samsum.py

import json
import datasets  

class SkillDataset(datasets.GeneratorBasedBuilder):
  VERSION = datasets.Version("0.1.0")
  
  def _info(self):
    return datasets.DatasetInfo(
      description="A dataset of skills and their descriptions",
      features=datasets.Features(
        {
          "summary": datasets.Value("string"),
          "dialogue": datasets.Value("string"),
        }
      ),
    )

  def _split_generators(self, dl_manager):
    mappings = json.load(open("mappings.json", "r"))

    training_data = mappings[:int(0.95 * len(mappings))]
    testing_data = mappings[int(0.95 * len(mappings)) :]

    return [
      datasets.SplitGenerator(
        name=datasets.Split.TRAIN,
        gen_kwargs={
          "data": training_data,
          "split": "train"
        }
      ),
      datasets.SplitGenerator(
        name=datasets.Split.TEST,
        gen_kwargs={
          "data": testing_data,
          "split": "test",
        }
      ),
    ]

  def _generate_examples(self, data):
    for i, mapping in enumerate(data):
      yield i, {
        "summary": mapping["summary"],
        "dialogue": mapping["dialogue"],
      }