R interface for tensorflow hub

We’re happy to announce that the primary model of Tfhub Now it’s in Cran. TFHub is an R Tensorflow Hub interface: a library for publication, discovery and consumption of reusable elements of computerized studying fashions. A module is an autonomous piece of a tensioning movement chart, together with its weights and property, which could be reused in several duties in a course of referred to as switch studying.

The Cran de TFHub model could be put in with:

After putting in the R bundle, you need to set up the tensorflow python bundle. You are able to do it by operating:

Beginning

The important operate of TFHub is layer_hub that works like a keras Capa, however means that you can load an entire pre-questioned deep studying mannequin.

For instance, you may:

library(tfhub)
layer_mobilenet <- layer_hub(
  deal with = "https://tfhub.dev/google/tf2-preview/mobilenet_v2/classification/4"
)

It will obtain the beforehand skilled Mobilenet mannequin within the Imagenet knowledge set. TFHub fashions are saved regionally and don’t have to obtain the subsequent time you utilize the identical mannequin.

Now you need to use layer_mobilenet As a standard layer of keras. For instance, you may outline a mannequin:

library(keras)
enter <- layer_input(form = c(224, 224, 3))
output <- layer_mobilenet(enter)
mannequin <- keras_model(enter, output)
abstract(mannequin)

Mannequin: "mannequin"
____________________________________________________________________
Layer (kind)                  Output Form               Param #    
====================================================================
input_2 (InputLayer)          ((None, 224, 224, 3))      0          
____________________________________________________________________
keras_layer_1 (KerasLayer)    (None, 1001)               3540265    
====================================================================
Whole params: 3,540,265
Trainable params: 0
Non-trainable params: 3,540,265
____________________________________________________________________

This mannequin can now be used to foretell Imagenet labels for a picture. For instance, let’s examine the outcomes of the well-known Grace Hopper picture:

img <- image_load("https://blogs.rstudio.com/tensorflow/posts/photos/grace-hopper.jpg", target_size = c(224,224)) %>% 
  image_to_array()
img <- img/255
dim(img) <- c(1, dim(img))
pred <- predict(mannequin, img)
imagenet_decode_predictions(pred(,-1,drop=FALSE))((1))

  class_name class_description    rating
1  n03763968  military_uniform 9.760404
2  n02817516          bearskin 5.922512
3  n04350905              swimsuit 5.729345
4  n03787032       mortarboard 5.400651
5  n03929855       pickelhaube 5.008665

Tensorflow Hub additionally presents many different beforehand skilled picture, textual content and video fashions. All doable fashions could be discovered within the tensorflow hub web site.

You will discover extra examples of layer_hub Use within the following articles on the tensorflow for R web site:

Use with recipes and the API of traits specs

TFHub additionally presents recipes Steps to facilitate the usage of beforehand skilled deep studying fashions of their computerized studying workflow.

For instance, you may outline a recipe that makes use of a beforehand skilled textual content inlaid mannequin with:

rec <- recipe(obscene ~ comment_text, knowledge = prepare) %>%
  step_pretrained_text_embedding(
    comment_text,
    deal with = "https://tfhub.dev/google/tf2-preview/gnews-swivel-20dim-with-oov/1"
  ) %>%
  step_bin2factor(obscene)

You may see an entire instance of execution right here.

You too can use TFHub with the brand new Function specs API carried out in tfdatasets. You may see an entire instance right here.

We hope that our readers have enjoyable experiencing with Hub fashions and/or can do them in good situation. When you encounter any downside, allow us to create an issue within the TFHub repository

Re-use

The textual content and figures are licensed below the attribution of Inventive Commons CC by 4.0. The figures which have been reused from different sources don’t fall below this license and could be acknowledged by a be aware of their title: “Determine of …”.

Quotation

For attribution, he cites this work as

Falbel (2019, Dec. 18). Posit AI Weblog: tfhub: R interface to TensorFlow Hub. Retrieved from https://blogs.rstudio.com/tensorflow/posts/2019-12-18-tfhub-0.7.0/

Bibtex appointment

@misc{tfhub,
  creator = {Falbel, Daniel},
  title = {Posit AI Weblog: tfhub: R interface to TensorFlow Hub},
  url = {https://blogs.rstudio.com/tensorflow/posts/2019-12-18-tfhub-0.7.0/},
  12 months = {2019}
}