Baseline clone

README.md

@@ -0,0 +1,31 @@
+10K slice of OpenWebText - An open-source replication of the WebText dataset from OpenAI.
+
+This is a small subset representing the first 10K records from the original dataset - created for testing.
+
+The full 8M-record dataset is [here](https://huggingface.co/datasets/openwebtext).
+
+```
+$ python -c "from datasets import load_dataset; ds=load_dataset('stas/openwebtext-10k'); print(ds)"
+DatasetDict({
+    train: Dataset({
+        features: ['text'],
+        num_rows: 10000
+    })
+})
+```
+
+* Records: 10,000
+* compressed size: ~15MB
+* uncompressed size: 50MB
+
+To convert to jsonlines:
+
+```
+from datasets import load_dataset
+dataset_name = "stas/openwebtext-10k"
+name = dataset_name.split('/')[-1]
+ds = load_dataset(dataset_name, split='train')
+ds.to_json(f"{name}.jsonl", orient="records", lines=True)
+```
+
+To see how this subset was created, here is the [instructions file](https://huggingface.co/datasets/stas/openwebtext-10k/blob/main/process.txt).

dataset_infos.json

@@ -0,0 +1 @@
+{"plain_text": {"description": "An open-source replication of the WebText dataset from OpenAI.\n\nThis is a small subset representing the first 10K records from the original dataset - created for testing.\n\nThe full 8M-record dataset is at https://huggingface.co/datasets/openwebtext\n", "citation": "@misc{Gokaslan2019OpenWeb,\n  title={OpenWebText Corpus},\n  author={Aaron Gokaslan*, Vanya Cohen*, Ellie Pavlick, Stefanie Tellex},\n  howpublished{\\url{http://Skylion007.github.io/OpenWebTextCorpus}},\n  year={2019}\n}\n", "homepage": "https://skylion007.github.io/OpenWebTextCorpus/", "license": "", "features": {"text": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "openwebtext10k", "config_name": "plain_text", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 49670861, "num_examples": 10000, "dataset_name": "openwebtext10k"}}, "download_checksums": {"https://cdn-datasets.huggingface.co/nlp/datasets/openwebtext/openwebtext-10k.tar.xz": {"num_bytes": 14723792, "checksum": "1dd150ffa3361ab32fa9f129d1b5ce20ac48728be16be436558f844d1761c572"}}, "download_size": 14723792, "post_processing_size": null, "dataset_size": 49670861, "size_in_bytes": 64394653}}

openwebtext-10k.py

@@ -0,0 +1,89 @@
+# coding=utf-8
+# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""The Open WebText Corpus"""
+
+
+import os
+import re
+from itertools import chain
+
+import datasets
+
+
+_CITATION = """\
+@misc{Gokaslan2019OpenWeb,
+  title={OpenWebText Corpus},
+  author={Aaron Gokaslan*, Vanya Cohen*, Ellie Pavlick, Stefanie Tellex},
+  howpublished{\\url{http://Skylion007.github.io/OpenWebTextCorpus}},
+  year={2019}
+}
+"""
+
+_DESCRIPTION = """\
+An open-source replication of the WebText dataset from OpenAI.
+
+This is a small subset representing the first 10K records from the original dataset - created for testing.
+
+The full 8M-record dataset is at https://huggingface.co/datasets/openwebtext
+"""
+
+_URL = "https://cdn-datasets.huggingface.co/nlp/datasets/openwebtext/openwebtext-10k.tar.xz"
+
+class Openwebtext10k(datasets.GeneratorBasedBuilder):
+    """The Open WebText dataset."""
+
+    BUILDER_CONFIGS = [
+        datasets.BuilderConfig(
+            name="plain_text",
+            description="Plain text",
+            version=datasets.Version("1.0.0"),
+        )
+    ]
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features({"text": datasets.Value("string")}),
+            homepage="https://skylion007.github.io/OpenWebTextCorpus/",
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        dl_dir = dl_manager.download_and_extract(_URL)
+        owt_dir = os.path.join(dl_dir, "openwebtext-10k")
+        subset_xzs = [
+            os.path.join(owt_dir, file_name)
+            for file_name in sorted(os.listdir(owt_dir))
+            if file_name.endswith("xz")  # filter out ...xz.lock
+        ]
+        ex_dirs = dl_manager.extract(subset_xzs, num_proc=round(os.cpu_count() * 0.75))
+        nested_txt_files = [
+            [
+                os.path.join(ex_dir, txt_file_name)
+                for txt_file_name in sorted(os.listdir(ex_dir))
+                if txt_file_name.endswith("txt")
+            ]
+            for ex_dir in ex_dirs
+        ]
+        txt_files = chain(*nested_txt_files)
+        return [
+            datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"txt_files": txt_files}),
+        ]
+
+    def _generate_examples(self, txt_files):
+        """Yields examples."""
+        for idx, filepath in enumerate(txt_files):
+            with open(filepath, encoding="utf-8") as f:
+                yield idx, {"text": re.sub("\n\n\n+", "\n\n", f.read()).strip()}

