参考：
dialogues
使用以下命令在 TFDS 中加载此数据集：
ds = tfds.load('huggingface:deal_or_no_dialog/dialogues')
- 说明：
A large dataset of human-human negotiations on a multi-issue bargaining task, where agents who cannot observe each other’s reward functions must reach anagreement (o a deal) via natural language dialogue.
- 许可：该项目在 CC-by-NC 下许可
- 版本：1.0.0
- 拆分：
| 拆分 | 样本 |
|---|---|
'test' |
1052 |
'train' |
10095 |
'validation' |
1087 |
- 特征：
{
"input": {
"feature": {
"count": {
"dtype": "int32",
"id": null,
"_type": "Value"
},
"value": {
"dtype": "int32",
"id": null,
"_type": "Value"
}
},
"length": -1,
"id": null,
"_type": "Sequence"
},
"dialogue": {
"dtype": "string",
"id": null,
"_type": "Value"
},
"output": {
"dtype": "string",
"id": null,
"_type": "Value"
},
"partner_input": {
"feature": {
"count": {
"dtype": "int32",
"id": null,
"_type": "Value"
},
"value": {
"dtype": "int32",
"id": null,
"_type": "Value"
}
},
"length": -1,
"id": null,
"_type": "Sequence"
}
}
self_play
使用以下命令在 TFDS 中加载此数据集：
ds = tfds.load('huggingface:deal_or_no_dialog/self_play')
- 说明：
A large dataset of human-human negotiations on a multi-issue bargaining task, where agents who cannot observe each other’s reward functions must reach anagreement (o a deal) via natural language dialogue.
- 许可：该项目在 CC-by-NC 下许可
- 版本：1.0.0
- 拆分：
| 拆分 | 样本 |
|---|---|
'train' |
8172 |
- 特征：
{
"input": {
"feature": {
"count": {
"dtype": "int32",
"id": null,
"_type": "Value"
},
"value": {
"dtype": "int32",
"id": null,
"_type": "Value"
}
},
"length": -1,
"id": null,
"_type": "Sequence"
}
}