DLYuanGod / TinyGPT-V

TinyGPT-V: Efficient Multimodal Large Language Model via Small Backbones
BSD 3-Clause "New" or "Revised" License
1.24k stars 76 forks source link

在Stage3可以使用中文语料进行train吗? #35

Open chenzhongde opened 3 weeks ago

chenzhongde commented 3 weeks ago

非常棒的工作! 我有一个数据集是商品的图片和中文标题,任务想要用图片生成标题,不知道可不可以在第三阶段使用我的数据集进行微调。