open-llm-leaderboard-old/details_vicgalle__gpt2-alpaca

该数据集是在评估模型vicgalle/gpt2-alpaca时自动创建的，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，每次运行的结果存储在不同的分割中，分割名称基于运行的时间戳。train分割始终指向最新的结果。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，并用于计算和展示在Open LLM Leaderboard上的聚合指标。

原始信息汇总

数据集概述

数据集 Evaluation run of vicgalle/gpt2-alpaca 是在评估模型 vicgalle/gpt2-alpaca 在 Open LLM Leaderboard 上的自动创建的。

数据集组成

• 数据集由 64 个配置组成，每个配置对应一个评估任务。
• 数据集从 2 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。
• 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_vicgalle__gpt2-alpaca", "harness_winogrande_5", split="train")

最新结果

以下是 2023-09-22T17:31:40.228869 运行的最新结果： python { "all": { "em": 0.009542785234899329, "em_stderr": 0.0009956233793266855, "f1": 0.05457529362416121, "f1_stderr": 0.001605303697316422, "acc": 0.2533543804262036, "acc_stderr": 0.0070256103461651745 }, "harness|drop|3": { "em": 0.009542785234899329, "em_stderr": 0.0009956233793266855, "f1": 0.05457529362416121, "f1_stderr": 0.001605303697316422 }, "harness|gsm8k|5": { "acc": 0.0, "acc_stderr": 0.0 }, "harness|winogrande|5": { "acc": 0.5067087608524072, "acc_stderr": 0.014051220692330349 } }

配置详情

• harness_arc_challenge_25

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|arc:challenge|25_2023-07-19T10:35:22.548714.parquet

• harness_drop_3

  • 分片: 2023_09_22T17_31_40.228869, latest
  • 路径: **/details_harness|drop|3_2023-09-22T17-31-40.228869.parquet

• harness_gsm8k_5

  • 分片: 2023_09_22T17_31_40.228869, latest
  • 路径: **/details_harness|gsm8k|5_2023-09-22T17-31-40.228869.parquet

• harness_hellaswag_10

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hellaswag|10_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: 多个路径，例如 **/details_harness|hendrycksTest-abstract_algebra|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_abstract_algebra_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-abstract_algebra|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_anatomy_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-anatomy|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_astronomy_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-astronomy|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_business_ethics_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-business_ethics|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_clinical_knowledge_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-clinical_knowledge|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_college_biology_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-college_biology|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_college_chemistry_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-college_chemistry|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_college_computer_science_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-college_computer_science|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_college_mathematics_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-college_mathematics|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_college_medicine_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-college_medicine|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_college_physics_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-college_physics|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_computer_security_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-computer_security|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_conceptual_physics_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-conceptual_physics|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_econometrics_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-econometrics|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_electrical_engineering_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-electrical_engineering|5_2023-07-19T10:35:22.548714.parquet

• harness_hendrycksTest_elementary_mathematics_5

  • 分片: 2023_07_19T10_35_22.548714, latest
  • 路径: **/details_harness|hendrycksTest-elementary_mathematics|5_2023-07-19T10:35:22.548714.parquet

以上是数据集 Evaluation run of vicgalle/gpt2-alpaca 的详细概述，包括数据集组成、数据加载示例、最新结果以及各个配置的详细信息。