ビッグベンチ

参考文献:

抽象的な物語の理解

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:bigbench/abstract_narrative_understanding')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 3000
'train' 2400
'validation' 600
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

時代錯誤

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:bigbench/anachronisms')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 230
'train' 184
'validation' 46
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

類似性

次のコマンドを使用して、このデータセットを TFDS にロードします。

ds = tfds.load('huggingface:bigbench/analogical_similarity')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 323
'train' 259
'validation' 64
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

分析含意

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/analytic_entailment')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 70
'train' 54
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

算術

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/arithmetic')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 15023
'train' 12019
'validation' 3004
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ascii_単語認識

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/ascii_word_recognition')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 5000
'train' 4000
'validation' 1000
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

著者認証_検証

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/authorship_verification')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 880
'train' 704
'validation' 176
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

自動分類

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/auto_categorization')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 328
'train' 263
'validation' 65
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

auto_debugging

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/auto_debugging')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 34
'train' 18
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

bbq_lite_json

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/bbq_lite_json')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 16076
'train' 12866
'validation' 3210
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

bridging_anaphora_resolution_barqa

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/bridging_anaphora_resolution_barqa')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 648
'train' 519
'validation' 129
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

因果関係の判断

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/causal_judgment')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 190
'train' 152
'validation' 38
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

原因と結果

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/cause_and_effect')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 153
'train' 123
'validation' 30
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

チェックメイトインワン

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/checkmate_in_one')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 3498
'train' 2799
'validation' 699
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

チェス_状態_追跡

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/chess_state_tracking')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 6000
'train' 4800
'validation' 1200
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

chinese_remainder_theorem

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/chinese_remainder_theorem')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 500
'train' 400
'validation' 100
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

cifar10_分類

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/cifar10_classification')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 20000
'train' 16000
'validation' 4000
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

コードラインの説明

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/code_line_description')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 60
'train' 44
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

コードネーム

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/codenames')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 85
'train' 68
'validation' 17
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

色

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/color')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 4000
'train' 3200
'validation' 800
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

共通形態素

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/common_morpheme')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 50
'train' 34
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

概念的な組み合わせ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/conceptual_combinations')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 103
'train' 84
'validation' 19
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

conlang_translation

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/conlang_translation')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 164
'train' 132
'validation' 32
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

context_parametric_knowledge_conflicts

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/contextual_parametric_knowledge_conflicts')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 17528
'train' 14023
'validation' 3505
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

クラッシュブロッサム

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/crash_blossom')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 38
'train' 22
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

crass_ai

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/crass_ai')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 44
'train' 28
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

冷凍生物学_スペイン語

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/cryobiology_spanish')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 146
'train' 117
'validation' 29
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

クリプトナイト

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/cryptonite')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 26157
'train' 20926
'validation' 5231
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

cs_algorithms

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/cs_algorithms')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1320
'train' 1056
'validation' 264
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ダークユーモア検出

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/dark_humor_detection')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 80
'train' 64
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

日付_理解

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/date_understanding')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 369
'train' 296
'validation' 73
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

曖昧さ回避_QA

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/disambiguation_qa')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 258
'train' 207
'validation' 51
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

談話_マーカー_予測

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/discourse_marker_prediction')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 857
'train' 686
'validation' 171
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

disfl_qa

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/disfl_qa')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 8000
'train' 6400
'validation' 1600
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ダイク_言語

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/dyck_languages')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1000
'train' 800
'validation' 200
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

初等数学QA

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/elementary_math_qa')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 38160
'train' 30531
'validation' 7629
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

emoji_movie

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/emoji_movie')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 100
'train' 80
'validation' 20
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

emojis_emotion_prediction

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/emojis_emotion_prediction')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 131
'train' 105
'validation' 26
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

経験的判断

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/empirical_judgments')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 99
'train' 80
'validation' 19
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

英語_ことわざ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/english_proverbs')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 34
'train' 18
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

英語_ロシア_ことわざ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/english_russian_proverbs')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 80
'train' 64
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

entailed_polarity

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/entailed_polarity')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 148
'train' 119
'validation' 29
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

entailed_polarity_hindi

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/entailed_polarity_hindi')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 138
'train' 111
'validation' 27
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

