avalanche
avalanche copied to clipboard
ContinualAI
Avalanche: an End-to-End Library for Continual Learning based on PyTorch.
Hello, I implemented the SIW approach from https://arxiv.org/pdf/2008.13710.pdf as a plugin. I just need you to check if my implementation respects the rules of Avalanche. I also added an example...
I noticed it is quite tricky at the moment to generate a benchmark with an unbalanced number of examples for each step. It would be nice to have an option...
Now that the VPS is online we should add automatic tests on GPU as well. We could also add more fancy tests to automatically expand supported package versions.
In order to test the support we are developing for Segmentation / Detection other than Classification and Regression tasks we should implement the split Pascal VOC2021 proposed here: [Incremental Learning...
We should support averaging results across multiple runs. This is related to #300.
### 1. Experience reference It might be useful, from inside the `before_training()` (and `before_eval()`) method of `StrategyPlugin` to have access to the current scenario. This would allow to do variable...
We should add MiniImagenet to the benchmarks. It's been used here: https://arxiv.org/abs/1902.10486
Can we set task labels to `None` or some other specific value instead of 0 for single task scenarios? Setting to None has the advantage that the evaluation and loggers...
We need to add some features statistics in tensorboard. The user should be able to specify them during the creation of the tensorboard object. For now the `TensorboardLogging `object is...
At the moment creating a custom benchmark using Avalanche (for instance when using `nc_scenario` or `ni_scenario`) means passing one or more "supported dataset"s to these generators. A supported dataset is...
Metadata
Owner
Metadata
Avalanche: an End-to-End Library for Continual Learning based on PyTorch.