bmm_model AttributeError: 'int' object has no attribute 'look_lookup'
uRENu opened this issue · comments
"the input parameter bmm_model is 0" see line 125 in train.py
should function track_training_loss(line 184 in train.py) be executed before function train_mixUp_HardBootBeta(line 159 in train.py)?
Hello uRENu, thank you for your interest in our code and apologies for the late reply. You are correct, the function track_training_loss should be called before train_mixUp_HardBootBeta to create the bmm_model. This is why the function train_mixUp_HardBootBeta is called only if epoch > bootstrap_ep_mixup. In almost all the setups we have a warm-up stage where the model trains without bootstrapping to obtain better features and perform a better noise detection. Please, feel free to ask any other questions.