認識論的推論

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/epistemic_reasoning')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 2000年
'train' 1600
'validation' 400
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

評価情報の本質性

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/evaluating_information_essentiality')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 68
'train' 52
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ファクトチェッカー

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/fact_checker')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 7154
'train' 5724
'validation' 1430
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ファンタジー推理

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/fantasy_reasoning')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 201
'train' 161
'validation' 40
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

some_shot_nlg

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/few_shot_nlg')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 153
'train' 123
'validation' 30
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

音声認識図

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/figure_of_speech_detection')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 59
'train' 43
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

正式な誤謬の三理論否定

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/formal_fallacies_syllogisms_negation')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 14200
'train' 11360
'validation' 2840
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

宝石

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/gem')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 14802
'train' 11845
'validation' 2957
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ジェンダーを含む文_ドイツ語

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/gender_inclusive_sentences_german')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 200
'train' 160
'validation' 40
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

一般知識

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/general_knowledge')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 70
'train' 54
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

幾何学的形状

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/geometric_shapes')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 359
'train' 288
'validation' 71
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ゴールステップ_ウィキハウ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/goal_step_wikihow')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 7053
'train' 5643
'validation' 1410
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

gre_reading_comprehension

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/gre_reading_comprehension')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 31
'train' 15
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

hh_アライメント

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/hhh_alignment')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 221
'train' 179
'validation' 42
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ヒンディー語_質問_回答

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/hindi_question_answering')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 6610
'train' 5288
'validation' 1322
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ヒンドゥー教の知識

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/hindu_knowledge')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 175
'train' 140
'validation' 35
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ヒングリッシュ毒性

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/hinglish_toxicity')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 200
'train' 160
'validation' 40
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

人間の臓器の感覚

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/human_organs_senses')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 42
'train' 26
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ハイパーバトン

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/hyperbaton')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 50000
'train' 40000
'validation' 10000
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

数学定理の識別

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/identify_math_theorems')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 53
'train' 37
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

識別奇数メタファー

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/identify_odd_metaphor')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 47
'train' 31
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

印象

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/implicatures')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 492
'train' 394
'validation' 98
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

暗黙的な関係

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/implicit_relations')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 85
'train' 68
'validation' 17
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

意図の認識

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/intent_recognition')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 693
'train' 555
'validation' 138
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

国際音声アルファベット_nli

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/international_phonetic_alphabet_nli')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 126
'train' 101
'validation' 25
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

international_phonetic_alphabet_transliterate

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/international_phonetic_alphabet_transliterate')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1003
'train' 803
'validation' 200
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

intersect_geometry

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/intersect_geometry')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 249999
'train' 200000
'validation' 49999
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

皮肉_識別

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/irony_identification')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 99
'train' 80
'validation' 19
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

漢字_ascii

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/kanji_ascii')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1092
'train' 875
'validation' 217
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

カンナダ語

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/kannada')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 316
'train' 253
'validation' 63
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

キー値マップ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/key_value_maps')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 101
'train' 80
'validation' 21
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

既知_未知

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/known_unknowns')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 46
'train' 30
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

言語ゲーム

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/language_games')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 2128
'train' 1704年
'validation' 424
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

言語識別

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/language_identification')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 10000
'train' 8000
'validation' 2000年
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

言語マッピング

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/linguistic_mappings')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 15527
'train' 12426
'validation' 3101
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

言語学_パズル

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/linguistics_puzzles')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 2000年
'train' 1600
'validation' 400
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

リスト関数

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/list_functions')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 10750
'train' 8700
'validation' 2050年
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ロジックグリッドパズル

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/logic_grid_puzzle')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1000
'train' 800
'validation' 200
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

論理引数

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/logical_args')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 32
'train' 16
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

論理的推論

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/logical_deduction')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1500
'train' 1200
'validation' 300
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

論理的誤謬検出

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/logical_fallacy_detection')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 2800
'train' 2240
'validation' 560
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

論理シーケンス

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/logical_sequence')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 39
'train' 23
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

数学的帰納法

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/mathematical_induction')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 69
'train' 53
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

マトリックス形状

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/matrixshapes')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 4462
'train' 3570
'validation' 892
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

