Hello. please advise resources for studying (tensors, autograd, NN abstractions, optimizers, data pipeline, training loop abstraction, distributed training, graph compilation) this is to get the necessary skills for creating a framework for computer vision from scratch (python and numpy). if something is missing in the list, then tell me what. and if you have implementations from scratch then please share. Thanks you
I would highly recommend to go through fastai and pytorch docs.
I got good understanding from pytorch docs and tutorials. It has specific tutorials for Computer Vision
Is implementation from scratch? there seems to be based on torch
The course sets itself up to where we cannot use any of the torch functionalities until we build them ourselves/andor improve them. Along with that it goes over the fundamentals of what you need to get a deep learning library up and running