simon-artzet/details_SmolLM3-SFT-GSM8K_private
收藏资源简介：
数据集是在对SmolLM3-SFT-GSM8K模型进行评估时自动创建的。该数据集包含一个配置，每个配置对应于一个评估任务。数据集基于一次运行，train 分片始终指向最新的结果。还有一个额外的results配置，用于存储运行的所有聚合结果。数据集文件为Parquet格式，可以使用HuggingFace的datasets库进行加载。最新结果包括all类别和特定的lighteval|gsm8k|0任务的extractive_match分数及其标准误差。
The dataset was automatically created during the evaluation of the SmolLM3-SFT-GSM8K model. It consists of one configuration, each corresponding to one of the evaluated tasks. The dataset is based on one run, with the train split always pointing to the latest results. There is also an additional results configuration that stores all aggregated results of the run. The dataset files are in Parquet format and can be loaded using the datasets library from HuggingFace. The latest results include an extractive_match score and its standard error for the all category and a specific lighteval|gsm8k|0 task.
