I am working on a BiLSTM for stock data, and I ran into this issue. How can I get around that issue? I already changed the batch size of the data, with no luck. The size of the _train data is 2114 x 1 and the _test data is 529 x 1. Here is what I have so far.

bilstm_mod % layer_embedding(input_dim = 2115, output_dim = 2114) %>% bidirectional(layer_lstm(units = 2114, dropout = 0.4, recurrent_dropout = 0.4)) %>% layer_dense(units = 2114, activation = "sigmoid")

bilstm_mod %>% compile( optimizer = "adam", loss = "binary_crossentropy", metrics = c("accuracy") )

Index

x_train

bilstm_mod %>% fit( x_train, y_train_rm, batch_size = 768, epochs = 4, validation_data = list(x_test, y_test_rm) )

bilstm_mod %>% evaluate(x_train, y_train_rm) bilstm_mod %>% evaluate(x_test, y_test_rm)