deployTFModel: Deploy a TensorFlow saved model

Description Usage Arguments Details References See Also

View source: R/deployTFModel.R

Description

Deploys a directory containing a Tensorflow saved model file.

Usage

1
deployTFModel(modelDir, ...)

Arguments

modelDir

Path to the saved model directory. MUST contain saved_model.pb or saved_model.pbtxt

...

Additional arguments to deployApp().

Details

Deploy a single Tensorflow saved model as a bundle. Should be passed a directory that contains the saved_model.pb or saved_model.pbtxt file, as well as any variables and assets necessary to load the model.

A saved model directory might look like this:

1
2
3
4
5
./1/
./1/saved_model.pb or ./1/saved_model.pbtxt
./1/variables/
./1/variables/variables.data-00000-of-00001
./1/variables/variables.index

For information on creating saved models, see the Keras method keras::export_savedmodel.keras.engine.training.Model() or the TensorFlow method tensorflow::export_savedmodel(). If using the TensorFlow package for R, the official TensorFlow guide for saving and restoring models may be useful.

References

https://www.tensorflow.org/programmers_guide/saved_model#overview_of_saving_and_restoring_models

See Also

Other Deployment functions: applications, deployAPI, deployApp, deployDoc, deploySite


rstudio/rsconnect documentation built on Jan. 15, 2019, 5:29 a.m.