yleo/openhermes2.5-dpo-binarized

Name: yleo/openhermes2.5-dpo-binarized
Creator: yleo
Published: 2024-02-12 22:52:55
License: 暂无描述

Hugging Face2024-02-12 更新2024-03-04 收录

下载链接：

https://hf-mirror.com/datasets/yleo/openhermes2.5-dpo-binarized

下载链接

链接失效反馈

官方服务：

资源简介：

--- dataset_info: features: - name: skip_prompt_formatting dtype: bool - name: category dtype: string - name: system_prompt dtype: 'null' - name: views dtype: 'null' - name: id dtype: 'null' - name: conversations list: - name: from dtype: string - name: value dtype: string - name: weight dtype: 'null' - name: language dtype: 'null' - name: avatarUrl dtype: 'null' - name: custom_instruction dtype: 'null' - name: source dtype: string - name: model dtype: 'null' - name: hash dtype: 'null' - name: model_name dtype: 'null' - name: topic dtype: 'null' - name: idx dtype: 'null' - name: title dtype: 'null' - name: input dtype: string - name: generation_model sequence: string - name: generation_prompt sequence: string - name: raw_generation_responses sequence: string - name: generations sequence: string - name: rating sequence: float32 - name: chosen dtype: string - name: rejected dtype: string - name: chosen_model dtype: string - name: rejected_model dtype: string - name: rejected_score dtype: float64 - name: chosen_score dtype: float64 splits: - name: train num_bytes: 930348 num_examples: 100 download_size: 593836 dataset_size: 930348 configs: - config_name: default data_files: - split: train path: data/train-* ---

数据集元信息：特征字段列表如下： 1. `skip_prompt_formatting`：布尔型数据 2. `category`：字符串型数据 3. `system_prompt`：空类型数据 4. `views`：空类型数据 5. `id`：空类型数据 6. `conversations`：复合列表类型，内部包含子字段： - `from`：字符串型数据 - `value`：字符串型数据 - `weight`：空类型数据 7. `language`：空类型数据 8. `avatarUrl`：空类型数据 9. `custom_instruction`：空类型数据 10. `source`：字符串型数据 11. `model`：空类型数据 12. `hash`：空类型数据 13. `model_name`：空类型数据 14. `topic`：空类型数据 15. `idx`：空类型数据 16. `title`：空类型数据 17. `input`：字符串型数据 18. `generation_model`：字符串序列数据 19. `generation_prompt`：字符串序列数据 20. `raw_generation_responses`：字符串序列数据 21. `generations`：字符串序列数据 22. `rating`：单精度浮点数（float32）序列数据 23. `chosen`：字符串型数据 24. `rejected`：字符串型数据 25. `chosen_model`：字符串型数据 26. `rejected_model`：字符串型数据 27. `rejected_score`：双精度浮点数（float64）数据 28. `chosen_score`：双精度浮点数（float64）数据数据划分信息： - 训练集（train）：占用字节数930348，样本总量100 全局统计参数：下载总大小为593836，数据集本体总大小为930348 数据集配置项： - 默认配置（default）：对应数据文件划分如下，训练集的文件路径为`data/train-*`

提供机构：

yleo

原始信息汇总