TensorFlow 2.0 is finally alpha-released. I use Python and R but tend to focus on R these days. So, what changes should R users be aware of? The RStudio team wrote a thorough post on the subject, providing a tour of the new
r-tensorflow ecosystem, with code samples.
The good news? If you’ve been using TensorFlow within
keras, like me, syntactically almost everything remains the same. You’ll notice very few differences, if any. For those unfamiliar with
keras, it is the frontend library for machine learning and AI that uses TensorFlow for AI and deep learning as a low-level backend. Two updates that I’m excited to explore include 1)
tfdatasets for data pipelining, and 2) TensorFlow Hub (available as the R
tfhub package) for publishing and using pre-trained models.
- RStudio Blog: TensorFlow 2.0 is here - what changes for R users?
- TensorFlow Hub: Existing Pretrained Models