주사

  • 설명 :

다양한 분할이 있는 SCAN 작업.

SCAN은 구성 학습 및 제로샷 일반화를 연구하기 위한 간단한 언어 기반 탐색 작업 세트입니다.

대부분의 분할은 https://github.com/brendenlake/SCAN 에 설명되어 있습니다. MCD 분할은 https://arxiv.org/abs/1912.09713.pdf 를 참조하십시오.

기본 사용법:

data = tfds.load('scan/length')

고급 예제:

import tensorflow_datasets as tfds
from tensorflow_datasets.datasets.scan import scan_dataset_builder

data = tfds.load(
    'scan',
    builder_kwargs=dict(
        config=scan_dataset_builder.ScanConfig(
            name='simple_p8', directory='simple_split/size_variations')))
FeaturesDict({
    'actions': Text(shape=(), dtype=string),
    'commands': Text(shape=(), dtype=string),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
행위 텍스트
명령 텍스트
@inproceedings{Lake2018GeneralizationWS,
  title={Generalization without Systematicity: On the Compositional Skills of
         Sequence-to-Sequence Recurrent Networks},
  author={Brenden M. Lake and Marco Baroni},
  booktitle={ICML},
  year={2018},
  url={https://arxiv.org/pdf/1711.00350.pdf},
}
@inproceedings{Keysers2020,
  title={Measuring Compositional Generalization: A Comprehensive Method on
         Realistic Data},
  author={Daniel Keysers and Nathanael Sch\"{a}rli and Nathan Scales and
          Hylke Buisman and Daniel Furrer and Sergii Kashubin and
          Nikola Momchev and Danila Sinopalnikov and Lukasz Stafiniak and
          Tibor Tihon and Dmitry Tsarkov and Xiao Wang and Marc van Zee and
          Olivier Bousquet},
  note={Additional citation for MCD splits},
  booktitle={ICLR},
  year={2020},
  url={https://arxiv.org/abs/1912.09713.pdf},
}

스캔/단순(기본 구성)

  • 다운로드 크기 : 17.82 MiB

  • 데이터 세트 크기 : 4.47 MiB

  • 분할 :

나뉘다
'test' 4,182
'train' 16,728

스캔/addprim_jump

  • 다운로드 크기 : 17.82 MiB

  • 데이터 세트 크기 : 4.53 MiB

  • 분할 :

나뉘다
'test' 7,706
'train' 14,670

스캔/addprim_turn_left

  • 다운로드 크기 : 17.82 MiB

  • 데이터 세트 크기 : 4.58 MiB

  • 분할 :

나뉘다
'test' 1,208
'train' 21,890

스캔/filler_num0

  • 다운로드 크기 : 17.82 MiB

  • 데이터 세트 크기 : 3.20 MiB

  • 분할 :

나뉘다
'test' 1,173
'train' 15,225

스캔/filler_num1

  • 다운로드 크기 : 17.82 MiB

  • 데이터 세트 크기 : 3.51 MiB

  • 분할 :

나뉘다
'test' 1,173
'train' 16,290

스캔/filler_num2

  • 다운로드 크기 : 17.82 MiB

  • 데이터 세트 크기 : 3.84 MiB

  • 분할 :

나뉘다
'test' 1,173
'train' 17,391

스캔/filler_num3

  • 다운로드 크기 : 17.82 MiB

  • 데이터 세트 크기 : 4.17 MiB

  • 분할 :

나뉘다
'test' 1,173
'train' 18,528

스캔/길이

  • 다운로드 크기 : 17.82 MiB

  • 데이터 세트 크기 : 4.47 MiB

  • 분할 :

나뉘다
'test' 3,920
'train' 16,990

스캔/template_around_right

  • 다운로드 크기 : 17.82 MiB

  • 데이터 세트 크기 : 4.17 MiB

  • 분할 :

나뉘다
'test' 4,476
'train' 15,225

스캔/template_jump_around_right

  • 다운로드 크기 : 17.82 MiB

  • 데이터 세트 크기 : 4.17 MiB

  • 분할 :

나뉘다
'test' 1,173
'train' 18,528

스캔/template_opposite_right

  • 다운로드 크기 : 17.82 MiB

  • 데이터 세트 크기 : 4.22 MiB

  • 분할 :

나뉘다
'test' 4,476
'train' 15,225

스캔/템플릿_오른쪽

  • 다운로드 크기 : 17.82 MiB

  • 데이터 세트 크기 : 4.26 MiB

  • 분할 :

나뉘다
'test' 4,476
'train' 15,225

스캔/mcd1

  • 다운로드 크기 : 17.89 MiB

  • 데이터 세트 크기 : 1.89 MiB

  • 분할 :

나뉘다
'test' 1,045
'train' 8,365

스캔/mcd2

  • 다운로드 크기 : 17.89 MiB

  • 데이터 세트 크기 : 1.84 MiB

  • 분할 :

나뉘다
'test' 1,045
'train' 8,365

스캔/mcd3

  • 다운로드 크기 : 17.89 MiB

  • 데이터 세트 크기 : 1.87 MiB

  • 분할 :

나뉘다
'test' 1,045
'train' 8,365