052 (0.029669) | 0.409967 / 0.258489 (0.151478) | 0.460480 / 0.293841 (0.166639) | 0.036700 / 0.128546 (-0.091847) | 0.009854 / 0.075646 (-0.065792) | 0.320936 / 0.419271 (-0.098335) | 0.061002 / 0.043533 (0.017469) | 0.413963 / 0.255139 (0.158824) | 0.426787 / 0.283200 (0.143588) | 0.029182 / 0.141683 (-0.112501) | 1.685136 / 1.452155 (0.232981) | 1.754590 / 1.492716 (0.261873) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.222698 / 0.018006 (0.204692) | 0.505929 / 0.000490 (0.505440) | 0.005291 / 0.000200 (0.005091) | 0.000097 / 0.000054 (0.000042) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.032527 / 0.037411 (-0.004884) | 0.094842 / 0.014526 (0.080317) | 0.110138 / 0.176557 (-0.066418) | 0.193786 / 0.737135 (-0.543349) | 0.112593 / 0.296338 (-0.183745) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.441671 / 0.215209 (0.226461) | 4.392961 / 2.077655 (2.315306) | 2.161111 / 1.504120 (0.656991) | 1.967080 / 1.541195 (0.425885) | 2.065411 / 1.468490 (0.596920) | 0.561080 / 4.584777 (-4.023697) | 4.159612 / 3.745712 (0.413900) | 6.435248 / 5.269862 (1.165386) | 3.732338 / 4.565676 (-0.833339) | 0.066156 / 0.424275 (-0.358119) | 0.008030 / 0.007607 (0.000423) | 0.532182 / 0.226044 (0.306137) | 5.315142 / 2.268929 (3.046213) | 2.680157 / 55.444624 (-52.764467) | 2.303799 / 6.876477 (-4.572677) | 2.530911 / 2.142072 (0.388838) | 0.669504 / 4.805227 (-4.135723) | 0.151940 / 6.500664 (-6.348724) | 0.066999 / 0.075469 (-0.008470) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.424275 / 1.841788 (-0.417513) | 21.550742 / 8.074308 (13.476434) | 16.031414 / 10.191392 (5.840022) | 0.194681 / 0.680424 (-0.485743) | 0.020389 / 0.534201 (-0.513812) | 0.429808 / 0.579283 (-0.149475) | 0.457503 / 0.434364 (0.023139) | 0.511522 / 0.540337 (-0.028816) | 0.682621 / 1.386936 (-0.704315) |\n\n</details>\nPyArrow==latest\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.007519 / 0.011353 (-0.003834) | 0.004445 / 0.011008 (-0.006563) | 0.071946 / 0.038508 (0.033438) | 0.082982 / 0.023109 (0.059873) | 0.459938 / 0.275898 (0.184040) | 0.504875 / 0.323480 (0.181395) | 0.005805 / 0.007986 (-0.002181) | 0.003740 / 0.004328 (-0.000589) | 0.071998 / 0.004250 (0.067747) | 0.062580 / 0.037052 (0.025527) | 0.462263 / 0.258489 (0.203774) | 0.506355 / 0.293841 (0.212514) | 0.036321 / 0.128546 (-0.092225) | 0.009830 / 0.075646 (-0.065816) | 0.079810 / 0.419271 (-0.339461) | 0.055291 / 0.043533 (0.011758) | 0.464093 / 0.255139 (0.208954) | 0.481109 / 0.283200 (0.197910) | 0.026909 / 0.141683 (-0.114774) | 1.652538 / 1.452155 (0.200383) | 1.750713 / 1.492716 (0.257997) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.267552 / 0.018006 (0.249546) | 0.502021 / 0.000490 (0.501531) | 0.001635 / 0.000200 (0.001435) | 0.000099 / 0.000054 (0.000044) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.033747 / 0.037411 (-0.003665) | 0.104242 / 0.014526 (0.089716) | 0.113829 / 0.176557 (-0.062728) | 0.176242 / 0.737135 (-0.560893) | 0.117002 / 0.296338 (-0.179336) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.476731 / 0.215209 (0.261522) | 4.727054 / 2.077655 (2.649399) | 2.589396 / 1.504120 (1.085276) | 2.511180 / 1.541195 (0.969985) | 2.634122 / 1.468490 (1.165632) | 0.563840 / 4.584777 (-4.020937) | 4.140212 / 3.745712 (0.394500) | 6.188789 / 5.269862 (0.918928) | 3.716897 / 4.565676 (-0.848780) | 0.065823 / 0.424275 (-0.358452) | 0.007705 / 0.007607 (0.000098) | 0.566580 / 0.226044 (0.340535) | 5.653306 / 2.268929 (3.384377) | 3.028756 / 55.444624 (-52.415868) | 2.592319 / 6.876477 (-4.284158) | 2.614250 / 2.142072 (0.472178) | 0.667135 / 4.805227 (-4.138093) | 0.153455 / 6.500664 (-6.347209) | 0.069321 / 0.075469 (-0.006148) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.541978 / 1.841788 (-0.299810) | 21.747360 / 8.074308 (13.673052) | 15.963657 / 10.191392 (5.772265) | 0.192843 / 0.680424 (-0.487581) | 0.020702 / 0.534201 (-0.513499) | 0.433620 / 0.579283 (-0.145663) | 0.467327 / 0.434364 (0.032963) | 0.507398 / 0.540337 (-0.032940) | 0.692797 / 1.386936 (-0.694140) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#396cf9419d12e3150e2051793b10f2c813780a90 \"CML watermark\")\n" ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[]

