five

open-llm-leaderboard/details_Kunhao__pile-7b-250b-tokens

收藏
Hugging Face2023-09-17 更新2024-03-04 收录
下载链接:
https://hf-mirror.com/datasets/open-llm-leaderboard/details_Kunhao__pile-7b-250b-tokens
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 Kunhao/pile-7b-250b-tokens 在 Open LLM Leaderboard 上的评估运行期间自动创建的。数据集由 64 个配置组成,每个配置对应一个被评估的任务。数据集是从 2 次运行中生成的,每次运行在每个配置中表示为特定的拆分。train 拆分始终指向最新的结果。一个名为 results 的额外配置存储了所有运行的聚合结果,这些结果用于计算和显示 Open LLM Leaderboard 上的指标。README 还提供了如何使用 Python 中的 datasets 库加载数据集的示例,并包含了特定运行的最新结果。
提供机构:
open-llm-leaderboard
原始信息汇总

数据集概述

数据集名称

  • 名称: Evaluation run of Kunhao/pile-7b-250b-tokens

数据集组成

  • 配置数量: 64个配置
  • 每个配置对应: 一个评估任务
  • 创建来源: 从2次运行中创建
  • 分割命名: 使用运行的时间戳命名
  • 训练分割: 指向最新结果

额外配置

  • 配置名称: "results"
  • 用途: 存储所有运行结果的聚合,用于计算和显示聚合指标

数据加载示例

python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_Kunhao__pile-7b-250b-tokens", "harness_winogrande_5", split="train")

最新结果

  • 最新结果来源: 2023-09-17T06:31:55.680940
  • 结果示例: python { "all": { "em": 0.0016778523489932886, "em_stderr": 0.00041913301788268413, "f1": 0.02978817114093967, "f1_stderr": 0.0010045845151481873, "acc": 0.26666299658982046, "acc_stderr": 0.008015854967176925 }, "harness|drop|3": { "em": 0.0016778523489932886, "em_stderr": 0.00041913301788268413, "f1": 0.02978817114093967, "f1_stderr": 0.0010045845151481873 }, "harness|gsm8k|5": { "acc": 0.00530705079605762, "acc_stderr": 0.002001305720948061 }, "harness|winogrande|5": { "acc": 0.5280189423835833, "acc_stderr": 0.014030404213405788 } }

配置详情

  • 配置名称: harness_arc_challenge_25

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|arc:challenge|25_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_drop_3

    • 分割: 2023_09_17T06_31_55.680940, latest
    • 路径: **/details_harness|drop|3_2023-09-17T06-31-55.680940.parquet
  • 配置名称: harness_gsm8k_5

    • 分割: 2023_09_17T06_31_55.680940, latest
    • 路径: **/details_harness|gsm8k|5_2023-09-17T06-31-55.680940.parquet
  • 配置名称: harness_hellaswag_10

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hellaswag|10_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: 多个路径,例如 **/details_harness|hendrycksTest-abstract_algebra|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_abstract_algebra_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-abstract_algebra|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_anatomy_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-anatomy|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_astronomy_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-astronomy|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_business_ethics_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-business_ethics|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_clinical_knowledge_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-clinical_knowledge|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_college_biology_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-college_biology|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_college_chemistry_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-college_chemistry|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_college_computer_science_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-college_computer_science|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_college_mathematics_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-college_mathematics|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_college_medicine_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-college_medicine|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_college_physics_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-college_physics|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_computer_security_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-computer_security|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_conceptual_physics_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-conceptual_physics|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_econometrics_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-econometrics|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_electrical_engineering_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-electrical_engineering|5_2023-08-17T19:43:31.029227.parquet
  • 配置名称: harness_hendrycksTest_elementary_mathematics_5

    • 分割: 2023_08_17T19_43_31.029227, latest
    • 路径: **/details_harness|hendrycksTest-elementary_mathematics|5_2023-08-17T19:43:31.029227.parquet

以上是数据集的详细概述,包括数据集的名称、组成、额外配置、数据加载示例、最新结果以及各个配置的详细信息。

二维码
社区交流群
二维码
科研交流群
商业服务