メタファー_ブール値

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/metaphor_boolean')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 680
'train' 544
'validation' 136
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

比喩理解

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/metaphor_understanding')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 234
'train' 188
'validation' 46
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

分_謎_qa

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/minute_mysteries_qa')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 477
'train' 383
'validation' 94
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

誤解

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/misconceptions')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 219
'train' 176
'validation' 43
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

誤解_ロシア語

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/misconceptions_russian')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 49
'train' 33
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

mnist_ascii

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/mnist_ascii')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 69984
'train' 55988
'validation' 13996
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

修正済み_算術

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/modified_arithmetic')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 6000
'train' 4800
'validation' 1200
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

道徳的許容性

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/moral_permissibility')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 342
'train' 274
'validation' 68
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

movie_dialog_同じまたは異なる

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/movie_dialog_same_or_different')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 50000
'train' 40000
'validation' 10000
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

映画_おすすめ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/movie_recommendation')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 500
'train' 400
'validation' 100
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

マルチデータラングリング

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/mult_data_wrangling')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 7854
'train' 6380
'validation' 1474年
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

マルチエモ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/multiemo')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1437281
'train' 1149873
'validation' 287408
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

自然な説明

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/natural_instructions')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 193250
'train' 154615
'validation' 38635
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/navigate')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1000
'train' 800
'validation' 200
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ナンセンスワード文法

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/nonsense_words_grammar')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 50
'train' 34
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

小説のコンセプト

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/novel_concepts')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 32
'train' 16
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

オブジェクトカウント

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/object_counting')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1000
'train' 800
'validation' 200
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

奇数ワンアウト

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/odd_one_out')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 86
'train' 69
'validation' 17
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

演算子

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/operators')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 210
'train' 168
'validation' 42
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

段落の分割

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/paragraph_segmentation')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 9000
'train' 7200
'validation' 1800
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

parsinlu_qa

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/parsinlu_qa')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1050
'train' 840
'validation' 210
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

parsinlu_reading_comprehension

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/parsinlu_reading_comprehension')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 518
'train' 415
'validation' 103
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

テーブル内のペンギン

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/penguins_in_a_table')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 149
'train' 120
'validation' 29
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

periodic_elements

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/periodic_elements')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 654
'train' 524
'validation' 130
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ペルシア語のイディオム

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/persian_idioms')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 66
'train' 50
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

フレーズ関連性

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/phrase_relatedness')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 100
'train' 80
'validation' 20
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

身体的直観

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/physical_intuition')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 81
'train' 65
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

物理

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/physics')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 229
'train' 184
'validation' 45
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

物理学_質問

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/physics_questions')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 54
'train' 38
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

play_dialog_同じまたは異なる

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/play_dialog_same_or_different')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 3264
'train' 2612
'validation' 652
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ポリッシュシーケンスラベリング

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/polish_sequence_labeling')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 12812
'train' 10250
'validation' 2562
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

前提条件_as_nli

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/presuppositions_as_nli')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 735
'train' 588
'validation' 147
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

qa_ウィキデータ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/qa_wikidata')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 20321
'train' 16257
'validation' 4064
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

質問の選択

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/question_selection')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1582年
'train' 1266
'validation' 316
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

本物または偽物のテキスト

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/real_or_fake_text')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 15088
'train' 12072
'validation' 3016
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

色付きオブジェクトについての推論

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/reasoning_about_colored_objects')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 2000年
'train' 1600
'validation' 400
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

リピートコピーロジック

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/repeat_copy_logic')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 32
'train' 16
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

言い換える

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/rephrase')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 78
'train' 62
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

なぞなぞ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/riddle_sense')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 49
'train' 33
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

破滅の名前

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/ruin_names')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 448
'train' 359
'validation' 89
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

顕著な翻訳エラー検出

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/salient_translation_error_detection')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 998
'train' 799
'validation' 199
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

科学的プレスリリース

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/scientific_press_release')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 50
'train' 34
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

semantic_parsing_in_context_sparc

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/semantic_parsing_in_context_sparc')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1155
'train' 924
'validation' 231
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

semantic_parsing_spider

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/semantic_parsing_spider')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1034
'train' 828
'validation' 206
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

