RStudio AI Weblog: TensorFlow and Keras 2.9

The discharge of Deep Studying with R, 2d Version coincides with new releases of TensorFlow and Keras. Those releases convey many refinements that permit for extra idiomatic and concise R code.

First, the set of Tensor strategies for base R generics has very much expanded. The set of R generics that paintings with TensorFlow Tensors is now fairly in depth:

strategies(elegance = "tensorflow.tensor")

 [1] -           !           !=          [           [<-        
 [6] *           /           &           %/%         %%         
[11] ^           +           <           <=          ==         
[16] >           >=          |           abs         acos       
[21] all         any         aperm       Arg         asin       
[26] atan        cbind       ceiling     Conj        cos        
[31] cospi       digamma     dim         exp         expm1      
[36] flooring       Im          is.finite   is.countless is.nan     
[41] duration      lgamma      log         log10       log1p      
[46] log2        max         imply        min         Mod        
[51] print       prod        vary       rbind       Re         
[56] rep         spherical       signal        sin         sinpi      
[61] kind        sqrt        str         sum         t          
[66] tan         tanpi      

Which means continuously you’ll be able to write the similar code for TensorFlow Tensors as you could possibly for R arrays. For instance, believe this small serve as from Bankruptcy 11 of the e-book:

reweight_distribution <-
  serve as(original_distribution, temperature = 0.5) {
    original_distribution %>%
      { exp(log(.) / temperature) } %>%
      { . / sum(.) }
  }

Notice that purposes like reweight_distribution() paintings with each 1D R vectors and 1D TensorFlow Tensors, since exp(), log(), /, and sum() are all R generics with strategies for TensorFlow Tensors.

In the similar vein, this Keras unencumber brings with it a refinement to the way in which customized elegance extensions to Keras are outlined. In part impressed by means of the brand new R7 syntax, there’s a new circle of relatives of purposes: new_layer_class(), new_model_class(), new_metric_class(), and so forth. This new interface considerably simplifies the quantity of boilerplate code required to outline customized Keras extensions—a pleasing R interface that serves as a facade over the mechanics of sub-classing Python categories. This new interface is the yang to the yin of %py_class%–a strategy to mime the Python elegance definition syntax in R. After all, the “uncooked” API of changing an R6Class() to Python by the use of r_to_py() remains to be to be had for customers that require complete keep watch over.

This unencumber additionally brings with it a cornucopia of small enhancements all through the Keras R interface: up to date print() and plot() strategies for fashions, improvements to freeze_weights() and load_model_tf(), new exported utilities like zip_lists() and %<>%. And let’s now not omit to say a brand new circle of relatives of R purposes for editing the educational charge right through coaching, with a set of integrated schedules like learning_rate_schedule_cosine_decay(), complemented by means of an interface for developing customized schedules with new_learning_rate_schedule_class().

You’ll to find the total unencumber notes for the R applications right here:

The discharge notes for the R applications inform most effective part the tale then again. The R interfaces to Keras and TensorFlow paintings by means of embedding a complete Python procedure in R (by the use of the reticulate bundle). One of the vital main advantages of this design is that R customers have complete get entry to to the entirety in each R and Python. In different phrases, the R interface at all times has function parity with the Python interface—anything else you’ll be able to do with TensorFlow in Python, you’ll be able to do in R simply as simply. This implies the discharge notes for the Python releases of TensorFlow are simply as related for R customers:

Thank you for studying!

Picture by means of Raphael Wild on Unsplash

Reuse

Textual content and figures are approved underneath Ingenious Commons Attribution CC BY 4.0. The figures which were reused from different resources do not fall underneath this license and can also be identified by means of a be aware of their caption: “Determine from …”.

Quotation

For attribution, please cite this paintings as

Kalinowski (2022, June 9). RStudio AI Weblog: TensorFlow and Keras 2.9. Retrieved from https://blogs.rstudio.com/tensorflow/posts/2022-06-09-tf-2-9/

BibTeX quotation

@misc{kalinowskitf29,
  creator = {Kalinowski, Tomasz},
  name = {RStudio AI Weblog: TensorFlow and Keras 2.9},
  url = {https://blogs.rstudio.com/tensorflow/posts/2022-06-09-tf-2-9/},
  yr = {2022}
}