open-llm-leaderboard-old/details_Lajonbot__Llama-2-7b-chat-hf-instruct-pl-lora_unload
收藏数据集概述
该数据集是在对模型 Lajonbot/Llama-2-7b-chat-hf-instruct-pl-lora_unload 进行评估运行期间自动创建的,用于 Open LLM Leaderboard。
数据集组成
- 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 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_Lajonbot__Llama-2-7b-chat-hf-instruct-pl-lora_unload", "harness_winogrande_5", split="train")
最新结果
以下是 2023-09-17T12:29:41.474491 运行的最新结果: python { "all": { "em": 0.0016778523489932886, "em_stderr": 0.0004191330178826897, "f1": 0.058242449664429756, "f1_stderr": 0.0013938395704941914, "acc": 0.39479575124777627, "acc_stderr": 0.009795911187121632 }, "harness|drop|3": { "em": 0.0016778523489932886, "em_stderr": 0.0004191330178826897, "f1": 0.058242449664429756, "f1_stderr": 0.0013938395704941914 }, "harness|gsm8k|5": { "acc": 0.06899166034874905, "acc_stderr": 0.006980995834838582 }, "harness|winogrande|5": { "acc": 0.7205998421468035, "acc_stderr": 0.01261082653940468 } }
配置详情
-
harness_arc_challenge_25
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|arc:challenge|25_2023-08-01T13:20:11.770272.parquet
-
harness_drop_3
- 分割: 2023_09_17T12_29_41.474491, latest
- 路径:
**/details_harness|drop|3_2023-09-17T12-29-41.474491.parquet
-
harness_gsm8k_5
- 分割: 2023_09_17T12_29_41.474491, latest
- 路径:
**/details_harness|gsm8k|5_2023-09-17T12-29-41.474491.parquet
-
harness_hellaswag_10
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hellaswag|10_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径: 多个路径,例如
**/details_harness|hendrycksTest-abstract_algebra|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_abstract_algebra_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_anatomy_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_astronomy_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_business_ethics_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_clinical_knowledge_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_college_biology_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_college_chemistry_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_college_computer_science_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_college_mathematics_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_college_medicine_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_college_physics_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_computer_security_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_conceptual_physics_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_econometrics_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-08-01T13:20:11.770272.parquet
-
harness_hendrycksTest_electrical_engineering_5
- 分割: 2023_08_01T13_20_11.770272, latest
- 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-08-01T13:20:11.770272.parquet
以上是数据集的详细概述,包括数据集的组成、加载示例、最新结果以及各个配置的详细信息。



