cannot copy param 0 weights from layer

最新推荐文章于 2019-12-13 20:15:45 发布

原创最新推荐文章于 2019-12-13 20:15:45 发布 · 1k 阅读

1 ·

CC 4.0 BY-SA版权

文章标签：

#caffetraining-bug

ubuntu中安装环境及问题记录专栏收录该内容

35 篇文章

订阅专栏

本文解决了一个在使用Caffe框架训练深度学习模型时遇到的参数形状不匹配问题，通过删除solverstate文件并修改模型中层的名字，成功解决了预训练网络参数与当前模型架构不一致的问题。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

I have this problem when i train my data by caffe .

F1107 12:42:42.507861  2369 net.cpp:774] Cannot copy param 0 weights from layer 'conv4_3_norm_mbox_conf'; shape mismatch.  Source param shape is 40 512 3 5 (307200); target param shape is 180 512 3 5 (1382400). To learn this layer's parameters from scratch rather than copying from a saved net, rename the layer.

这是由于预训练网络的参数与当前模型架构对不上，只要把出现错误的层名字改了就ok了！！之前识别的是两类（background和text），然后源码每个点是有20个priorbox，所以num_output是40，现在我要识别11类，输出应该是220才对。

I have solved this problem by delete the file like "VGG_text_text_polygon_precise_fix_order_384x384_iter_120000.solverstate"

Delete all solverstate file and problem have solved .

Later, I realize that the true reason is that I have altered another caffemodel to train my data , so I can solve this problem!

The latest solvement is to rename the layer's name which you add, or you can change the layers'name in "model_libs.py". It works!

But the train val is very low so I think this is a bad solvement ?Oh my god ~I don't know what I can do !