Skip to content

MVbench的测试情况 #9

@ruishuzhao

Description

@ruishuzhao

作者好,我在本地复现了模型的训练过程。采用videochat2相同的训练集,并且修改了你所提到的两个数据集(videochat1 videochatgpt)的标注内容。
采用4个epoch,在mvbench上的性能大概是51.2%(开源模型本地复现性能54.85%)
基于存在差异较大,
请问,在训练过程中有什么需要注意的事项吗?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions