elichen-skymizer/coco_qwen2_vl_7b_ins_ppl_dataset-1500-verbose

这个数据集包含了问题、答案和相关的图像信息。每个数据点都有一个唯一的question_id和id。数据集中的图像可以通过image字段访问，问题通过question字段访问，答案则是一个字符串列表。此外，数据集还包含了处理后的输入ID列表input_ids，输入令牌长度input_tokens_len，生成的文本generated_texts，生成的令牌长度generated_tokens_len等信息。数据集分为验证集，共有1500个示例。

The dataset contains questions, answers, and associated image information. Each data point has a unique question_id and id. Images in the dataset can be accessed through the image field, questions through the question field, and answers are provided as a list of strings. Additionally, the dataset includes processed input ID lists (input_ids), input token lengths (input_tokens_len), generated texts (generated_texts), and generated token lengths (generated_tokens_len). The dataset is split into a validation set with a total of 1500 examples.