Load openvino for image segmentation

Hello everyone,

I am currently trying to speed up the performance of my pytorch model. This is what I did to load my pth file.

DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
self.model = UNET(in_channels=1, out_channels=1).to(DEVICE)
load_checkpoint(torch.load(config["load_checkpoint"], map_location=DEVICE), self.model)
 model.eval()
        img_tensor = tfms(img).unsqueeze(0)
        with torch.no_grad():
            preds = torch.sigmoid(model(img_tensor))
            preds = (preds > 0.5).float()
            tensor_to_pil = transforms.ToPILImage()(preds.squeeze_(0))


I read in some articles, we can speed up the performance by using openvino.
Here I tried

ie = IECore()
# Loading IR files
net = IENetwork(model="Unet.xml", weights=Unet.bin")
# Loading the network to the inference engine
self.mod = ie.load_network(network=net, device_name="CPU")
img_tensor = tfms(img).unsqueeze(0)
with torch.no_grad():
            preds = torch.sigmoid(mod(img_tensor))
            preds = (preds > 0.5).float()
            tensor_to_pil = transforms.ToPILImage()(preds.squeeze_(0))

and I got an error

TypeError: 'openvino.inference_engine.ie_api.ExecutableNetwork' object is not callable

how can I solve this?
BR,
Thalia