process.txt

@@ -0,0 +1,68 @@
+# --------------------------------------------------------------------------
+# Original process as following (until the end)
+
+
+# this is a small derivative from 8M-big openwebtext dataset for testing
+
+# how this build script and dataset_infos.json were generated
+
+#
+
+mkdir openwebtext-10k
+cd openwebtext-10k
+
+# data
+wget https://zenodo.org/record/3834942/files/openwebtext.tar.xz
+tar xf openwebtext.tar.xz
+cd openwebtext
+rename.pl 's|-|-00|; s|-00(\d\d\d)|-$1|; s|-00(\d\d)|-0$1|;' *xz
+
+# now open the first 30 archives
+mkdir subset
+cp urlsf_subset00-0[0-2]*_data.xz subset
+cd subset
+find . -name "*xz" -exec tar xf {} \;
+mkdir 10k
+find . -name "*txt" | sort | head -10000 | xargs mv -t 10k
+tar cfJ 10k.xz -C 10k .
+mkdir openwebtext-10k
+mv 10k.xz openwebtext-10k
+tar cfJ openwebtext-10k.tar.xz openwebtext-10k
+# the openwebtext subdir gets created on the fly
+aws s3 cp openwebtext-10k.tar.xz s3://datasets.huggingface.co/nlp/datasets/openwebtext/
+
+# script
+wget https://raw.githubusercontent.com/huggingface/datasets/master/datasets/openwebtext/openwebtext.py
+mv openwebtext.py openwebtext-10k.py
+perl -pi -e 's|openwebtext|openwebtext-10k|g' openwebtext-10k.py
+perl -pi -e 's|https://zenodo.org/record/3834942/files/|https://cdn-datasets.huggingface.co/nlp/datasets/openwebtext/|g' openwebtext-10k.py
+perl -pi -e 's|Openwebtext|Openwebtext10k|g' openwebtext-10k.py
+
+
+
+# manually check that the script is correct - edit the descriptions
+
+# create a new dataset entry on the hub
+https://huggingface.co/new-dataset
+
+# once created clone it
+git clone https://huggingface.co/datasets/stas/openwebtext-10k
+cp openwebtext-10k.py process.txt openwebtext-10k
+cd openwebtext-10k
+
+git add openwebtext-10k.py process.txt
+git commit -m "build script" openwebtext-10k.py process.txt
+git push
+
+# test and generate config file
+cd ..
+datasets-cli test ./openwebtext-10k --save_infos --all_configs
+
+# add and push the generated config
+cd openwebtext-10k
+git add dataset_infos.json
+git commit -m "add dataset_infos.json" dataset_infos.json
+git push
+
+# test that the dataset is working
+python -c "from datasets import load_dataset; ds=load_dataset('stas/openwebtext-10k'); print(ds)"