the batch_norm layer problem

[NanYoMy]

label-reg/labelreg/layers.py

Line 40 in 3fe086f

return tf.nn.relu(tf.contrib.layers.batch_norm(tf.nn.conv3d_transpose(input_, w, shape_out, strides, "SAME")))

Hi HU,
I notices that the is_training parameter of batch_norm layers(tf.contrib.layers.batch_norm) in your project didn't set to False during the inference period.
But according to the helper document of TensorFlow. this parameter should be set to False during the inference time.
https://www.tensorflow.org/api_docs/python/tf/contrib/layers/batch_norm

[YipengHu]

Thanks! You are absolutely right! However, in practice, it makes little difference with one single pass on the test data, so it wasn't set only for simplicity of the demo. This should be updated as you suggested.

…

On Sun, 15 Sep 2019, 05:58 NanYoMy, ***@***.***> wrote: https://github.com/YipengHu/label-reg/blob/3fe086f8bb9d029acf9fdecba3cee92009cfb3dc/labelreg/layers.py#L40 Hi HU, I notices that the *is_training* parameter of batch_norm layers(tf.contrib.layers.batch_norm) in your project didn't set to *False* during the inference period. But according to the helper document of TensorFlow. this parameter should be set to False during the inference time. https://www.tensorflow.org/api_docs/python/tf/contrib/layers/batch_norm <https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.tensorflow.org%2Fapi_docs%2Fpython%2Ftf%2Fcontrib%2Flayers%2Fbatch_norm&data=02%7C01%7C%7Ca6ad2c5722824800ae7e08d739995aa0%7C1faf88fea9984c5b93c9210a11d9a5c2%7C0%7C0%7C637041203124729950&sdata=XcEHi18RQ7bG%2Fd9k%2F%2BpZ%2F%2B1BeCvBkx78eQM%2BXkyShEE%3D&reserved=0> — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#22?email_source=notifications&email_token=AATX37LPZNNOGWU7OROK55LQJW6HLA5CNFSM4IWZGHE2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HLNGKGQ>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AATX37MRBQRFI7KTKJ52P23QJW6HLANCNFSM4IWZGHEQ> .

[NanYoMy]

really appreciated!