escuadrón_adversario

Referencias:

escuadrón_adversario

Utilice el siguiente comando para cargar este conjunto de datos en TFDS:

ds = tfds.load('huggingface:squad_adversarial/squad_adversarial')
  • Descripción :
Here are two different adversaries, each of which uses a different procedure to pick the sentence it adds to the paragraph:
AddSent: Generates up to five candidate adversarial sentences that don't answer the question, but have a lot of words in common with the question. Picks the one that most confuses the model.
AddOneSent: Similar to AddSent, but just picks one of the candidate sentences at random. This adversary is does not query the model in any way.
  • Licencia : Licencia MIT
  • Versión : 1.1.0
  • Divisiones :
Separar Ejemplos
'AddOneSent' 1787
'AddSent' 3560
  • Características :
{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

AñadirEnviado

Utilice el siguiente comando para cargar este conjunto de datos en TFDS:

ds = tfds.load('huggingface:squad_adversarial/AddSent')
  • Descripción :
Here are two different adversaries, each of which uses a different procedure to pick the sentence it adds to the paragraph:
AddSent: Generates up to five candidate adversarial sentences that don't answer the question, but have a lot of words in common with the question. Picks the one that most confuses the model.
AddOneSent: Similar to AddSent, but just picks one of the candidate sentences at random. This adversary is does not query the model in any way.
  • Licencia : Licencia MIT
  • Versión : 1.1.0
  • Divisiones :
Separar Ejemplos
'validation' 3560
  • Características :
{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}

AddOneSent

Utilice el siguiente comando para cargar este conjunto de datos en TFDS:

ds = tfds.load('huggingface:squad_adversarial/AddOneSent')
  • Descripción :
Here are two different adversaries, each of which uses a different procedure to pick the sentence it adds to the paragraph:
AddSent: Generates up to five candidate adversarial sentences that don't answer the question, but have a lot of words in common with the question. Picks the one that most confuses the model.
AddOneSent: Similar to AddSent, but just picks one of the candidate sentences at random. This adversary is does not query the model in any way.
  • Licencia : Licencia MIT
  • Versión : 1.1.0
  • Divisiones :
Separar Ejemplos
'validation' 1787
  • Características :
{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}