vincent-4/allava4v-train-regenerated

ALLaVA-4V训练数据集（重新生成版）是原始ALLaVA-4V训练数据集的重新生成版本，使用Qwen/Qwen3-VL-8B-Instruct模型进行处理。该数据集目前处于生成中，包含部分数据子集，格式为Parquet，原始JSONL格式包含252,924个样本。适用于视觉问答和图像文本到文本任务，支持英语语言。The ALLaVA-4V Train Dataset (Regenerated) is a regenerated version of the original ALLaVA-4V training dataset, processed using the Qwen/Qwen3-VL-8B-Instruct model. This dataset is currently being generated and contains a subset of the data in Parquet format, with the original JSONL format including 252,924 samples. It is suitable for visual question answering and image-text-to-text tasks, and supports the English language.