文の曖昧さ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/sentence_ambiguity')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 60
'train' 44
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

類似点_抽象化

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/similarities_abstraction')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 76
'train' 60
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

simp_turing_concept

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/simp_turing_concept')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 6390
'train' 5112
'validation' 1278
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

simple_arithmetic_json

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/simple_arithmetic_json')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 30
'train' 14
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

simple_arithmetic_json_multiple_choice

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/simple_arithmetic_json_multiple_choice')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 8
'train' 0
'validation' 0
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

simple_arithmetic_json_subtasks

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/simple_arithmetic_json_subtasks')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 30
'train' 15
'validation' 15
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

simple_arithmetic_multiple_targets_json

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/simple_arithmetic_multiple_targets_json')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 10
'train' 0
'validation' 0
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

シンプルな倫理的質問

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/simple_ethical_questions')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 115
'train' 92
'validation' 23
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

シンプルテキスト編集

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/simple_text_editing')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 47
'train' 31
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

嫌がる

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/snarks')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 181
'train' 145
'validation' 36
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ソーシャルアイカ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/social_iqa')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1935年
'train' 1548年
'validation' 387
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ソーシャルサポート

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/social_support')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 897
'train' 718
'validation' 179
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

スポーツ理解

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/sports_understanding')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 986
'train' 789
'validation' 197
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

奇妙な物語

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/strange_stories')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 174
'train' 140
'validation' 34
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

戦略QA

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/strategyqa')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 2289
'train' 1832年
'validation' 457
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

十分な情報

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/sufficient_information')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 39
'train' 23
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

自殺の危険性

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/suicide_risk')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 40
'train' 24
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

スワヒリ語_英語_ことわざ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/swahili_english_proverbs')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 153
'train' 123
'validation' 30
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

スウェーデン語からドイツ語へのことわざ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/swedish_to_german_proverbs')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 72
'train' 56
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

シンボルの解釈

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/symbol_interpretation')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 990
'train' 795
'validation' 195
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

時間的シーケンス

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/temporal_sequences')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1000
'train' 800
'validation' 200
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

時制

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/tense')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 286
'train' 229
'validation' 57
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

タイムダイヤル

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/timedial')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 2550
'train' 2040年
'validation' 510
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

トピックチャット

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/topical_chat')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 22295
'train' 17836
'validation' 4459
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

tracking_shuffled_objects

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/tracking_shuffled_objects')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 3750
'train' 3000
'validation' 750
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

寓話の理解

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/understanding_fables')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 189
'train' 152
'validation' 37
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

undo_permutation

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/undo_permutation')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 300
'train' 240
'validation' 60
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

単位変換

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/unit_conversion')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 23936
'train' 19151
'validation' 4785
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

単位の解釈

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/unit_interpretation')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 100
'train' 80
'validation' 20
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

unnatural_in_context_learning

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/unnatural_in_context_learning')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 73420
'train' 58736
'validation' 14684
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ビタミンc_fact_verification

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/vitaminc_fact_verification')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 54668
'train' 43735
'validation' 10933
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

タオとは何か

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/what_is_the_tao')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 36
'train' 20
'validation' 16
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

どのウィキ編集

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/which_wiki_edit')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 571
'train' 457
'validation' 114
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

ウィノホワイ

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/winowhy')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 2862
'train' 2290
'validation' 572
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

単語の並べ替え

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/word_sorting')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 1900年
'train' 1520
'validation' 380
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

word_unscrambling

次のコマンドを使用して、このデータセットを TFDS にロードします。 

ds = tfds.load('huggingface:bigbench/word_unscrambling')
  • 説明： 
The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to
probe large language models, and extrapolate their future capabilities.
  • ライセンス: Apache ライセンス 2.0
  • バージョン: 0.0.0
  • 分割:
スプリット例
'default' 8917
'train' 7134
'validation' 1783年
  • 特徴： 
{
    "idx": {
        "dtype": "int32",
        "id": null,
        "_type": "Value"
    },
    "inputs": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_targets": {
        "feature": {
            "dtype": "string",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    },
    "multiple_choice_scores": {
        "feature": {
            "dtype": "int32",
            "id": null,
            "_type": "Value"
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}