Checks if eval() is called before validating or testing a model. Some layers behave differently during training and evaluation.
1def pytorch_miss_call_to_eval_noncompliant(model): 2 import torch 3 # Noncompliant: miss call to `eval()` after load. 4 model.load_state_dict(torch.load("model.pth"))
1def pytorch_miss_call_to_eval_compliant(model): 2 model.load_state_dict(torch.load("model.pth")) 3 # Compliant: `eval()` is called after load. 4 model.eval()