tests/testthat/test-timeseries.R

In keras: R Interface to 'Keras'

context("timeseries")



test_call_succeeds("timeseries_generator", required_version = "2.1.5", {

  data <- matrix(1:50, nrow = 50, ncol = 1)
  targets <- matrix(1:50, nrow = 50, ncol = 1)
  data_gen <- timeseries_generator(data, targets, length = 10, sampling_rate = 2, batch_size = 2)

  model <- keras_model_sequential() %>%
    layer_lstm(units = 5, input_shape = c(5, 1)) %>%
    layer_dense(units = 1)

  model %>% compile(
    optimizer = "rmsprop",
    loss = "binary_crossentropy",
    metrics = "accuracy"
  )
  expect_warning_if(tensorflow::tf_version() >= "2.1", {
    model %>% fit_generator(data_gen, steps_per_epoch = 10,
                            validation_data = data_gen, validation_steps = 2)
  })
})