rnn_train.py

from keras.models import Sequential
from keras.layers import Dense, LSTM, Dropout, Activation
from keras.optimizers import Adam
import numpy as np
import pickle

def lstm_model(input_shape):
   hidden_units = 512  # may increase/decrease depending on capacity needed
   timesteps = 20
   input_dim = input_shape
   num_classes = 22    # num of classes for output
   model = Sequential()
   model.add(LSTM(hidden_units, input_shape=(timesteps, input_dim)))
   model.add(Dense(256))
   model.add(Dropout(0.1))
   model.add(Activation('relu'))

   model.add(Dense(128))
   model.add(Dropout(0.1))
   model.add(Activation('relu'))

   model.add(Dense(64))
   model.add(Dropout(0.1))
   model.add(Activation('relu'))
   #
   # model.add(Dense(32))
   # model.add(Dropout(0.25))
   # model.add(Activation('relu'))
   # output layer
   # model.add(Dense(2))


   model.add(Dense(num_classes))
   model.add(Activation('softmax'))
   adam = Adam(lr=0.001)
   model.compile(loss='categorical_crossentropy', optimizer=adam, metrics=['accuracy'])
   # model.compile(loss='mse', optimizer=adam, metrics=['accuracy'])
   model.summary()
   return model

def train(load_path, save_path, epochs):
   print("loading samples")

   pickle_in = open(load_path, "rb")

   data = pickle.load(pickle_in)
   print("samples loaded")
   xt = data['bigdata1']  # input_data shape = (num_trials, timesteps, input_dim)
   # yt = data['y_data'].reshape(24000,6)  # out_data shape = (num_trials, num_classes)
   print xt.shape[2]
   yt = data['class']
   from sklearn.preprocessing import OneHotEncoder, normalize
   #xt = normalize(xt)
   enc = OneHotEncoder()
   yt = enc.fit_transform(yt).toarray()
   print(yt[:10])
   print(xt.shape, yt.shape)
   batch_size = 64

   epochs = epochs
   model = lstm_model(xt.shape[2])
   model.fit(xt, yt, epochs=epochs, batch_size=batch_size, shuffle=True)
   model.save(save_path)

if __name__ == '__main__':

	import argparse
	parser = argparse.ArgumentParser()
	parser.add_argument('data_path', type=str)
	parser.add_argument('-e', type=int, default=50)
	parser.add_argument('-s', type=str)

	args = parser.parse_args()

	train(args.data_path, args.s, args.e)