[ { "color": "0075ca", "default": true, "description": "Improvements or additions to documentation", "id": 1935892861, "name": "documentation", "node_id": "MDU6TGFiZWwxOTM1ODkyODYx", "url": "https://api.github.com/repos/huggingface/datasets/labels/documentation" } ]

[ "_The documentation is not available anymore as the PR was closed or merged._", "Thanks ! Let us know if we can help :)\r\n\r\nSmall pref for having multiple PRs", "Awesome, thanks! Sorry this one is a little big, I'll open some smaller ones next :)" ]

[]

[]

[ { "color": "e99695", "default": false, "description": "Requesting to add a new dataset", "id": 2067376369, "name": "dataset request", "node_id": "MDU6TGFiZWwyMDY3Mzc2MzY5", "url": "https://api.github.com/repos/huggingface/datasets/labels/dataset%20request" } ]

[ "Hey, It's already present [here](https://huggingface.co/datasets/squad_v2) ", "Hi! This dataset is indeed already available on the Hub. Closing." ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._" ]

[ { "color": "d73a4a", "default": true, "description": "Something isn't working", "id": 1935892857, "name": "bug", "node_id": "MDU6TGFiZWwxOTM1ODkyODU3", "url": "https://api.github.com/repos/huggingface/datasets/labels/bug" } ]

[ "Hi! Thanks for reporting! This looks like a bug. I've just opened a PR with the fix.", "Awesome! Thank you. I'll close the issue once the PR gets merged. :-)", "I guess we should reopen after the revert by:\r\n- #5006" ]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._" ]

[]

[ "Hi,\r\n\r\nthe `_data` attribute is missing due to `MyDataset.__init__` not calling the parent `__init__`. However, I don't think it's a good idea to subclass the `datasets.Dataset` class (e.g. it's kind of dangerous to override `datasets.Dataset.__getitem__`). Instead, it's better to follow the \"association over inheritance\" approach with a simple wrapper class that delegates calls to a wrapped `Dataset` (map, etc.). Btw, the library offers the `datasets.Dataset.from_pandas` class method to directly create a `datasets.Dataset` from the dataframe." ]

