Hamed

New mixin-based template structure

7

I'm creating this draft PR for the new mixin-based template structure. For now, the main goal is to implement working version of the templates that we want and get feedback...

AvalancheConcatDataset is slow for online buffer updates

3

As pointed out in PR #1024, in online strategies buffers are updated after every (sub-)experience, therefore creating an instance of `AvalancheConcatDataset` after every experience slows down the update process. This...

Storage policy's buffer doesn't keep the targets attribute

1

All storage policies in Avalanche initialize their internal buffer with an empty dataset, i.e.: ```python self._buffer: AvalancheDataset = concat_datasets([]) ``` When a new dataset is used to update the current...

`ReplayPlugin` is slow when `num_workers` > 0

6

This is an expected issue for the `ReplayPlugin` to be slower when the number of workers is larger than zero, especially when the size of the buffer's dataloader is much...

Fix input type checking for indices in `classification_subset`

3

`classification_subset` expects a list of indices as its input, but currently it also works with a tensor of indices. However, when you want to concatenate the subset dataset with another...

I'm creating this issue to implement the Meta-BGD strategy from this paper: [Task Agnostic Continual Learning via Meta Learning](https://arxiv.org/pdf/1906.05201.pdf) It's a commonly used baseline in meta-learning-based strategies.