def main():
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model = inceptionv1(pretrained=True)
model.to(device).eval()
CPPN = False
SPATIAL_DECORRELATION = True
CHANNEL_DECORRELATION = True
if CPPN:
# CPPN parameterization
param_f = lambda: param.cppn(224)
opt = lambda params: torch.optim.Adam(params, 5e-3)
# Some objectives work better with CPPN than others
obj = "mixed4d_3x3_bottleneck_pre_relu_conv:139"
else:
param_f = lambda: param.image(
224, fft=SPATIAL_DECORRELATION, decorrelate=CHANNEL_DECORRELATION)
opt = lambda params: torch.optim.Adam(params, 5e-2)
obj = "mixed4a:476"
render.render_vis(model, obj, param_f, opt)
from lucent.modelzoo import inceptionv1
import torch
import torchvision.models
from PIL import Image
from lucent.optvis import transform
from torch import nn
from torchvision.transforms import transforms
from consts import IMAGENET_MEAN
from consts import IMAGENET_STD
MODEL_LOADERS = {
"inception_v3":
lambda: torchvision.models.inception_v3(pretrained=True, aux_logits=False),
"inception_v1":
lambda: inceptionv1(pretrained=True),
}
def load_model(model_name: str) -> nn.Module:
"""
Load a pretrained torchvision model.
"""
assert model_name in MODEL_LOADERS, f"Model {model_name} cannot be loaded"
model = MODEL_LOADERS[model_name]()
model.eval()
return model
def track_activations(model: nn.Module) -> Dict[str, torch.Tensor]:
def test_inceptionv1_import_layer_repr():
model = inceptionv1()
layer_names = util.get_model_layers(model, getLayerRepr=True)
for layer_name in important_layer_names:
assert layer_names[layer_name] == 'CatLayer()'
def test_inceptionv1_graph_import():
model = inceptionv1()
layer_names = util.get_model_layers(model)
for layer_name in important_layer_names:
assert layer_name in layer_names
def inceptionv1_model():
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model = inceptionv1().to(device).eval()
return model
def inceptionv1_model():
model = inceptionv1().to(device).eval()
return model