Brando Miranda

> Hi! > > We currently have not released our tensorflow checkpoints - I will update this response if this changes in the future. > In the meantime, though, we...

btw would they be compatible with pytorch?

progress on this?

@tristandeleu I get a data integrity issue...what does that mean? Do I download it again or what do you recommend? ``` raise RuntimeError('MiniImagenet integrity check failed') ```

what does "Integrity issue here indeed means a quota issue on Google Drive's end" mean? > On Oct 22, 2020, at 2:28 AM, Tristan Deleu wrote: > > Integrity issue...

what issues are you having? torche meta is just a dataloader library, it should be easy (but tedious) to do it since there is nothing that stops torchmeta from working....

@tristandeleu sorry for reviving this thread now. Is torchmeta still not compatible with DDP?

> In the (non-higher) example of MAML you are also looping through each task individually. Unfortunately this is necessary in PyTorch because each task has their own set of parameters,...

> The data-loaders should definitely work with DDP, since they are inheriting from regular PyTorch `DataLoader`. However when I said that DDP was not supported in Torchmeta at the moment,...

for myself in answering 2, I think it's unclear (my guess unlikely) that higher is compatible with DDP, see https://github.com/facebookresearch/higher/issues/116, https://github.com/facebookresearch/higher/issues/98