def test_cross_layer_equalization_mobilenet_v2(self):
torch.manual_seed(10)
model = MobileNetV2().to(torch.device('cpu'))
print(model)
model = model.eval()
equalize_model(model, (1, 3, 224, 224))
def cross_layer_equalization_auto():
model = models.resnet18(pretrained=True)
input_shape = (1, 3, 224, 224)
model = model.eval()
# Performs BatchNorm fold, Cross layer scaling and High bias folding
equalize_model(model, input_shape)
def visualize_changes_in_model_after_and_before_cle():
"""
Code example for visualizating model before and after Cross Layer Equalization optimization
"""
visualization_url, process = start_bokeh_server_session(8002)
model = models.resnet18(pretrained=True).to(torch.device('cpu'))
model = model.eval()
model_copy = copy.deepcopy(model)
batch_norm_fold.fold_all_batch_norms(model_copy, (1, 3, 224, 224))
equalize_model(model, (1, 3, 224, 224))
visualize_model.visualize_changes_after_optimization(
model_copy, model, visualization_url)
def test_cross_layer_equalization_mobilenet_v2_visualize_after_optimization(
self):
bokeh_visualizations_url, process = start_bokeh_server_session(8006)
torch.manual_seed(10)
model = MobileNetV2().to(torch.device('cpu'))
bokeh_session = BokehServerSession(bokeh_visualizations_url,
session_id="cle")
model = model.eval()
model_copy = copy.deepcopy(model)
# model_copy_again = copy.deepcopy(model)
batch_norm_fold.fold_all_batch_norms(model_copy, (1, 3, 224, 224))
equalize_model(model, (1, 3, 224, 224))
visualize_model.visualize_changes_after_optimization(
model_copy, model, bokeh_visualizations_url)
bokeh_session.server_session.close("test complete")
os.killpg(os.getpgid(process.pid), signal.SIGTERM)
def test_cross_layer_equalization_vgg(self):
torch.manual_seed(10)
model = models.vgg16().to(torch.device('cpu'))
model = model.eval()
equalize_model(model, (1, 3, 224, 224))
def run_pytorch_cross_layer_equalization(config, model):
cross_layer_equalization.equalize_model(model.cpu(), config.input_shape)
return model