five

open-llm-leaderboard-old/details_dvruette__oasst-pythia-12b-6000-steps

收藏
Hugging Face2023-10-19 更新2024-06-22 收录
下载链接:
https://hf-mirror.com/datasets/open-llm-leaderboard-old/details_dvruette__oasst-pythia-12b-6000-steps
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[dvruette/oasst-pythia-12b-6000-steps](https://huggingface.co/dvruette/oasst-pythia-12b-6000-steps)的评估运行中自动创建的。数据集由64个配置组成,每个配置对应一个评估任务。数据集由2次运行创建,每次运行都可以在特定配置中找到,分割名称使用运行的时间戳。"train"分割始终指向最新结果。此外,"results"配置存储了所有运行的聚合结果,并用于计算和显示在[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[dvruette/oasst-pythia-12b-6000-steps](https://huggingface.co/dvruette/oasst-pythia-12b-6000-steps)的评估运行中自动创建的。数据集由64个配置组成,每个配置对应一个评估任务。数据集由2次运行创建,每次运行都可以在特定配置中找到,分割名称使用运行的时间戳。"train"分割始终指向最新结果。此外,"results"配置存储了所有运行的聚合结果,并用于计算和显示在[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 dvruette/oasst-pythia-12b-6000-steps 进行评估运行期间自动创建的。数据集包含64个配置,每个配置对应一个评估任务。数据集从2次运行中创建,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。

数据集结构

  • 配置数量: 64
  • 运行次数: 2
  • 分割类型: 每个配置包含特定时间戳的分割和 "latest" 分割

数据加载示例

python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_dvruette__oasst-pythia-12b-6000-steps", "harness_winogrande_5", split="train")

最新结果

以下是 2023-10-19T11:18:23.511791 运行 的最新结果: python { "all": { "em": 0.0030411073825503355, "em_stderr": 0.0005638896908753164, "f1": 0.05902055369127535, "f1_stderr": 0.001454819126381737, "acc": 0.31875455144380627, "acc_stderr": 0.00777684564039083 }, "harness|drop|3": { "em": 0.0030411073825503355, "em_stderr": 0.0005638896908753164, "f1": 0.05902055369127535, "f1_stderr": 0.001454819126381737 }, "harness|gsm8k|5": { "acc": 0.00530705079605762, "acc_stderr": 0.0020013057209480696 }, "harness|winogrande|5": { "acc": 0.6322020520915549, "acc_stderr": 0.013552385559833591 } }

配置详情

以下是部分配置的详细信息:

  • harness_arc_challenge_25

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|arc:challenge|25_2023-08-22T17:23:24.296836.parquet
  • harness_drop_3

    • 分割: 2023_10_19T11_18_23.511791, latest
    • 路径: **/details_harness|drop|3_2023-10-19T11-18-23.511791.parquet
  • harness_gsm8k_5

    • 分割: 2023_10_19T11_18_23.511791, latest
    • 路径: **/details_harness|gsm8k|5_2023-10-19T11-18-23.511791.parquet
  • harness_hellaswag_10

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hellaswag|10_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: 多个路径,例如 **/details_harness|hendrycksTest-abstract_algebra|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_abstract_algebra_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-abstract_algebra|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_anatomy_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-anatomy|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_astronomy_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-astronomy|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_business_ethics_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-business_ethics|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_clinical_knowledge_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-clinical_knowledge|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_college_biology_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-college_biology|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_college_chemistry_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-college_chemistry|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_college_computer_science_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-college_computer_science|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_college_mathematics_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-college_mathematics|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_college_medicine_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-college_medicine|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_college_physics_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-college_physics|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_computer_security_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-computer_security|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_conceptual_physics_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-conceptual_physics|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_econometrics_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-econometrics|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_electrical_engineering_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-electrical_engineering|5_2023-08-22T17:23:24.296836.parquet
  • harness_hendrycksTest_elementary_mathematics_5

    • 分割: 2023_08_22T17_23_24.296836, latest
    • 路径: **/details_harness|hendrycksTest-elementary_mathematics|5_2023-08-22T17:23:24.296836.parquet

以上是数据集的概述和部分配置的详细信息。

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