[ { "color": "25b21e", "default": false, "description": "A bug in a metric script", "id": 2067393914, "name": "metric bug", "node_id": "MDU6TGFiZWwyMDY3MzkzOTE0", "url": "https://api.github.com/repos/huggingface/datasets/labels/metric%20bug" } ]

[ "Hi ! Thanks for reporting.\r\nWe're indeed using `jiwer` to compute the WER.\r\n\r\nMaybe instead of calling `jiwer.wer` once for all the preditions/references we can compute the WER iteratively to avoid memory issues ? I'm not too familial with `jiwer` but this must be possible.\r\n\r\nCurrently the code to compute the WER is defined here:\r\n\r\nhttps://github.com/huggingface/nlp/blob/349ac4398a3bcae6356f14c5754483383a60e8a4/metrics/wer/wer.py#L93-L94", "Hi,\r\n\r\nI've just pushed a pull request that is related to this issue https://github.com/huggingface/datasets/pull/2169. It's not iterative, but it should avoid memory errors. It's based on the editdistance python library. An iterative implementation should be as easy as storing scores and words stepwise and dividing at the end. ", "I see, this was solved by other thread. Ok, let me know if you want to switch the implementation for any reason :)", "Thanks for diving into this anyway ^^'\r\nAs you said this actually got solved a few days ago", "Someone created an issue https://github.com/jitsi/jiwer/issues/40 at jiwer which shows that this is still a problem in the current version. Would be curious to figure out how this can be fixed by jiwer... :) I assume that it runs of out memory because it's trying to compute the WER over (too many) test samples?", "Hi !\r\n\r\nIt's computed iteratively so not sure what could go wrong\r\n\r\nhttps://github.com/huggingface/datasets/blob/8afd0ba8c27800a55ea69d9fcd702dc97d9c16d8/metrics/wer/wer.py#L100-L106\r\n\r\n@NiklasHoltmeyer what version of `datasets` are you running ?\r\n", "One possible explanation might be that it is the user who is passing all the sentences in a single element to `wer.compute`?\r\n\r\nAs current implementation iterates over the elements of `predictions` and `references`, this can be problematic if `predictions` and `references` contain a single huge element each. \r\n\r\nThis could be the case, for example, with a single string with all sentences:\r\n```python\r\nresult[\"predicted\"] = \"One sentence. Other sentence.\"\r\n```\r\nor with a __double__ nested list of sentence lists\r\n```python\r\nresult[\"predicted\"] = [[ [\"One sentence.\"], [\"Other sentence\"] ]]\r\n```\r\n\r\nThe user should check the dimensions of the data structure passed to `predictions` and `references`.", "Hi all,\r\n\r\nin my case I was using and older version of datasets and, as @albertvillanova points out, passing the full list of sentences for the metric calculation. The problem was in the way jiwer implements WER, as it tries to compute WER for the full list at once instead of doing it element-wise. I think that with the latest implementation of datasets, or by using the alternative WER function that I've contributed on this [pull request](https://github.com/huggingface/datasets/pull/2169) there shouldn't be memory errors.", "@lhoestq i was using Datasets==1.5.0 with 1.6.1 it worked (atleast the first run) but 1.5.0 is not compatible with my preprocessing. i cant save my dataset to a parquet file while using the latest datasets version\r\n\r\n-> \r\n```\r\n File \"../preprocess_dataset.py\", line 132, in <module>\r\n pq.write_table(train_dataset.data, f'{resampled_data_dir}/{data_args.dataset_config_name}.train.parquet')\r\n File \"/usr/local/lib/python3.8/dist-packages/pyarrow/parquet.py\", line 1674, in write_table\r\n writer.write_table(table, row_group_size=row_group_size)\r\n File \"/usr/local/lib/python3.8/dist-packages/pyarrow/parquet.py\", line 588, in write_table\r\n self.writer.write_table(table, row_group_size=row_group_size)\r\nTypeError: Argument 'table' has incorrect type (expected pyarrow.lib.Table, got ConcatenationTable)\r\n``` \r\n\r\nif i do \r\n```\r\nimport pyarrow.parquet as pq\r\n...\r\n...\r\npq.write_table(train_dataset.data, 'train.parquet')\r\npq.write_table(eval_dataset.data, 'eval.parquet')\r\n```\r\n\r\nwhile using 1.6.1. and its working with 1.5.0\r\n", "Hi ! You can pass dataset.data.table instead of dataset.data to pq.write_table", "This seems to be working so far! Thanks!" ]

