Skip to content

Qwen2-VL视频数据训练报错:cannot reshape array of size xxx into shape (xx,2,3,9,2,14,7,2,14) #5712

@yunkchen

Description

@yunkchen

Reminder

  • I have read the README and searched the existing issues.

System Info

  • llamafactory version: 0.9.1.dev0
  • Platform: Linux-4.19.91-009.ali4000.alios7.x86_64-x86_64-with-glibc2.35
  • Python version: 3.10.12
  • PyTorch version: 2.3.0a0+ebedce2
  • Transformers version: 4.45.2
  • Datasets version: 2.21.0
  • Accelerate version: 0.34.2
  • PEFT version: 0.12.0
  • TRL version: 0.9.6
  • DeepSpeed version: 0.15.1

Reproduction

https://pai-aigc-photog-wlcb.oss-cn-wulanchabu.aliyuncs.com/TDS1M/video/335946748474.mp4

可用来复现的其中一个视频

Expected behavior

部分视频编码方式或最后帧出错问题,会导致total_frame与遍历的总数差一帧;
建议换成torchvision来解码;
此外希望能增加数据出错的try逻辑,来避免一些数据下载失败导致的训练中断。

Others

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    solvedThis problem has been already solved

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions