https://www.youtube.com/watch?v=nbJ-2G2GXL0
Siraj Raval explains what PyTorch is. Torch was written in Lua and some people rewrote it in python as PyTorch. Symbolic programing and imperative programming are important concepts to understand to understand what PyTorch is. Somehow this is related to deep learning, which has something to do with neural networks.
Tensorflow - best documentation for machine learning. ok tensorflow is similar to pytorch.
Distributed cross validation
Dynamic Computation Graphing
Writing a whole program before running it.
Static graphs work well for fixed-size networks, but sometimes dynamic graphs are more useful when the input changes. I'm not sure what kinds of changes would make the switch beneficial, but I hope to understand that soon.
autograd package, I guess this is a framework.
Sorry this one post is a mess. There was just so much info and most of it was jargon.
No comments:
Post a Comment