Not able to use vgg_like_recipe, alexnet_recipe and default cnn recipe
When I am using the vgg_like_recipe for training the model I am getting an error like this
ValueError: Negative dimension size caused by subtracting 3 from 2 for 'cnn_arch/convolutional_block/conv2d_10/Conv2D' (op: 'Conv2D') with input shapes: [10,2,3,512], [3,3,512,512].
Here the batch size is 10.This is what my train data looks like
/wav/data (Table(1664,), fletcher32, shuffle, zlib(1)) '' description := { "data": Float32Col(shape=(94, 2433), dflt=0.0, pos=0), "filename": StringCol(itemsize=100, shape=(), dflt=b'', pos=1), "id": UInt32Col(shape=(), dflt=0, pos=2), "label": UInt8Col(shape=(), dflt=0, pos=3), "offset": Float64Col(shape=(), dflt=0.0, pos=4)} byteorder := 'little' chunkshape := (1,)
The model trains well on the default ketos resnet recipe but I am unable to run it on vgg_like_recipe or alexnet_recipe.