[]

[]

[]

[ "_The documentation is not available anymore as the PR was closed or merged._", "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==8.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.006087 / 0.011353 (-0.005265) | 0.003567 / 0.011008 (-0.007442) | 0.079668 / 0.038508 (0.041160) | 0.063647 / 0.023109 (0.040538) | 0.323082 / 0.275898 (0.047184) | 0.348679 / 0.323480 (0.025199) | 0.004726 / 0.007986 (-0.003259) | 0.002955 / 0.004328 (-0.001373) | 0.062724 / 0.004250 (0.058473) | 0.050194 / 0.037052 (0.013142) | 0.321407 / 0.258489 (0.062918) | 0.355053 / 0.293841 (0.061212) | 0.026992 / 0.128546 (-0.101554) | 0.007994 / 0.075646 (-0.067653) | 0.260562 / 0.419271 (-0.158710) | 0.050933 / 0.043533 (0.007400) | 0.316644 / 0.255139 (0.061505) | 0.336759 / 0.283200 (0.053560) | 0.022581 / 0.141683 (-0.119101) | 1.481259 / 1.452155 (0.029104) | 1.535191 / 1.492716 (0.042475) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.194111 / 0.018006 (0.176104) | 0.448146 / 0.000490 (0.447656) | 0.000321 / 0.000200 (0.000121) | 0.000056 / 0.000054 (0.000001) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.023908 / 0.037411 (-0.013503) | 0.073316 / 0.014526 (0.058790) | 0.085588 / 0.176557 (-0.090968) | 0.145377 / 0.737135 (-0.591759) | 0.084788 / 0.296338 (-0.211550) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.439327 / 0.215209 (0.224118) | 4.384833 / 2.077655 (2.307179) | 2.322943 / 1.504120 (0.818823) | 2.147737 / 1.541195 (0.606542) | 2.226725 / 1.468490 (0.758235) | 0.502957 / 4.584777 (-4.081820) | 3.098106 / 3.745712 (-0.647606) | 4.194642 / 5.269862 (-1.075220) | 2.598820 / 4.565676 (-1.966856) | 0.057942 / 0.424275 (-0.366333) | 0.006857 / 0.007607 (-0.000750) | 0.511517 / 0.226044 (0.285472) | 5.121797 / 2.268929 (2.852868) | 2.756506 / 55.444624 (-52.688118) | 2.424602 / 6.876477 (-4.451875) | 2.608342 / 2.142072 (0.466270) | 0.589498 / 4.805227 (-4.215729) | 0.126065 / 6.500664 (-6.374600) | 0.061456 / 0.075469 (-0.014013) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.239928 / 1.841788 (-0.601860) | 18.423532 / 8.074308 (10.349224) | 13.935148 / 10.191392 (3.743756) | 0.129913 / 0.680424 (-0.550511) | 0.016744 / 0.534201 (-0.517457) | 0.333468 / 0.579283 (-0.245815) | 0.359615 / 0.434364 (-0.074749) | 0.383678 / 0.540337 (-0.156659) | 0.533007 / 1.386936 (-0.853929) |\n\n</details>\nPyArrow==latest\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.005980 / 0.011353 (-0.005373) | 0.003640 / 0.011008 (-0.007368) | 0.062500 / 0.038508 (0.023992) | 0.059843 / 0.023109 (0.036733) | 0.360993 / 0.275898 (0.085095) | 0.401981 / 0.323480 (0.078501) | 0.005495 / 0.007986 (-0.002490) | 0.002862 / 0.004328 (-0.001467) | 0.062491 / 0.004250 (0.058240) | 0.050778 / 0.037052 (0.013726) | 0.371007 / 0.258489 (0.112518) | 0.405154 / 0.293841 (0.111313) | 0.027390 / 0.128546 (-0.101156) | 0.008042 / 0.075646 (-0.067604) | 0.067590 / 0.419271 (-0.351681) | 0.042485 / 0.043533 (-0.001048) | 0.361305 / 0.255139 (0.106166) | 0.388669 / 0.283200 (0.105469) | 0.024143 / 0.141683 (-0.117540) | 1.451508 / 1.452155 (-0.000647) | 1.490431 / 1.492716 (-0.002285) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.175976 / 0.018006 (0.157970) | 0.428923 / 0.000490 (0.428434) | 0.002099 / 0.000200 (0.001899) | 0.000068 / 0.000054 (0.000014) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.026346 / 0.037411 (-0.011065) | 0.078084 / 0.014526 (0.063558) | 0.087287 / 0.176557 (-0.089269) | 0.144179 / 0.737135 (-0.592957) | 0.088286 / 0.296338 (-0.208053) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.450436 / 0.215209 (0.235227) | 4.488801 / 2.077655 (2.411146) | 2.479303 / 1.504120 (0.975184) | 2.305396 / 1.541195 (0.764201) | 2.370370 / 1.468490 (0.901879) | 0.502355 / 4.584777 (-4.082422) | 3.094733 / 3.745712 (-0.650979) | 4.062367 / 5.269862 (-1.207495) | 2.587506 / 4.565676 (-1.978170) | 0.058245 / 0.424275 (-0.366030) | 0.006487 / 0.007607 (-0.001120) | 0.524147 / 0.226044 (0.298102) | 5.236876 / 2.268929 (2.967947) | 2.897134 / 55.444624 (-52.547490) | 2.574631 / 6.876477 (-4.301846) | 2.620307 / 2.142072 (0.478235) | 0.586963 / 4.805227 (-4.218265) | 0.125761 / 6.500664 (-6.374903) | 0.062264 / 0.075469 (-0.013205) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.299668 / 1.841788 (-0.542120) | 19.004441 / 8.074308 (10.930133) | 13.841867 / 10.191392 (3.650475) | 0.159674 / 0.680424 (-0.520750) | 0.016699 / 0.534201 (-0.517502) | 0.331868 / 0.579283 (-0.247415) | 0.344604 / 0.434364 (-0.089760) | 0.379391 / 0.540337 (-0.160947) | 0.514790 / 1.386936 (-0.872146) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#47a006a90e9711b33db70b0ef2d2cefaadfa2179 \"CML watermark\")\n", "<details>\n<summary>Show benchmarks</summary>\n\nPyArrow==8.0.0\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.005792 / 0.011353 (-0.005561) | 0.003519 / 0.011008 (-0.007489) | 0.079133 / 0.038508 (0.040625) | 0.057858 / 0.023109 (0.034749) | 0.314206 / 0.275898 (0.038308) | 0.346939 / 0.323480 (0.023459) | 0.004583 / 0.007986 (-0.003403) | 0.002824 / 0.004328 (-0.001504) | 0.061652 / 0.004250 (0.057402) | 0.048520 / 0.037052 (0.011467) | 0.318018 / 0.258489 (0.059529) | 0.350350 / 0.293841 (0.056509) | 0.026284 / 0.128546 (-0.102262) | 0.007827 / 0.075646 (-0.067819) | 0.259624 / 0.419271 (-0.159647) | 0.052318 / 0.043533 (0.008786) | 0.317400 / 0.255139 (0.062261) | 0.340530 / 0.283200 (0.057331) | 0.025181 / 0.141683 (-0.116501) | 1.459208 / 1.452155 (0.007053) | 1.529158 / 1.492716 (0.036442) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.169692 / 0.018006 (0.151686) | 0.432638 / 0.000490 (0.432148) | 0.003675 / 0.000200 (0.003475) | 0.000071 / 0.000054 (0.000017) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.022956 / 0.037411 (-0.014456) | 0.071860 / 0.014526 (0.057334) | 