Tony Morris

interesting... we are already forced to use many tools around python to help being a bit deterministic, more robust tools would help...

Please let us know what you need. Happy to write it.

Dreaming: I want a tensorflow/pytorch python tensor compile-time dimension checker, Is that possible? :D

I don't know what that is. But let's go with yes. Want to help me out with knowing what that means?

you can read that IMHO http://nlp.seas.harvard.edu/NamedTensor  ... it's recent and shows a few issues... tensor are just matrix with dimension batch x dim1 x dim2 x ... and all matrix computations aren't possible depending on those dims... a type-dependent system could help but python is farfar

tensorflow & pytorch have a runtime compile phase that try to check dimensions as much as possible but it has limits and you have to run and there is no inference... dimensions shouldn't be just the value "5" for example but a {name:"batch" + value:5} ...

You (the programmer) assert the matrix dimension and the tools somehow try to check that it is true?

yes, it tries but for now, compile phase is limited, some dimension remain dynamic in tensorflow until you run the "fitting/training" phase (you know the batch size after loading the dataset)