open-llm-leaderboard/details_CHIH-HUNG__llama-2-13b-FINETUNE3_3.3w-r16-q_k_v_o_gate_up_down
收藏数据集概述
数据集来源
该数据集是在评估模型 CHIH-HUNG/llama-2-13b-FINETUNE3_3.3w-r16-q_k_v_o_gate_up_down 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。
- 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_CHIH-HUNG__llama-2-13b-FINETUNE3_3.3w-r16-q_k_v_o_gate_up_down", "harness_winogrande_5", split="train")
最新结果
以下是 2023-10-25T21:42:44.819201 运行 的最新结果： python { "all": { "em": 0.24790268456375839, "em_stderr": 0.004421985207771805, "f1": 0.2875272651006712, "f1_stderr": 0.004396913707433761, "acc": 0.4423710172435768, "acc_stderr": 0.010335262808146502 }, "harness|drop|3": { "em": 0.24790268456375839, "em_stderr": 0.004421985207771805, "f1": 0.2875272651006712, "f1_stderr": 0.004396913707433761 }, "harness|gsm8k|5": { "acc": 0.11599696739954511, "acc_stderr": 0.008820485491442497 }, "harness|winogrande|5": { "acc": 0.7687450670876085, "acc_stderr": 0.01185004012485051 } }
配置详情
-
harness_arc_challenge_25
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|arc:challenge|25_2023-10-04T05-31-22.658790.parquet
-
harness_drop_3
- 分割: 2023_10_25T21_42_44.819201, latest
- 路径:
**/details_harness|drop|3_2023-10-25T21-42-44.819201.parquet
-
harness_gsm8k_5
- 分割: 2023_10_25T21_42_44.819201, latest
- 路径:
**/details_harness|gsm8k|5_2023-10-25T21-42-44.819201.parquet
-
harness_hellaswag_10
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hellaswag|10_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径: 多个文件路径，例如
**/details_harness|hendrycksTest-abstract_algebra|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_abstract_algebra_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_anatomy_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_astronomy_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_business_ethics_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_clinical_knowledge_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_college_biology_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_college_chemistry_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_college_computer_science_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_college_mathematics_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_college_medicine_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_college_physics_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_computer_security_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_conceptual_physics_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_econometrics_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-10-04T05-31-22.658790.parquet
-
harness_hendrycksTest_electrical_engineering_5
- 分割: 2023_10_04T05_31_22.658790, latest
- 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-10-04T05-31-22.658790.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。
