Tristan Deleu

This would be a great contribution indeed! You can have a look at the [`gradient_update_parameters`](https://github.com/tristandeleu/pytorch-meta/blob/9d34b837440df4e3790fc2faaffaacfa257f4223/torchmeta/utils/gradient_based.py) for inspiration. Having a general transform sounds like an excellent idea, I would also suggest...

Hi, sorry for the late reply. There is no official guide on how to use Torchmeta with your own dataset. However, I suggest you to look at [this comment](https://github.com/tristandeleu/pytorch-meta/issues/113#issuecomment-773972272) and...

Profiling the code of Torchmeta is a great idea! I don't have much experience with this (other than using [pyinstrument](https://github.com/joerick/pyinstrument) occasionally), so any input on that is welcome!

That would be a great addition! I am not as familiar with the standard few-shot object detection in the literature, so any help is welcome.

Most datasets currently available in Torchmeta rely on a hierarchy of three objects: - [`Dataset`](https://github.com/tristandeleu/pytorch-meta/blob/8e739fa7b2d572c2044ca63215a99d0d3315bce5/torchmeta/utils/data/task.py#L8), which is simply a PyTorch dataset, which is responsible for getting the individual examples for...

Documentation is definitely incomplete, and PRs to add more documentation are welcome! > For example, usually when creating a meta-set one has a number of data-sets. How many data-sets are...

Any contribution is welcome, thank you for your help! Having a general definition for these terms would be great indeed, and the docs could use some tutorials to introduce some...

Once again apologies for the late reply. You are absolutely right, having these datasets would be great, and they are really missing at the moment (especially Meta-Dataset). I have looked...

Speaking of Meta-Dataset in particular, I have heard that the team at University of Freiburg might have a PyTorch version of Meta-Dataset, which could be a good starting point.

Having the same data when sampling a task multiple times, even in regression problems, is a design choice in Torchmeta for reproducibility which I explained in #69, so that is...