Architecture & DevOps for ML Products and Services
Getting started with MLOPs through Architecture & CI/CD best practices at ITP, and extend it with new concepts from the world of Machine Learning.
PyTorch is a Python package that provides two high-level features:
Tensor computation (like NumPy) with strong GPU acceleration
Deep neural networks built on a tape-based autograd system
Pytorch is easier to learn for Software Engineers than Tensorflow.
A lot of R&D these days is happening in Pytorch, and critical building blocks e.g. the Transformers NLP lib show a preference for Pytorch.
With regards to Mobile & Ops, Pytorch has been catching up in 2021 with TorchServe and Pytorch Live.