NVIDIA Xavier TensorRT TensorRTX 生成engine时报错

XhranXhran

已于 2023-11-29 18:45:55 修改

阅读量5k

点赞数 6

CC 4.0 BY-SA版权

文章标签：深度学习目标检测视觉检测 linux

于 2022-04-14 17:10:44 首次发布

本文链接：https://blog.csdn.net/qq_44942565/article/details/124176642

用户在尝试将自己训练的YOLOv5s模型转换为TensorRT引擎时遇到错误。模型可以生成.wts权重文件，但在构建 TensorRT .engine 文件时发生错误，错误信息指出在某个卷积层中，权重计数不匹配。具体表现为期望的权重数量与实际提供的数量不符，可能原因是模型配置或权重加载有问题，需要调整模型的类别数量。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

自己训练得到的yolov5s无法部署。能生成.wts文件，但在生成.engine时报错，报错内容：

Loading weights: best.wts
[04/13/2022-18:13:44] [E] [TRT] (Unnamed Layer* 214) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/13/2022-18:13:44] [E] [TRT] (Unnamed Layer* 214) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
[04/13/2022-18:13:44] [E] [TRT] (Unnamed Layer* 214) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/13/2022-18:13:44] [E] [TRT] (Unnamed Layer* 214) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
[04/13/2022-18:13:44] [E] [TRT] (Unnamed Layer* 214) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/13/2022-18:13:44] [E] [TRT] (Unnamed Layer* 214) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
[04/13/2022-18:13:44] [E] [TRT] (Unnamed Layer* 214) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/13/2022-18:13:44] [E] [TRT] (Unnamed Layer* 214) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
Building engine, please wait for a while...
[04/13/2022-18:13:44] [E] [TRT] (Unnamed Layer* 214) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/13/2022-18:13:44] [E] [TRT] (Unnamed Layer* 214) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
[04/13/2022-18:13:44] [E] [TRT] Could not compute dimensions for (Unnamed Layer* 214) [Convolution]_output, because the network is not valid.
[04/13/2022-18:13:44] [E] [TRT] Network validation failed.
Build engine successfully!
yolov5: /home/xhr/tensorrtx-yolov5-v5.0/yolov5/yolov5.cpp:245: void APIToModel(unsigned int, nvinfer1::IHostMemory**, bool&, float&, float&, std::__cxx11::string&): Assertion `engine != nullptr' failed.

why？

查了一下，需要修改下类别数量。