In PyTorchTrainer class, inside fit() function within the epochs loop when a DataLoader is used, the model is set to self.model.eval() mode while printing the loss, but it is not set back to self.model.train().
A simple fix would be to add self.model.train() right at the start of the epoch loop.
In PyTorchTrainer class, inside
fit()function within the epochs loop when aDataLoaderis used, the model is set toself.model.eval()mode while printing the loss, but it is not set back toself.model.train().A simple fix would be to add
self.model.train()right at the start of the epoch loop.