add data loading script and readme

README.md

@@ -1,3 +1,11 @@
----
-license: mit
----
+# Distilled CNN/DailyMail Dataset
+
+This folder contains the distilled data and dataset loading script to build a dataset on top of it.
+
+- `cnn_bart_pl` is downloaded from [Saved Pseudo-Labels](https://github.com/huggingface/transformers/blob/main/examples/research_projects/seq2seq-distillation/precomputed_pseudo_labels.md), which is generated by facebook/bart-large-cnn, this corresponds to version "1.0.0". It contains train/validataion/test splits.
+- `pegasus_cnn_cnn_pls` is also downloaded from [Saved Pseudo-Labels](https://github.com/huggingface/transformers/blob/main/examples/research_projects/seq2seq-distillation/precomputed_pseudo_labels.md). It is generated by sshleifer/pegasus-cnn-ft-v2, and it corresponds to version "2.0.0". It only includes the train split.
+
+
+## Updates
+- 03/16/2023
+    1. Remove "(CNN)" in the beginning of articles.

distil_cnndm.py

@@ -0,0 +1,139 @@
+# coding=utf-8
+# References:
+# (1) https://huggingface.co/datasets/cnn_dailymail/blob/main/cnn_dailymail.py
+# (2) https://huggingface.co/docs/datasets/dataset_script
+
+"""Distilled CNN/DailyMail Summarization dataset."""
+
+import os
+
+import datasets
+import nltk
+
+
+_DESCRIPTION = """\
+Distilled CNN/DailyMail non-anonymized summarization dataset.
+There are two features:
+  - article: text of news article, used as the document to be summarized
+  - highlights: joined text of highlights with <s> and </s> around each
+    highlight, which is the target summary
+
+The distilled data is obtained by running facebook/bart-large-cnn on the 
+training dataset. The files used here is directly downloaded from 
+https://github.com/huggingface/transformers/blob/main/examples/research_projects/seq2seq-distillation/precomputed_pseudo_labels.md.
+"""
+
+_CITATION = ""
+
+_DL_URLS = {
+    "cnn_bart_pl": "https://cdn-datasets.huggingface.co/pseudo/cnn_dm/cnn_bart_pl.tgz",
+    "cnn_pegasus_pl": "https://cdn-datasets.huggingface.co/pseudo/cnn_dm/pegasus_cnn_cnn_pls.tgz",
+}
+
+# as mentioned in https://github.com/huggingface/transformers/blob/main/examples/research_projects/seq2seq-distillation/precomputed_pseudo_labels.md#available-pseudo-labels,
+#   about 5K are missing, and the training should be 282173.
+_NUM_EXAMPLES = {"train": 282173, "val": 13368, "test": 11490}
+
+# maps from datasets.Split to the one used in the downloaded data.
+_SPLIT_MAP = {"train": "train", "test": "test", "validation": "val"}
+
+_SUPPORTED_VERSIONS = [
+    # Using the pseudo labels generated by BART.
+    datasets.Version("1.0.0", "Using cased version and the one generated by BART."),
+    # Using the pseudo labels generated by Pegasus.
+    datasets.Version("2.0.0", "Using cased version and the one generated by PEGASUS."),
+]
+
+_DEFAULT_VERSION = datasets.Version("2.0.0", "Using cased version.")
+
+
+class DistillCNNDMConfig(datasets.BuilderConfig):
+    """BuilderConfig for DistillCNNDM."""
+
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)
+
+
+class DistillCNNDM(datasets.GeneratorBasedBuilder):
+    """Distilled CNN/DailyMail non-anonymized summarization dataset."""
+
+    BUILDER_CONFIGS = [
+        DistillCNNDMConfig(name=str(version), description="Plain text", version=version)
+        for version in _SUPPORTED_VERSIONS
+    ]
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(
+                {
+                    "article": datasets.Value("string"),
+                    "highlights": datasets.Value("string"),
+                }
+            ),
+            supervised_keys=None,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager: datasets.DownloadManager):
+        """Split generators.
+
+        Note that the validation data have prefix val instead of
+            validation, so we use a split mapping.
+
+        Although dl_manager is not used, we still need to keep it.
+        """
+        dl_paths = dl_manager.download(_DL_URLS)
+        if self.config.version == "1.0.0":
+            return [
+                datasets.SplitGenerator(
+                    name=split,
+                    gen_kwargs={
+                        "src_path": os.path.join(
+                            dl_paths["cnn_bart_pl"], f"{_SPLIT_MAP[split]}.source"
+                        ),
+                        "tgt_path": os.path.join(
+                            dl_paths["cnn_bart_pl"], f"{_SPLIT_MAP[split]}.target"
+                        ),
+                        "num_examples": _NUM_EXAMPLES[_SPLIT_MAP[split]],
+                    },
+                )
+                for split in [
+                    datasets.Split.TRAIN,
+                    datasets.Split.VALIDATION,
+                    datasets.Split.TEST,
+                ]
+            ]
+        elif self.config.version == "2.0.0":
+            return [
+                datasets.SplitGenerator(
+                    name=datasets.Split.TRAIN,
+                    gen_kwargs={
+                        "src_path": os.path.join(
+                            dl_paths["cnn_pegasus_pl"], "train.source"
+                        ),
+                        "tgt_path": os.path.join(
+                            dl_paths["cnn_pegasus_pl"], "train.target"
+                        ),
+                        "num_examples": 287112,
+                    },
+                )
+            ]
+
+    def _generate_examples(self, src_path, tgt_path, num_examples):
+        """This function returns the examples in the raw text form.
+
+        The output article and highlights formats resemble those given
+            by `load_dataset("cnn_dailymail", "3.0.0")`.
+        """
+        with open(src_path) as src, open(tgt_path) as tgt:
+            for idx in range(num_examples):
+                article = src.readline().strip()
+                if article[:5] == "(CNN)":
+                    article = article[5:]
+                highlights = tgt.readline().strip()
+                highlights = "\n".join(nltk.sent_tokenize(highlights))
+                yield idx, {
+                    "article": article,
+                    "highlights": highlights,
+                }