README.md

metadata

dataset_info:
  features:
    - name: clip-RN50
      sequence: float64
    - name: face
      sequence:
        sequence:
          sequence:
            sequence: float64
    - name: face_box
      sequence:
        sequence: float64
    - name: face_embedding
      sequence: float64
    - name: face_landmarks
      sequence:
        sequence:
          sequence: float64
    - name: face_prob
      sequence: float64
    - name: id
      dtype: string
    - name: imagenet-RN50
      sequence: float64
    - name: input
      dtype: string
    - name: kilt_id
      dtype: string
    - name: meta
      struct:
        - name: left_context
          dtype: string
        - name: mention
          dtype: string
        - name: obj_surface
          struct:
            - name: text
              sequence: 'null'
        - name: partial_evidence
          struct:
            - name: end_paragraph_id
              sequence: 'null'
            - name: meta
              sequence: 'null'
            - name: section
              sequence: 'null'
            - name: start_paragraph_id
              sequence: 'null'
            - name: title
              sequence: 'null'
            - name: wikipedia_id
              sequence: 'null'
        - name: right_context
          dtype: string
        - name: sub_surface
          struct:
            - name: text
              sequence: 'null'
        - name: subj_aliases
          struct:
            - name: text
              sequence: 'null'
        - name: template_questions
          struct:
            - name: text
              sequence: 'null'
    - name: original_question
      dtype: string
    - name: output
      struct:
        - name: answer
          sequence: string
        - name: meta
          sequence: 'null'
        - name: original_answer
          dtype: string
        - name: provenance
          list:
            - name: bleu_score
              sequence: float64
            - name: end_character
              sequence: int64
            - name: end_paragraph_id
              sequence: int64
            - name: meta
              sequence: 'null'
            - name: section
              sequence: string
            - name: start_character
              sequence: int64
            - name: start_paragraph_id
              sequence: int64
            - name: title
              sequence: string
            - name: wikipedia_id
              sequence: string
    - name: url
      dtype: string
    - name: wikidata_id
      dtype: string
    - name: search_indices
      sequence: int64
    - name: search_provenance_indices
      sequence: int64
    - name: search_irrelevant_indices
      sequence: int64
  splits:
    - name: train
      num_bytes: 313513980
      num_examples: 1165
    - name: validation
      num_bytes: 299175833
      num_examples: 1216
    - name: test
      num_bytes: 304124128
      num_examples: 1237
  download_size: 139611306
  dataset_size: 916813941
license: cc-by-4.0

Dataset Card for "miniviquae_dataset"

Reduced version of ViQuAE, see https://github.com/PaulLerner/ViQuAE