Hi there.
I encountered mvncStatus.TIMEOUT when I'm trying to run 1d version of pix2pix model.
Following is a minimum model that reproduce the error.
input shape is [1, 1, 1024, 1]
Removing 2nd conv2d will not raise error.
Changing stride (1,1) -> (2,2) will not raise error too.
Would anyone tell me how to solve this problem?
def create_generator(input):
with tf.variable_scope("encoder"):
ini = tf.random_normal_initializer(0, 0.02)
out = tf.layers.conv2d(input, 2, kernel_size=[1,4], strides=(1, 2), padding="same", kernel_initializer=ini)
out = tf.layers.conv2d(out, 4, kernel_size=[1,4], strides=(1, 2), padding="same", kernel_initializer=ini)
return out
I put complete code of minimum reproduction here
https://github.com/hamling-ling/MovidiusTimeout
and this is how to reproduce
git clone https://github.com/hamling-ling/MovidiusTimeout
cd MovidiusTimeout
./repro.sh
