gan例子infer错误
Created by: githubusr1
我使用https://github.com/PaddlePaddle/models/tree/develop/PaddleCV/PaddleGAN,里的infer.py和https://paddle-gan-models.bj.bcebos.com/stargan_G.tar.gz的预训练模型去预测老是出错: C++ Callstacks: Enforce failed. Expected in_dims[1] == filter_dims[1] * groups, but received in_ dims[1]:16 != filter_dims[1] * groups:8. The number of input channels should be equal to filter channels * groups. at [D: \1.4.1\paddle\paddle\fluid\operators\conv_op.cc:64] PaddlePaddle Call Stacks: Windows not support stack backtrace yet.
详细输出信息见附件output.txt文件。文件基本没改。我调试的一点看法是, exe.run初始化能print看到里网络输出了,(1,3,128,128)数据,表示网络没问题,feed数据后就exe.run就报错,判断是输入不对,检查输入input:(1,3,128,128), label trg (1,13)和data定义一样,但就是报错,没办法了,请帮助。infer.py,data_reaer.py有小小改动,需要发给你,基本和没改一样。 环境信息: python:3.7.1 paddle: 1.4.1 os: win7 64bit gpu: no output.txt