open-llm-leaderboard-old/details_Mikivis__gpt2-large-lora-sft1
收藏数据集概述
数据集来源
该数据集是在对模型 Mikivis/gpt2-large-lora-sft1 进行评估运行期间自动创建的,评估结果展示在 Open LLM Leaderboard 上。
数据集结构
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。
结果汇总
一个额外的配置 "results" 存储所有运行的汇总结果,用于计算和显示在 Open LLM Leaderboard 上的聚合指标。
数据加载示例
以下是加载特定运行详细信息的示例代码: python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_Mikivis__gpt2-large-lora-sft1", "harness_winogrande_5", split="train")
最新结果
以下是 2023-10-27T12:39:31.040652 运行 的最新结果: python { "all": { "em": 0.010591442953020135, "em_stderr": 0.0010483469790502319, "f1": 0.08433934563758401, "f1_stderr": 0.001890128394422962, "acc": 0.27229676400947117, "acc_stderr": 0.006998242518864891 }, "harness|drop|3": { "em": 0.010591442953020135, "em_stderr": 0.0010483469790502319, "f1": 0.08433934563758401, "f1_stderr": 0.001890128394422962 }, "harness|gsm8k|5": { "acc": 0.0, "acc_stderr": 0.0 }, "harness|winogrande|5": { "acc": 0.5445935280189423, "acc_stderr": 0.013996485037729782 } }
配置详情
数据集包含多个配置,每个配置对应不同的任务和运行结果。以下是部分配置的详细信息:
-
harness_arc_challenge_25
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|arc:challenge|25_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|arc:challenge|25_2023-09-11T17-00-06.378151.parquet
-
harness_drop_3
- 分割:2023_10_27T12_39_31.040652
- 路径:
**/details_harness|drop|3_2023-10-27T12-39-31.040652.parquet - 分割:latest
- 路径:
**/details_harness|drop|3_2023-10-27T12-39-31.040652.parquet
-
harness_gsm8k_5
- 分割:2023_10_27T12_39_31.040652
- 路径:
**/details_harness|gsm8k|5_2023-10-27T12-39-31.040652.parquet - 分割:latest
- 路径:
**/details_harness|gsm8k|5_2023-10-27T12-39-31.040652.parquet
-
harness_hellaswag_10
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hellaswag|10_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hellaswag|10_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_5
- 分割:2023_09_11T17_00_06.378151
- 路径:多个路径,如
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-11T17-00-06.378151.parquet等 - 分割:latest
- 路径:多个路径,如
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-11T17-00-06.378151.parquet等
-
harness_hendrycksTest_abstract_algebra_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_anatomy_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_astronomy_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_business_ethics_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_clinical_knowledge_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_college_biology_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_college_chemistry_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_college_computer_science_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_college_mathematics_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_college_medicine_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_college_physics_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_computer_security_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_conceptual_physics_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_econometrics_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-09-11T17-00-06.378151.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-09-11T17-00-06.378151.parquet
-
harness_hendrycksTest_electrical_engineering_5
- 分割:2023_09_11T17_00_06.378151
- 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-09-11T17-00-06.378151.parquet



