MemoryError

[golmo]

Hello.

When I try to plot MNIST images across layers, only the two first (conv2d_4 and conv2d_5) are output ok, my CNN model has:

conv2d_7
conv2d_8 <<< last layer with OK output
flatten_4 <<< MemoryError
dense_7
dense_8

MemoryError Traceback (most recent call last)
in
9 size = layer_activation.shape[1] #The feature map has shape (1, size, size, n_features).
10 n_cols = n_features // images_per_row # Tiles the activation channels in this matrix
---> 11 display_grid = np.zeros((size * n_cols, images_per_row * size))
12 for col in range(n_cols): # Tiles each filter into a big horizontal grid
13 for row in range(images_per_row):

Best regards.

Gregorio Del Olmo.

[VitalyArtemiev]

conv2d_7
conv2d_8 <<< last layer with OK output
flatten_4 <<< MemoryError

This is for visualising convolution filters from conv2d layers. You are trying to apply it to a dense layer. The code is trying to allocate a matrix to display the output but is overwhelmed because the dimensions that are passed to it are nonsensical.

do something like:

if "conv" in layer_name:
   display_grid = np.zeros((size * n_cols, images_per_row * size))
   ...