One of my favorite things about TensorFlow 2.0 is that it brings high-level UX and low-level flexibility together fluently.
This is the result of 4 years of watching practitioners and researchers use the product, and thinking about how we can make life better for both.
This is the result of 4 years of watching practitioners and researchers use the product, and thinking about how we can make life better for both.
You no longer have on one hand, a high-level API that's easy to use but inflexible, & on the other hand a low-level API that's flexible but only approachable by experts.
Instead, you have a spectrum of workflows, from the high-level to the low-level. And they're all compatible.
Instead, you have a spectrum of workflows, from the high-level to the low-level. And they're all compatible.
You have an array of ways to build models, each representing a certain level of trade-off between usability and flexibility. You can get started easily and then gradually dive into workflows where you're writing more and more logic from scratch.
The same apply to model training as well -- it can be as simple as `fit` or as advanced as writing a training loop where you are responsible for every last detail.
You can know very little about ML and get started in minutes -- and you'd be using objects in every way compatible with the workflow of a researcher building the next AlphaGo.
