【Caffe】Check failed status == CUDNN_STATUS_SUCCESS (3 vs. 0) CUDNN_STATUS_BAD_PARAM

Nov 9, 2017#Caffe134

AI-generated summary

The author encountered an error while using Python to call a Caffe model. The error message states that the status is not equal to CUDNN_STATUS_SUCCESS, but instead CUDNN_STATUS_BAD_PARAM. The author searched for solutions online but did not find any. Eventually, they found a solution on a Github page, which involved adding "engine: CAFFE" to the convolution_param in the convolutional layer.

最近在搞一些事情，有些情况用 python 调用 caffe 模型时会报错（注意不是编译 caffe 时报的）

Check failed: status == CUDNN_STATUS_SUCCESS (3 vs. 0) CUDNN_STATUS_BAD_PARAM

Check failed: status == CUDNN_STATUS_SUCCESS (4 vs. 0) CUDNN_STATUS_BAD_PARAM

在网上查了很多资料，不过都没解决我的问题。

意外想到曾经在某个 Github 主页上看到过一句话。最终解决了我的问题。

就是在卷积层 convolution_param 中添加 engine: CAFFE