search

Joker316701882 / Deep-Image-Matting

This is tensorflow implementation for paper "Deep Image Matting"
626 stars 188 forks source link

Support for Tensorflow CPU version? #15

Open kvintessents opened 6 years ago

kvintessents commented 6 years ago

tf.nn.max_pool_with_argmax is not supported by Tensoflow CPU-only version. I get this error:

InvalidArgumentError (see above for traceback): No OpKernel was registered to support Op 'MaxPoolWithArgmax' with these attrs.  Registered devices: [CPU], Registered kernels:
  <no registered kernels>

         [[Node: pool1 = MaxPoolWithArgmax[T=DT_FLOAT, Targmax=DT_INT64, ksize=[1, 2, 2, 1], padding="SAME", strides=[1, 2, 2, 1]](conv1_2)]]