Shell Hu
I confirm the same issue of NAN.
Sorry for the super late reply! In case it's still relevant, you may want to print `config` at `utils/config.py:53` to see if it's a dict. It's probably easydict had a...
Hi @cszer, thanks for your questions! 1. bs=1 means we only have 1 episode at a time, which contains a support set of images and a query set of images....
That sounds a good fix. Could you send a pull request so that I can merge your code? Cheers!
@codeshop715 Unfortunately the ViT models were not optimized and it requires a 48G GPU for training Meta-Dataset. There is a trick to reduce memory is stopping grad on the ViT...
Thank you for your interests! > Can I use ProtoNet_Finetune during meta-train step? If so, how to modify the forward function of ProtoNet_Finetune? In theory, yes, but not for our...
Hello, Thank you for your interests! And hugely regret for the late reply! > the paper mentions 600 episodes for the test set and 2000 episodes for the training set...
Thanks for your interests! If still relevant: It looks like [engine.py#L120](https://github.com/hushell/pmf_cvpr22/blob/main/engine.py#L120) was not executed, otherwise `metric_logger` should have a `loss` attribute. You may want to set up a breakpoint at...
Hi @ajkailash, I would recommend to deploy your SSL model directly on the target tasks without meta-training (but fine-tuning on the support set would be necessary to achieve good performance,...
Thanks for your interests! I'll need to figure out the internal code permission for the checkpoint release. Stay tuned :)