A version of this is extremely useful in Deep Learning code, where a lot of time is spent trying to guess the shape and dims of tensor variables. To help with t

A version of this is extremely useful in Deep Learning code, where a lot of time is spent trying to guess the shape and dims of tensor variables. To help with this I often name variables like inputs_b_t_f where b = batches, t = time steps, f = features. This combined with something like einops [1] ( intuitive reshaping library) can be a huge time saver. [1] https://github.com/arogozhnikov/einops https://github.com/arogozhnikov/einops