Albin Soutif

Add the possibility to pretrain on multiple tasks

3

For the moment, the nc_benchmark generator function allows for a nc_first_task option, which is good for pre-training in the class-incremental learning scenario. However, the same kind of option is not...

Opencv loads BGR images thus lowering detection accuracy

1

I have noticed that Opencv imread was loading color channels in BGR order, what I did not know before. Since I suppose the original weights are made to deal with...

Changes in Dynamic Modules

8

I tried to implement the changes that would fix bugs described in Issue #1591 and also allow users to use Dynamic modules more easily outside of avalanche strategies. To do...

IncrementalClassifiers inside MultiHeadClassifiers are adapted on all experiences

2

🐛 **Describe the bug** MultiHead classifier contains a dictionnary of Incremental classifiers, both MultiHeadClassifier and IncrementalClassifier are DynamicModules, so they get adapted through the following loop: ```{python} def avalanche_model_adaptation(model: nn.Module,...

In the eval loop, before_eval_forward is called, but in the training loop, before_forward is called. It can be confusing since it's not clear whether this callback is used in the...

Forgetting measure does not work in online scenarios

4

🐛 **Describe the bug** Forgetting measure fails to update correctly in online scenarios 🐜 **To Reproduce** Add ExperienceForgetting measure to the evaluation metrics in split_cifar10/online_replay.py, the metric does not fail...

```{python} benchmark = SplitCifar100(20, shuffle=True) benchmark = benchmark_with_validation(benchmark, val_size=0.05) print(benchmark.valid_stream[0].classes_in_this_experience) # Shows 125 samples of the same class ``` The default value to create the validation stream is shuffle=False. This...

feature request: semi-supervised, unsupervised, and novelty detection common benchmarks

1

More and more CL papers are tackling unsupervised and a mix of supervised and unsupervised learning in the different tasks. I don't think there is a clean way to address...

new_instances_benchmark problems

2

🐛 **Describe the bug** One of this problem comprises a bug, which gives the following error when calling classes_in_this_experience on an experience. The same bug already happened for class_incremental_benchmark in...

Linear Probing tools

1

I would like to add some linear probing tools since it has been a widely used technique in evaluation of continual learning and non continual learning methods. I don't know...