five

open-llm-leaderboard-old/details_dball__zephyr-tiny-dpo-qlora

收藏
Hugging Face2024-02-19 更新2024-06-22 收录
下载链接:
https://hf-mirror.com/datasets/open-llm-leaderboard-old/details_dball__zephyr-tiny-dpo-qlora
下载链接
链接失效反馈
官方服务:
资源简介:
该数据集是在模型[dball/zephyr-tiny-dpo-qlora](https://huggingface.co/dball/zephyr-tiny-dpo-qlora)在[Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)上的评估运行期间自动创建的。数据集由63个配置组成,每个配置对应一个评估任务。数据集是从1次运行中创建的,每次运行可以在每个配置中找到特定的分割,分割名称使用运行的时间戳。"train"分割始终指向最新的结果。此外,"results"配置存储了所有运行的聚合结果,并用于计算和显示[Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)上的聚合指标。

该数据集是在模型[dball/zephyr-tiny-dpo-qlora](https://huggingface.co/dball/zephyr-tiny-dpo-qlora)在[Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)上的评估运行期间自动创建的。数据集由63个配置组成,每个配置对应一个评估任务。数据集是从1次运行中创建的,每次运行可以在每个配置中找到特定的分割,分割名称使用运行的时间戳。"train"分割始终指向最新的结果。此外,"results"配置存储了所有运行的聚合结果,并用于计算和显示[Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)上的聚合指标。
提供机构:
open-llm-leaderboard-old
原始信息汇总

数据集概述

数据集摘要

该数据集是在对模型 dball/zephyr-tiny-dpo-qlora 进行评估运行期间自动创建的,用于 Open LLM Leaderboard。数据集包含 63 个配置,每个配置对应一个评估任务。数据集从 1 次运行中创建,每个运行的详细信息可以在每个配置中找到,以运行的时间戳命名的特定分割形式存储。"train" 分割始终指向最新的结果。

数据集加载

要加载特定运行的详细信息,可以使用以下代码: python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_dball__zephyr-tiny-dpo-qlora", "harness_winogrande_5", split="train")

最新结果

以下是 2024-02-19T15:34:03.345858 运行的最新结果: python { "all": { "acc": 0.2646550018920259, "acc_stderr": 0.031081365682318642, "acc_norm": 0.2657170752389051, "acc_norm_stderr": 0.03183611681877473, "mc1": 0.22399020807833536, "mc1_stderr": 0.014594964329474209, "mc2": 0.36398152821435165, "mc2_stderr": 0.013862212711505954 }, "harness|arc:challenge|25": { "acc": 0.3506825938566553, "acc_stderr": 0.013944635930726087, "acc_norm": 0.3660409556313993, "acc_norm_stderr": 0.014077223108470139 }, "harness|hellaswag|10": { "acc": 0.46176060545708025, "acc_stderr": 0.00497516738206183, "acc_norm": 0.6166102370045807, "acc_norm_stderr": 0.00485218262127425 }, "harness|hendrycksTest-abstract_algebra|5": { "acc": 0.24, "acc_stderr": 0.04292346959909284, "acc_norm": 0.24, "acc_norm_stderr": 0.04292346959909284 }, ... }

配置详情

数据集包含多个配置,每个配置对应不同的任务和数据文件。以下是部分配置的示例:

  • config_name: harness_arc_challenge_25

    • data_files:
      • split: 2024_02_19T15_34_03.345858
        • path:
          • **/details_harness|arc:challenge|25_2024-02-19T15-34-03.345858.parquet
      • split: latest
        • path:
          • **/details_harness|arc:challenge|25_2024-02-19T15-34-03.345858.parquet
  • config_name: harness_gsm8k_5

    • data_files:
      • split: 2024_02_19T15_34_03.345858
        • path:
          • **/details_harness|gsm8k|5_2024-02-19T15-34-03.345858.parquet
      • split: latest
        • path:
          • **/details_harness|gsm8k|5_2024-02-19T15-34-03.345858.parquet
  • config_name: harness_hellaswag_10

    • data_files:
      • split: 2024_02_19T15_34_03.345858
        • path:
          • **/details_harness|hellaswag|10_2024-02-19T15-34-03.345858.parquet
      • split: latest
        • path:
          • **/details_harness|hellaswag|10_2024-02-19T15-34-03.345858.parquet
  • config_name: harness_hendrycksTest_5

    • data_files:
      • split: 2024_02_19T15_34_03.345858
        • path:
          • **/details_harness|hendrycksTest-abstract_algebra|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-anatomy|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-astronomy|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-business_ethics|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-clinical_knowledge|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-college_biology|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-college_chemistry|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-college_computer_science|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-college_mathematics|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-college_medicine|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-college_physics|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-computer_security|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-conceptual_physics|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-econometrics|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-electrical_engineering|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-elementary_mathematics|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-formal_logic|5_2024-02-19T15-34-03.345858.parquet
          • **/details_harness|hendrycksTest-global_facts|5_2024-02-19T15-34-03.345858.parquet }
5,000+
优质数据集
54 个
任务类型
进入经典数据集
二维码
社区交流群

面向社区/商业的数据集话题

二维码
科研交流群

面向高校/科研机构的开源数据集话题

数据驱动未来

携手共赢发展

商业合作