yfor

I also faced same problem,

I have solved this problem by modifying several places: 1. https://github.com/bamos/dcgan-completion.tensorflow/blob/6b0a06c300e65f209b0283ab7c076a40c42e40bb/model.py#L74 change to: ``` batch_norm(name='d_bn{}'.format(i,)) for i in range(1,4)] ``` 2. https://github.com/bamos/dcgan-completion.tensorflow/blob/6b0a06c300e65f209b0283ab7c076a40c42e40bb/model.py#L78 change to: ``` batch_norm(name='g_bn{}'.format(i,)) for i in range(log_size...

同样问题

我也遇到了这个问题, 你解决了嘛? @codeman008

@wangzilu 当时使用的tensorflow1.4.1，cuda8.0, cudnn7.0， 不知道你遇到什么错误。。

1. `from tensorflow.python.keras._impl.keras.layers.normalization import BatchNormalization` 这个里面使用的是keras的BN，好像是fuseBN，tensorRT3好像是不支持的； 2. acc下降看是否还和其他超参有关，如学习率、batchsize等；