import torch, sys sys.path.append('.') import config from model.CMTLNet import CMTLNet model = CMTLNet(in_channels=3, num_classes=2, Arcm=0.5, multi=3).to('cpu') # Simulate a small batch to see the Fc dimension dummy_input = torch.zeros(2, 3, 64, 64, 64) # typical shape with torch.no_grad(): Fl, Fc = model.ShareTask.get_Sharefeatures(dummy_input) print(f"Fc shape: {Fc.shape}") risk_features = model.avgpool(Fc) risk_features = torch.flatten(risk_features, 1) print(f"risk_features shape after pool+flatten: {risk_features.shape}")