is there a data loader for the IBM cross domain challenge? (Cross-Domain Few-Shot Learning (CD-FSL) Benchmark)

[brando90]

Hi,

I was wondering if there was a data loader using l2l ideally for the Cross-Domain Few-Shot Learning (CD-FSL) Benchmark? References:

• https://github.com/IBM/cdfsl-benchmark
• https://arxiv.org/pdf/1912.07200.pdf

ideally with an l2l model/example would be great!

[seba-1511]

Hi @brando90,

We don’t have a plan to add this dataset yet. Would you be interested in contributing it?

[brando90]

Hi @brando90,

We don’t have a plan to add this dataset yet. Would you be interested in contributing it?

it's not in my critical path at the moment, but I think in a few months it will be. I would be more than happy to contribute it once I have it ready.

[brando90]

the best way to test the implementation would be to build a testloader for it. Right?

e.g. this line

    tasksets = l2l.vision.benchmarks.get_tasksets('mini-imagenet',
                                                  train_samples=2*shots,
                                                  train_ways=ways,
                                                  test_samples=2*shots,
                                                  test_ways=ways,
                                                  root='~/data',
    )

learn2learn/examples/vision/maml_miniimagenet.py

Line 75 in 10361ce

tasksets = l2l.vision.benchmarks.get_tasksets('mini-imagenet',