遇见数据集

open-llm-leaderboard/details_CHIH-HUNG__llama-2-13b-huangyt_FINETUNE2_3w

收藏
Hugging Face2023-10-16 更新2024-03-04 收录
官方服务:

资源简介:

该数据集是在Open LLM Leaderboard上对模型CHIH-HUNG/llama-2-13b-huangyt_FINETUNE2_3w进行评估时自动生成的。数据集包含64个配置,每个配置对应一个评估任务。数据集由2次运行生成,每次运行在每个配置中表示为特定的分割,分割名称由运行的时间戳命名。train分割始终指向最新的结果。此外,还有一个名为results的配置,存储了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。README还提供了如何使用Python代码加载运行细节的示例,并包含了特定运行的最新结果。

This dataset was automatically generated during the evaluation of the model CHIH-HUNG/llama-2-13b-huangyt_FINETUNE2_3w on the Open LLM Leaderboard. It consists of 64 configurations, each corresponding to a single evaluation task. The dataset is generated from two separate runs: for each configuration, each run is represented as a dedicated split, with the split name derived from the timestamp of the corresponding run. The `train` split always points to the most recent evaluation results. Additionally, there is a configuration named `results` that stores the aggregated results across all runs, which is utilized to compute and display the aggregated metrics on the Open LLM Leaderboard. The README also includes example Python code for loading run-specific details, alongside the most recent results for particular runs.

提供机构:
open-llm-leaderboard
原始信息汇总

数据集概述

数据集来源

该数据集是在评估模型 CHIH-HUNG/llama-2-13b-huangyt_FINETUNE2_3w 在 Open LLM Leaderboard 上的运行过程中自动创建的。

数据集结构

  • 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。
  • 一个额外的配置 "results" 存储所有运行的聚合结果,用于计算和显示在 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_CHIH-HUNG__llama-2-13b-huangyt_FINETUNE2_3w", "harness_winogrande_5", split="train")

最新结果

以下是 2023-10-16T18:08:49.562791 运行的最新结果: python { "all": { "em": 0.19316275167785235, "em_stderr": 0.004042912227684817, "f1": 0.2405106963087243, "f1_stderr": 0.004012764038516629, "acc": 0.44387175953656505, "acc_stderr": 0.010404181547690496 }, "harness|drop|3": { "em": 0.19316275167785235, "em_stderr": 0.004042912227684817, "f1": 0.2405106963087243, "f1_stderr": 0.004012764038516629 }, "harness|gsm8k|5": { "acc": 0.1197877179681577, "acc_stderr": 0.008944213403553045 }, "harness|winogrande|5": { "acc": 0.7679558011049724, "acc_stderr": 0.011864149691827948 } }

配置详情

  • harness_arc_challenge_25

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|arc:challenge|25_2023-09-01T02:00:37.235761.parquet
  • harness_drop_3

    • 分割:2023_10_16T18_08_49.562791, latest
    • 路径:**/details_harness|drop|3_2023-10-16T18-08-49.562791.parquet
  • harness_gsm8k_5

    • 分割:2023_10_16T18_08_49.562791, latest
    • 路径:**/details_harness|gsm8k|5_2023-10-16T18-08-49.562791.parquet
  • harness_hellaswag_10

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hellaswag|10_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:多个路径,例如 **/details_harness|hendrycksTest-abstract_algebra|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_abstract_algebra_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_anatomy_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-anatomy|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_astronomy_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-astronomy|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_business_ethics_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-business_ethics|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_clinical_knowledge_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-clinical_knowledge|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_college_biology_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-college_biology|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_college_chemistry_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-college_chemistry|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_college_computer_science_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-college_computer_science|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_college_mathematics_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-college_mathematics|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_college_medicine_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-college_medicine|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_college_physics_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-college_physics|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_computer_security_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-computer_security|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_conceptual_physics_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-conceptual_physics|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_econometrics_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-econometrics|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_electrical_engineering_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-electrical_engineering|5_2023-09-01T02:00:37.235761.parquet
  • harness_hendrycksTest_elementary_mathematics_5

    • 分割:2023_09_01T02_00_37.235761, latest
    • 路径:**/details_harness|hendrycksTest-elementary_mathematics|5_2023-09-01T02:00:37.235761.parquet

以上是数据集的详细概述,包括数据集的来源、结构、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