def __init__(self, ascii, shapes):
super(Critic, self).__init__()
filters = 64
#self.init_shape = (filters, shapes[0][0], shapes[0][1])
#self.preprocess = nn.Sequential(
# nn.Linear(z_size, reduce(mul, self.init_shape, 1)),
# nn.ReLU(True))
self.blocks = nn.ModuleList()
in_ch = len(ascii)
out_ch = filters
for s in shapes[1:-1]:
block = nn.Sequential(utils.Resize(s),
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch), nn.LeakyReLU(True))
in_ch = out_ch
out_ch = in_ch * 2
self.blocks.append(block)
self.postprocess = nn.Sequential(utils.Resize(shapes[-1]),
nn.AdaptiveAvgPool2d(1))
self.winner = nn.Sequential(nn.Linear(in_ch, 1), nn.Sigmoid())
self.steps = nn.Sequential(nn.Linear(in_ch, 1), nn.Sigmoid())
self.compiles = nn.Sequential(nn.Linear(in_ch, 1), nn.Sigmoid())
def __init__(self, mapping, shapes, z_shape, dropout):
super(Generator, self).__init__()
self.z_size = z_shape[0]
filters = 512
self.init_shape = (filters, *shapes[0])
self.preprocess = nn.Sequential(
nn.Linear(self.z_size, reduce(mul, self.init_shape), bias=False),
nn.LeakyReLU(True))
self.blocks = nn.ModuleList()
in_ch = filters
for s in shapes[1:-1]:
out_ch = in_ch // 2
block = nn.Sequential(
utils.Resize(s),
nn.Conv2d(in_ch, out_ch, 3, padding=1, bias=False),
nn.LeakyReLU(True),
nn.Conv2d(out_ch, out_ch, 3, padding=1, bias=False),
nn.BatchNorm2d(out_ch),
nn.LeakyReLU(True),
)
in_ch = out_ch
self.blocks.append(block)
out_ch = len(mapping)
self.output = nn.Sequential(
utils.Resize(shapes[-1]),
nn.Conv2d(in_ch, out_ch, 3, padding=1, bias=True), nn.Softmax2d())
def __init__(self, num_outputs, hidden_size=512, shapes=[]):
super(Decoder, self).__init__()
in_ch = 32
self.initial_shape = (in_ch, *shapes[0])
self.features = nn.Linear(hidden_size, in_ch * reduce(mul, shapes[0]))
#[(3, 4),(6, 8),(12, 16)]
self.blocks = nn.ModuleList()
halfway = math.ceil(len(shapes) / 2)
for i, s in enumerate(shapes):
if (i == len(shapes) - 1):
out_ch = num_outputs
elif (i + 1 == halfway and len(shapes) % 2 == 0):
out_ch = in_ch
elif (i + 1 < halfway):
out_ch = in_ch * 2
else:
out_ch = in_ch // 2
block = nn.Sequential(utils.Resize(s),
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.ReLU())
in_ch = out_ch
self.blocks.append(block)