five

open-llm-leaderboard-old/details_ewqr2130__alignment-handbook-zephyr-7b-sft-full-dpo-5e7-cont2

收藏
Hugging Face2024-01-19 更新2024-06-22 收录
下载链接:
https://hf-mirror.com/datasets/open-llm-leaderboard-old/details_ewqr2130__alignment-handbook-zephyr-7b-sft-full-dpo-5e7-cont2
下载链接
链接失效反馈
官方服务:
资源简介:
该数据集是在评估模型ewqr2130/alignment-handbook-zephyr-7b-sft-full-dpo-5e7-cont2的过程中自动创建的。数据集包含了63个配置,每个配置对应一个评估任务。数据集由一次运行创建,每个运行可以在每个配置中找到,使用运行的时间戳命名。此外,还有一个名为"results"的配置,用于存储所有运行的聚合结果,并用于计算和显示在Open LLM Leaderboard上的聚合指标。

该数据集是在评估模型ewqr2130/alignment-handbook-zephyr-7b-sft-full-dpo-5e7-cont2的过程中自动创建的。数据集包含了63个配置,每个配置对应一个评估任务。数据集由一次运行创建,每个运行可以在每个配置中找到,使用运行的时间戳命名。此外,还有一个名为"results"的配置,用于存储所有运行的聚合结果,并用于计算和显示在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

该数据集是在评估模型 ewqr2130/alignment-handbook-zephyr-7b-sft-full-dpo-5e7-cont2Open LLM Leaderboard 上的运行过程中自动创建的。

数据集组成

  • 该数据集包含 63 个配置,每个配置对应一个评估任务。
  • 数据集来自 1 次运行,每个运行在每个配置中都有一个特定的分割,分割名称使用运行的时间戳。
  • "train" 分割始终指向最新的结果。
  • 一个额外的配置 "results" 存储所有运行的聚合结果,用于计算和显示 Open LLM Leaderboard 上的聚合指标。

数据加载示例

python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_ewqr2130__alignment-handbook-zephyr-7b-sft-full-dpo-5e7-cont2", "harness_winogrande_5", split="train")

最新结果

以下是 2024-01-19T21:12:05.940031 运行的最新结果

python { "all": { "acc": 0.5959098573921959, "acc_stderr": 0.03332692183681072, "acc_norm": 0.6019409558870633, "acc_norm_stderr": 0.034019817201103926, "mc1": 0.2998776009791922, "mc1_stderr": 0.016040352966713623, "mc2": 0.42358153067078547, "mc2_stderr": 0.015672254683217784 }, "harness|arc:challenge|25": { "acc": 0.5648464163822525, "acc_stderr": 0.014487986197186043, "acc_norm": 0.6032423208191127, "acc_norm_stderr": 0.014296513020180644 }, "harness|hellaswag|10": { "acc": 0.6366261700856403, "acc_stderr": 0.004799882248494813, "acc_norm": 0.8288189603664609, "acc_norm_stderr": 0.0037589728166275913 }, "harness|hendrycksTest-abstract_algebra|5": { "acc": 0.33, "acc_stderr": 0.047258156262526045, "acc_norm": 0.33, "acc_norm_stderr": 0.047258156262526045 }, ... }

配置详情

  • config_name: harness_arc_challenge_25

    • split: 2024_01_19T21_12_05.940031
      • path: **/details_harness|arc:challenge|25_2024-01-19T21-12-05.940031.parquet
    • split: latest
      • path: **/details_harness|arc:challenge|25_2024-01-19T21-12-05.940031.parquet
  • config_name: harness_gsm8k_5

    • split: 2024_01_19T21_12_05.940031
      • path: **/details_harness|gsm8k|5_2024-01-19T21-12-05.940031.parquet
    • split: latest
      • path: **/details_harness|gsm8k|5_2024-01-19T21-12-05.940031.parquet
  • config_name: harness_hellaswag_10

    • split: 2024_01_19T21_12_05.940031
      • path: **/details_harness|hellaswag|10_2024-01-19T21-12-05.940031.parquet
    • split: latest
      • path: **/details_harness|hellaswag|10_2024-01-19T21-12-05.940031.parquet
  • config_name: harness_hendrycksTest_5

    • split: 2024_01_19T21_12_05.940031
      • path:
        • **/details_harness|hendrycksTest-abstract_algebra|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-anatomy|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-astronomy|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-business_ethics|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-clinical_knowledge|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-college_biology|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-college_chemistry|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-college_computer_science|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-college_mathematics|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-college_medicine|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-college_physics|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-computer_security|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-conceptual_physics|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-econometrics|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-electrical_engineering|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-elementary_mathematics|5_2024-01-19T21-12-05.940031.parquet
        • **/details_harness|hendrycksTest-formal_logic|5_2024-01-19T21-12-05.940031.parquet }
二维码
社区交流群
二维码
科研交流群
商业服务