0.082159 / 0.176557 (-0.094398) | 0.142560 / 0.737135 (-0.594576) | 0.082333 / 0.296338 (-0.214006) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.397923 / 0.215209 (0.182714) | 3.958757 / 2.077655 (1.881102) | 1.925837 / 1.504120 (0.421717) | 1.758114 / 1.541195 (0.216919) | 1.808845 / 1.468490 (0.340354) | 0.501116 / 4.584777 (-4.083661) | 3.007739 / 3.745712 (-0.737973) | 3.295755 / 5.269862 (-1.974106) | 2.123843 / 4.565676 (-2.441833) | 0.057174 / 0.424275 (-0.367101) | 0.006426 / 0.007607 (-0.001182) | 0.468196 / 0.226044 (0.242152) | 4.677392 / 2.268929 (2.408464) | 2.334179 / 55.444624 (-53.110446) | 1.989283 / 6.876477 (-4.887194) | 2.140091 / 2.142072 (-0.001981) | 0.590700 / 4.805227 (-4.214527) | 0.124066 / 6.500664 (-6.376598) | 0.059931 / 0.075469 (-0.015538) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.224547 / 1.841788 (-0.617240) | 17.866979 / 8.074308 (9.792671) | 13.142009 / 10.191392 (2.950617) | 0.147081 / 0.680424 (-0.533343) | 0.016777 / 0.534201 (-0.517424) | 0.327766 / 0.579283 (-0.251517) | 0.343988 / 0.434364 (-0.090376) | 0.383268 / 0.540337 (-0.157070) | 0.528109 / 1.386936 (-0.858827) |\n\n</details>\nPyArrow==latest\n\n<details>\n<summary>Show updated benchmarks!</summary>\n\n### Benchmark: benchmark_array_xd.json\n\n| metric | read_batch_formatted_as_numpy after write_array2d | read_batch_formatted_as_numpy after write_flattened_sequence | read_batch_formatted_as_numpy after write_nested_sequence | read_batch_unformated after write_array2d | read_batch_unformated after write_flattened_sequence | read_batch_unformated after write_nested_sequence | read_col_formatted_as_numpy after write_array2d | read_col_formatted_as_numpy after write_flattened_sequence | read_col_formatted_as_numpy after write_nested_sequence | read_col_unformated after write_array2d | read_col_unformated after write_flattened_sequence | read_col_unformated after write_nested_sequence | read_formatted_as_numpy after write_array2d | read_formatted_as_numpy after write_flattened_sequence | read_formatted_as_numpy after write_nested_sequence | read_unformated after write_array2d | read_unformated after write_flattened_sequence | read_unformated after write_nested_sequence | write_array2d | write_flattened_sequence | write_nested_sequence |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.006145 / 0.011353 (-0.005208) | 0.003634 / 0.011008 (-0.007374) | 0.062887 / 0.038508 (0.024379) | 0.062659 / 0.023109 (0.039550) | 0.362962 / 0.275898 (0.087064) | 0.405149 / 0.323480 (0.081669) | 0.004821 / 0.007986 (-0.003164) | 0.002888 / 0.004328 (-0.001441) | 0.062982 / 0.004250 (0.058732) | 0.051929 / 0.037052 (0.014877) | 0.366825 / 0.258489 (0.108336) | 0.409830 / 0.293841 (0.115989) | 0.027263 / 0.128546 (-0.101283) | 0.007972 / 0.075646 (-0.067674) | 0.067413 / 0.419271 (-0.351858) | 0.044233 / 0.043533 (0.000700) | 0.365087 / 0.255139 (0.109948) | 