robintibor

ah amazing thanks that clears it up! Other questions, am I understading correctly there is some code that preprocesses MIMIC-CXR and that is not in this repo? Like, one cannot...

I see. One more question that came up: Did you try to handle subjects with multiple values for ethnicity in any way? For example, following code shows there are 168...

Also in all of that, just make very clear to the user whether the user is seeing a reduced, faster version of the example and what the user needed to...

So do you mean adding this as an additional option? or how do you mean solve directly in the core? Like you want to replace the current style of cropped...

Yeah I also think adding average pooling and not using cropped mode could be a simple solution for this use case, do you think this is enough info for you...

Yeah in two weeks could be a good time to discuss that

For all things cropped decoding keep in mind what is the most general case: * per-timestep targets * need to train on every target (i.e., each timestep) without gap during...

Hi @neerajwagh yes like @bruAristimunha said this is something we definitely should address. It might be possible to get it working atm by storing the npz in a way that...

Thanks for starting this! Having NMT be a subclass of TUH seems a potentially confusing decision to me. We also had some code here to load NMT: https://gist.github.com/gemeinl/e932169ed2eca5bf8b25432f0d54b095 Is there...

> > Thanks for starting this! Having NMT be a subclass of TUH seems a potentially confusing decision to me. We also had some code here to load NMT: https://gist.github.com/gemeinl/e932169ed2eca5bf8b25432f0d54b095...