def demo():
"""
Build a network and show its backprop graph.
"""
from plugins.backprop_visualizer import make_dot_backprop
from plugins.genome_visualizer import make_dot_genome
genome = [
[[1], [0, 0], [1, 0, 0], [1]],
[ # Phase will be ignored, there are no active connections (residual is not counted as active).
[0], [0, 0], [0, 0, 0], [1]
],
[[1], [0, 0], [0, 0, 0], [0, 0, 0, 0], [1]]
]
channels = [(3, 8), (8, 8), (8, 8)]
data = torch.randn(16, 3, 32, 32)
chopped = [gene[:-1] for gene in genome]
make_dot_genome(chopped).view()
model = ResidualGenomeDecoder(genome, channels, repeats=[1, 2,
3]).get_model()
out = model(torch.autograd.Variable(data))
make_dot_backprop(out).view()
def demo():
from plugins.backprop_visualizer import make_dot_backprop
genome = [[[1], [0, 0], [1, 1, 1], [1], [2]],
[[0], [0, 0], [0, 0, 0], [1], [2]],
[[1], [0, 0], [1, 0, 1], [1, 1, 0, 1], [0], [2]]]
channels = [(3, 8), (8, 16), (16, 32)]
data = torch.randn(16, 3, 32, 32)
model = VariableGenomeDecoder(genome, channels).get_model()
out = model(torch.autograd.Variable(data))
make_dot_backprop(out).view()
def demo():
from plugins.backprop_visualizer import make_dot_backprop
genome = [
[[1], [0, 0], [1, 1, 1], [1]],
[ # Phase will be ignored, there are no active connections (residual is not counted as active).
[0], [0, 0], [0, 0, 0], [1]
],
[[1], [0, 0], [1, 1, 1], [0, 0, 1, 0], [1]]
]
channels = [(3, 8), (8, 8), (8, 8)]
data = torch.randn(16, 3, 32, 32)
model = ResidualGenomeDecoder(genome, channels).get_model()
out = model(torch.autograd.Variable(data))
make_dot_backprop(out).view()
def demo():
"""
Build a network and show its backprop graph.
"""
from plugins.backprop_visualizer import make_dot_backprop
from plugins.genome_visualizer import make_dot_genome
genome = [[[1], [0, 1], [0, 1, 0], [1, 1, 0, 1], [1, 0, 1, 0, 0], [0]],
[[0], [0, 1], [0, 0, 1], [0, 1, 0, 1], [1, 0, 1, 1, 1], [0]],
[[1], [0, 0], [0, 1, 0], [0, 0, 1, 1], [1, 1, 0, 1, 1], [1]]]
channels = [(3, 32), (32, 32), (32, 32)]
data = torch.randn(16, 3, 32, 32)
make_dot_genome(genome).view()
model = DenseGenomeDecoder(genome, channels).get_model()
out = model(torch.autograd.Variable(data))
make_dot_backprop(out).view()