import torch.nn as nn | |
from diffusers.configuration_utils import ConfigMixin, register_to_config | |
from diffusers.models.modeling_utils import ModelMixin | |
class GuiderNet(ModelMixin, ConfigMixin): | |
def __init__(self, in_channels=3, mid_channels=4, out_channels=8): | |
super().__init__() | |
self.layers = nn.Sequential( | |
nn.Conv2d(in_channels, mid_channels, 4, 2, 1), | |
nn.SiLU(), | |
nn.Conv2d(mid_channels, mid_channels, 4, 2, 1), | |
nn.SiLU(), | |
nn.Conv2d(mid_channels, out_channels, 4, 2, 1), | |
) | |
def forward(self, x): | |
return self.layers(x) | |