five

open-llm-leaderboard-old/details_zyh3826__20231206094523-pretrain-Llama-2-13b-hf-76000

收藏
Hugging Face2023-12-16 更新2024-06-22 收录
下载链接:
https://hf-mirror.com/datasets/open-llm-leaderboard-old/details_zyh3826__20231206094523-pretrain-Llama-2-13b-hf-76000
下载链接
链接失效反馈
官方服务:
资源简介:
该数据集是在Open LLM Leaderboard上对模型zyh3826/20231206094523-pretrain-Llama-2-13b-hf-76000进行评估时自动创建的。数据集由63个配置组成,每个配置对应一个评估任务。数据集从1次运行中创建,每次运行可以在每个配置中找到,运行的时间戳作为分割名称。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,用于计算和显示在Open LLM Leaderboard上的聚合指标。

该数据集是在Open LLM Leaderboard上对模型zyh3826/20231206094523-pretrain-Llama-2-13b-hf-76000进行评估时自动创建的。数据集由63个配置组成,每个配置对应一个评估任务。数据集从1次运行中创建,每次运行可以在每个配置中找到,运行的时间戳作为分割名称。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,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

该数据集是在对模型 zyh3826/20231206094523-pretrain-Llama-2-13b-hf-76000 进行评估运行期间自动创建的,用于 Open LLM Leaderboard

数据集组成

  • 数据集由 63 个配置组成,每个配置对应一个评估任务。
  • 数据集从 1 次运行中创建,每个运行可以在每个配置中作为一个特定的分片找到,分片名称使用运行的时间戳。
  • "train" 分片始终指向最新的结果。
  • 额外的 "results" 配置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_zyh3826__20231206094523-pretrain-Llama-2-13b-hf-76000", "harness_winogrande_5", split="train")

最新结果

这些是最新结果,来自运行 2023-12-16T19:10:08.159006

python { "all": { "acc": 0.24943893194371924, "acc_stderr": 0.030400489062706072, "acc_norm": 0.25014496177092693, "acc_norm_stderr": 0.031209015064341802, "mc1": 0.25458996328029376, "mc1_stderr": 0.015250117079156482, "mc2": 0.4471244819837127, "mc2_stderr": 0.014622242508536614 }, "harness|arc:challenge|25": { "acc": 0.27303754266211605, "acc_stderr": 0.01301933276263575, "acc_norm": 0.310580204778157, "acc_norm_stderr": 0.013522292098053055 }, "harness|hellaswag|10": { "acc": 0.4026090420235013, "acc_stderr": 0.0048942100113032235, "acc_norm": 0.5203146783509262, "acc_norm_stderr": 0.0049856612829985835 }, "harness|hendrycksTest-abstract_algebra|5": { "acc": 0.22, "acc_stderr": 0.04163331998932269, "acc_norm": 0.22, "acc_norm_stderr": 0.04163331998932269 }, ... }

配置详情

  • config_name: harness_arc_challenge_25

    • data_files:
      • split: 2023_12_16T19_10_08.159006
        • **/details_harness|arc:challenge|25_2023-12-16T19-10-08.159006.parquet
      • split: latest
        • **/details_harness|arc:challenge|25_2023-12-16T19-10-08.159006.parquet
  • config_name: harness_gsm8k_5

    • data_files:
      • split: 2023_12_16T19_10_08.159006
        • **/details_harness|gsm8k|5_2023-12-16T19-10-08.159006.parquet
      • split: latest
        • **/details_harness|gsm8k|5_2023-12-16T19-10-08.159006.parquet
  • config_name: harness_hellaswag_10

    • data_files:
      • split: 2023_12_16T19_10_08.159006
        • **/details_harness|hellaswag|10_2023-12-16T19-10-08.159006.parquet
      • split: latest
        • **/details_harness|hellaswag|10_2023-12-16T19-10-08.159006.parquet
  • config_name: harness_hendrycksTest_5

    • data_files:
      • split: 2023_12_16T19_10_08.159006
        • **/details_harness|hendrycksTest-abstract_algebra|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-anatomy|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-astronomy|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-business_ethics|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-clinical_knowledge|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-college_biology|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-college_chemistry|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-college_computer_science|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-college_mathematics|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-college_medicine|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-college_physics|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-computer_security|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-conceptual_physics|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-econometrics|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-electrical_engineering|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-elementary_mathematics|5_2023-12-16T19-10-08.159006.parquet
        • **/details_harness|hendrycksTest-formal_logic|5_2023-12-16T19-10-08.159006.parquet
        • ...
二维码
社区交流群
二维码
科研交流群
商业服务