open-llm-leaderboard/details_Yukang__Llama-2-13b-longlora-16k-ft
收藏数据集概述
数据集来源
该数据集是在对模型 Yukang/Llama-2-13b-longlora-16k-ft 进行评估运行期间自动创建的,用于 Open LLM Leaderboard。
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" 分割总是指向最新的结果。
- 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_Yukang__Llama-2-13b-longlora-16k-ft", "harness_winogrande_5", split="train")
最新结果
以下是 最新结果来自 run 2023-10-25T13:51:53.444348 的示例: python { "all": { "em": 0.0, "em_stderr": 0.0, "f1": 0.0, "f1_stderr": 0.0, "acc": 0.2478295185477506, "acc_stderr": 0.007025978032038448 }, "harness|drop|3": { "em": 0.0, "em_stderr": 0.0, "f1": 0.0, "f1_stderr": 0.0 }, "harness|gsm8k|5": { "acc": 0.0, "acc_stderr": 0.0 }, "harness|winogrande|5": { "acc": 0.4956590370955012, "acc_stderr": 0.014051956064076896 } }
配置详情
-
harness_arc_challenge_25
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|arc:challenge|25_2023-10-10T13-32-51.379088.parquet
-
harness_drop_3
- 分割:2023_10_25T13_51_53.444348, latest
- 路径:
**/details_harness|drop|3_2023-10-25T13-51-53.444348.parquet
-
harness_gsm8k_5
- 分割:2023_10_25T13_51_53.444348, latest
- 路径:
**/details_harness|gsm8k|5_2023-10-25T13-51-53.444348.parquet
-
harness_hellaswag_10
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hellaswag|10_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:多个路径,例如
**/details_harness|hendrycksTest-abstract_algebra|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_abstract_algebra_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_anatomy_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_astronomy_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_business_ethics_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_clinical_knowledge_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_college_biology_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_college_chemistry_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_college_computer_science_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_college_mathematics_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_college_medicine_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_college_physics_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_computer_security_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_conceptual_physics_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_econometrics_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_electrical_engineering_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-10-10T13-32-51.379088.parquet
-
harness_hendrycksTest_elementary_mathematics_5
- 分割:2023_10_10T13_32_51.379088, latest
- 路径:
**/details_harness|hendrycksTest-elementary_mathematics|5_2023-10-10T13-32-51.379088.parquet
以上是数据集的概述和配置详情。



