Antonio Carta
Antonio Carta
Hi @radandreicristian, this is an important issue. We do some research on text datasets at UNIPI, so we are definitely interested in its support. Currently: - I'm working on refactoring...
You have to adapt the architecture before using `load_state_dict`. Here is how I do it: ``` def load_mt_model(model_fname, stream): model = ResNet18(100) model = as_multitask(model, 'linear') for e in list(stream):...
New PR at #1142
@lrzpellegrini can I merge the PR?
you are correct, this is a bug
Can you explain what is wrong here? Given a fixed memory size, each time that you add new samples you also have to resize the buffer. If your memory is...
I understands now. You are correct and we can probably avoid this and do the resize only once (call 3). It shouldn't make a large difference in runtime but it's...
I would prefer to keep the `training` module (and corresponding tag) for CL **algorithms**. Model selection techniques and evaluation methods (such as averaging multiple results) should belong to the `evaluation`...
> It is happening because in TaskBalancedDataLoader uses the task_set and not the adaped_dataset (data). It seems to be different. This is the problem. Ideally, the `task_set` should keep the...
> @AntonioCarta we should consider adding aliases for the legacy names so that they map to the new classes. Something like BaseStrategy -> SupervisedTemplate, StrategyPlugin -> SupervisedPlugin. The problem is...