Astris/Ensemble_of_models_VQA_v0.5

该数据集包含id、image、prompt、responses和img_ptr等特征。responses是一个列表，包含model和response两个字段。数据集分为一个训练集，包含240,794个示例，总大小为416,480,434字节。数据集的下载大小为222,975,230字节。数据集的prompts和GPT-4V completions来源于liuhaotian/LLaVA-Instruct-150K，其中40%的prompts被发送到Gemini 1.5 Pro，其余的发送到Gemini 1.5 Flash。

This dataset is primarily used for image and text interaction tasks, including features such as ID, image, prompt, and responses. The responses feature is a list containing the model and response content. The dataset is divided into a training set with approximately 240,000 entries. The prompts and completions in the dataset are obtained from liuhaotian/LLaVA-Instruct-150K, with 40% of the prompts going to Gemini 1.5 Pro and the rest to Gemini 1.5 Flash.

原始信息汇总

数据集概述

语言

• 英文（en）

数据集大小

• 100K<n<1M

数据集信息

特征

• id: 字符串类型
• image: 字符串类型
• prompt: 字符串类型
• responses: 列表类型
  • model: 字符串类型
  • response: 字符串类型
• img_ptr: 整数类型（int32）

数据分割

• train:
  • 字节数: 416480434
  • 样本数: 240794

下载和数据集大小

• 下载大小: 222975230
• 数据集大小: 416480434

配置

• config_name: default
  • data_files:
    • split: train
    • path: data/train-*