PlaidML_Keras_R_layer_output_test_6_1_6_2.R

# Reference: https://github.com/jjallaire/deep-learning-with-r-notebooks

library(keras)
use_backend(backend = "plaidml")

# 6-1 original ver
# get_layer(model, index = 2)  %>% 
#  set_weights(list(embedding_matrix)) %>% 
#  freeze_weights()

# 6-1 plaidML ver
# index = 1のままだとembedding_matrixの頭が全部0なので動かない
get_layer(model, index = 2)  %>% 
  set_weights(list(embedding_matrix)) %>% 
  freeze_weights()
  
# 6-2 original ver  
# model <- keras_model_sequential() %>% 
#   layer_embedding(input_dim = 10000, output_dim = 32) %>% 
#   layer_simple_rnn(units = 32)  # This last layer only returns the last outputs.
# summary(model)

# 6-2 plaidML ver
# input_length = XX を指定しないと動かない。
# このコード以降全て同様

library(keras)
use_backend(backend = "plaidml")

model <- keras_model_sequential() %>% 
  layer_embedding(input_dim = 10000, output_dim = 32, input_length = 32) %>% 
  layer_simple_rnn(units = 32)

summary(model)

# 6-2 original ver 
# model <- keras_model_sequential() %>%
#   layer_embedding(input_dim = max_features, output_dim = 32) %>%
#   layer_simple_rnn(units = 32) %>%
#   layer_dense(units = 1, activation = "sigmoid")
# 

# 6-2 plaidML ver
# input_length = maxlenを指定, rnnだけじゃなくlstmも同様
model <- keras_model_sequential() %>%
  layer_embedding(input_dim = max_features, output_dim = 32, input_length = maxlen) %>%
  layer_simple_rnn(units = 32) %>%
  layer_dense(units = 1, activation = "sigmoid")