0.393845 / 0.283200 (0.110646) | 0.027740 / 0.141683 (-0.113943) | 1.497896 / 1.452155 (0.045741) | 1.549419 / 1.492716 (0.056703) |\n\n### Benchmark: benchmark_getitem\\_100B.json\n\n| metric | get_batch_of\\_1024\\_random_rows | get_batch_of\\_1024\\_rows | get_first_row | get_last_row |\n|--------|---|---|---|---|\n| new / old (diff) | 0.225510 / 0.018006 (0.207503) | 0.417054 / 0.000490 (0.416564) | 0.002184 / 0.000200 (0.001984) | 0.000075 / 0.000054 (0.000020) |\n\n### Benchmark: benchmark_indices_mapping.json\n\n| metric | select | shard | shuffle | sort | train_test_split |\n|--------|---|---|---|---|---|\n| new / old (diff) | 0.025503 / 0.037411 (-0.011908) | 0.076164 / 0.014526 (0.061638) | 0.086110 / 0.176557 (-0.090446) | 0.140387 / 0.737135 (-0.596748) | 0.086956 / 0.296338 (-0.209382) |\n\n### Benchmark: benchmark_iterating.json\n\n| metric | read 5000 | read 50000 | read_batch 50000 10 | read_batch 50000 100 | read_batch 50000 1000 | read_formatted numpy 5000 | read_formatted pandas 5000 | read_formatted tensorflow 5000 | read_formatted torch 5000 | read_formatted_batch numpy 5000 10 | read_formatted_batch numpy 5000 1000 | shuffled read 5000 | shuffled read 50000 | shuffled read_batch 50000 10 | shuffled read_batch 50000 100 | shuffled read_batch 50000 1000 | shuffled read_formatted numpy 5000 | shuffled read_formatted_batch numpy 5000 10 | shuffled read_formatted_batch numpy 5000 1000 |\n|--------|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 0.469667 / 0.215209 (0.254458) | 4.689915 / 2.077655 (2.612261) | 2.685000 / 1.504120 (1.180880) | 2.516160 / 1.541195 (0.974965) | 2.531733 / 1.468490 (1.063243) | 0.501675 / 4.584777 (-4.083102) | 3.000579 / 3.745712 (-0.745133) | 2.853376 / 5.269862 (-2.416486) | 1.810677 / 4.565676 (-2.754999) | 0.057632 / 0.424275 (-0.366643) | 0.006390 / 0.007607 (-0.001217) | 0.543986 / 0.226044 (0.317941) | 5.432837 / 2.268929 (3.163908) | 3.138797 / 55.444624 (-52.305827) | 2.813141 / 6.876477 (-4.063336) | 2.803681 / 2.142072 (0.661609) | 0.588736 / 4.805227 (-4.216491) | 0.125696 / 6.500664 (-6.374968) | 0.062492 / 0.075469 (-0.012977) |\n\n### Benchmark: benchmark_map_filter.json\n\n| metric | filter | map fast-tokenizer batched | map identity | map identity batched | map no-op batched | map no-op batched numpy | map no-op batched pandas | map no-op batched pytorch | map no-op batched tensorflow |\n|--------|---|---|---|---|---|---|---|---|---|\n| new / old (diff) | 1.337163 / 1.841788 (-0.504624) | 18.611715 / 8.074308 (10.537407) | 13.953016 / 10.191392 (3.761624) | 0.154670 / 0.680424 (-0.525754) | 0.016523 / 0.534201 (-0.517678) | 0.333898 / 0.579283 (-0.245385) | 0.336520 / 0.434364 (-0.097844) | 0.389032 / 0.540337 (-0.151305) | 0.529202 / 1.386936 (-0.857734) |\n\n</details>\n</details>\n\n![](https://cml.dev/watermark.png#01d4b3330f2cc243a3f3b0cd61ec5558466c40fd \"CML watermark\")\n" ]

[]