TanmayPani
TanmayPani
Can confirm that this exact issue also happens on Fedora 41 GNOME (Wayland)
> Can you elaborate more on what kind of use-cases would this be useful for ? trying to think when stacking different batches comes in handy? For me, the biggest...
> The learning rate depends on the batch size but since we have N batches here, how is that handled? In other words, how do you ensure the training of...
> side implementation thought: we probably not want to loop over each `self.sources` serially but do it in parallel - need to think of dataset's thread safety in that case...
@divyanshk I think I have an working implementation ready, what would be the procedure for iterating on it between us? Or do we go for the pull request directly?
Thanks @divyanshk, I forked the repo, should be done pushing the StackedBatchSampler, and associated BaseNode subclass implementations and their tests by Tuesday.