sumittagadiya / image-super-resolution Goto Github PK

Hi, Sumit
thanks for your implementation!

I'm currently facing an error while attempting to instantiate the model using the conv2d_weightnorm function, as outlined in the WDSR paper. The issue arises during the initialization process.

def conv2d_weightnorm(filters, kernel_size, padding='same', activation=None, **kwargs):
return tfp.layers.weight_norm.WeightNorm(Conv2D(filters, kernel_size, padding=padding, activation=activation, **kwargs),
data_init=False)

However, I received the following error:

In an attempt to resolve this, I modified the code as follows:

def conv2d_weightnorm(filters, kernel_size, padding='same', activation=None, **kwargs):
    conv_layer = tf.keras.layers.Conv2D(filters, kernel_size, padding=padding, activation=activation, **kwargs)
    return tfp.layers.weight_norm.WeightNorm(conv_layer, data_init=False)

However, this resulted in the following error when trying to instantiate the model:

I found a workaround by using tensorflow.disable_eager_execution(), but this introduces graph-related issues that I am unsure how to resolve.

I would greatly appreciate any guidance or solutions you can provide to overcome this problem.

regards,
Gabriel