the command : ./yolov5_det -s yolov5s.wts yolov5s.engine s, error!!!

[coding-and-coding]

Env

• GPU: RTX3090
• OS: Ubuntu 20.04
• Cuda version: 11.4
• TensorRT version: 8.5.2.2
• torch: 1.12.0

About this repo

• branch: tensorrtx(yolov5-v7.0)
• model: yolov5s, v7.0, official weight

Your problem

• my command: ./yolov5_det -s yolov5s.wts yolov5s.engine s

• output: [E] [TRT] 9: [graphShapeAnalyzer.cpp::addVolumeCheck::739] Error Code 9: Internal Error (prob: dimension 0 never exceeds -5)
  [05/24/2024-19:01:38] [E] [TRT] 4: [network.cpp::validate::2980] Error Code 4: Internal Error (Could not compute dimensions for (Unnamed Layer* 0) [Convolution]_output, because the network is not valid.)

• expect output: yolov5s.engine

[wang-xinyu]

This error is weird. Are you sure the yolov5s.wts was generated with the official yolov5s-v7.0.pt

[coding-and-coding]

您好： 还是用中文交流吧，我下载的是官方的预训练pt文件，链接是https://github.com/ultralytics/yolov5/releases/download/v7.0/yolov5s.pt ，用的tensorrtx也是对应7.0版本的，我后来用yolov5 5.0版本搭配.5.0版本的tensorrtx跑通了。这个问题我考虑是因为7.0版本yolov5模型新增了分割和分类的检测头，我感觉有可能是yolov5s.pt生成yolov5s.wts时产生的问题，即gen_wts.py的脚本可能有bug，不过因为我对yolov5 7.0的模型不是特别清楚，所以暂时没去排查 PS: 官方还发布了分类的权重文件https://github.com/ultralytics/yolov5/releases/download/v7.0/yolov5s-cls.pt和分割的权重文件 https://github.com/ultralytics/yolov5/releases/download/v7.0/yolov5s-seg.pt，但我没有用到 您有空可以分析一下。 At 2024-05-30 11:15:39, "Wang Xinyu" ***@***.***> wrote: This error is weird. Are you sure the yolov5s.wts was generated with the official yolov5s-v7.0.pt — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>