py_dropout_model: Change dropout rate in recurrent layer or dropout layer of...

In thfuchs/tsRNN: Time Series Forecasting using Recurrent Neural Networks by Keras

Description

Wrapper for python function. Model requires fixed training mode (else dropout turned off automatically during testing and changing dropout rate has no effect)

Usage

py_dropout_model(model, dropout = 0.1)

Arguments

model	Keras model, required
dropout	new dropout rate

Details

Currently requires python warnings wrapper to suppress tensorflow warning

Value

Keras model

Examples

# Load Model with permanently activated dropout and generate test data
model <- keras::load_model_hdf5(
  system.file("tinytest_data/simple_rnn_dropout.hdf5", package = "tsRNN")
)
test <- array(runif(18, 0.1, 3), dim = c(6L, 2L, 1L))

# 1. Deactivate dropout for testing
model_01 <- py_dropout_model(model, 0)
model_01(test)

# 2. Change dropout level
model_02 <- py_dropout_model(model, 0.8)
model_02(test)

thfuchs/tsRNN documentation built on April 17, 2021, 11:03 p.m.