def test(fileName=None):
# training parameters
hidden_units = 128
num_layers = 2
training_batch_size = 1 #100
num_val_batches = 1
model='rnn'
# data_genrator
data_generator = DataGenerator(training_batch_size,fileName)
seq_length = data_generator.seq_length
input_length = data_generator.encoder.size_x()
output_length = data_generator.encoder.size_y()
model_create_time = 0
model_load_time = 0
with tf.Session(config=config.get_tf_config()) as sess:
#print "building model"
start = time.time()
model = Seq2SeqModel(session=sess,
hidden_units=hidden_units,
model=model,
num_layers=num_layers,
seq_length=seq_length,
input_length=input_length,
output_length=output_length,
batch_size=training_batch_size,
scope="model")
end = time.time()
model_create_time = end-start
model.load('vrep/version1/model.ckpt-967')
start = time.time()
model_load_time = start-end
saver = tf.train.Saver(write_version=saver_pb2.SaverDef.V1)
save_path = saver.save(sess, "{}/model.ckpt".format("output"), global_step=967)
def train(training_data_version, output_dir):
# training parameters
hidden_units = 128
num_layers = 2
num_epochs = 1000
training_batch_size = 20
num_val_batches = 1
keep_prob = 1.0
learning_rate = 0.1
#model='lstm'
model = 'rnn'
# data_genrator
model_config_file = './output/' + output_dir + "/params.yaml"
import yaml
with open(model_config_file, 'r') as stream:
model_params = yaml.load(stream)
if 'model' in model_params.keys():
model = model_params['model']
if 'hidden_units' in model_params.keys():
hidden_units = model_params['hidden_units']
if 'num_layers' in model_params.keys():
num_layers = model_params['num_layers']
if 'training_batch_size' in model_params.keys():
training_batch_size = model_params['training_batch_size']
data_generator = DataGenerator(training_batch_size, training_data_version)
print len(data_generator.xseqs)
print data_generator.seq_length
print 'data number of batches', data_generator.num_batches
batches_per_epoch = data_generator.num_batches
seq_length = data_generator.seq_length
input_length = data_generator.encoder.size_x()
output_length = data_generator.encoder.size_y()
with tf.Session(config=config.get_tf_config()) as sess:
print "building model"
model = Seq2SeqModel(session=sess,
hidden_units=hidden_units,
model=model,
num_layers=num_layers,
seq_length=seq_length,
input_length=input_length,
output_length=output_length,
keep_prob=keep_prob,
learning_rate=learning_rate,
batch_size=training_batch_size,
scope="model")
#summary_writer = tf.train.SummaryWriter('output', sess.graph)
model.init_variables()
print "finished building model"
model.fit(data_generator,
num_epochs=num_epochs,
batches_per_epoch=batches_per_epoch,
num_val_batches=num_val_batches,
output_dir="output/" + output_dir)
#model.debug(data_generator)
print "finished training"
def launch_learning_server(model_name, random_id):
import rospy
from learning_ros_server.srv import LearningServer
server_name = get_learning_server_name(random_id)
rospy.init_node(server_name)
global h_to_a_model
with tf.Session(config=config.get_tf_config()) as sess:
h_to_a_model = load_model(model_name, sess)
rospy.Service(server_name, LearningServer,
handle_learning_server_request)
rospy.spin()
def get_learning_model_output(model_name, raw_input, rnn_state, batch_size):
data_generator = rnn_model.DataGenerator(batch_size, raw_input)
seq_length = data_generator.seq_length
print len(data_generator.xseqs)
print data_generator.seq_length
print 'data number of batches', data_generator.num_batches
with tf.Session(config=config.get_tf_config()) as sess:
h_to_a_model = load_model(model_name, sess, seq_length,
data_generator.batch_size)
x, y = data_generator.next_batch()
target = np.argmax(y, axis=2) #target = batch size*seq length *1
probs, outputs, new_rnn_state, image_summary = h_to_a_model.predict_and_return_state(
x, rnn_state,
summary=False) #output = seqlength*batch size* hiddenunits
return (probs, outputs, new_rnn_state, target, x)
def train(model_name, output_dir, model_input):
data_generator = rnn_model.DataGenerator(1, model_input)
num_val_batches = data_generator.num_batches
print len(data_generator.xseqs)
print data_generator.seq_length
print 'data number of batches', data_generator.num_batches
#num_val_batches = 1
seq_length = data_generator.seq_length
#generate training data for two svms
with tf.Session(config=config.get_tf_config()) as sess:
h_to_a_model = load_model(model_name, sess, seq_length)
#summary_writer = tf.summary.FileWriter('output', sess.graph, seq_length)
correct_prediction_outputs = []
wrong_prediction_outputs = []
for _ in xrange(num_val_batches):
x, y = data_generator.next_batch(
) # x/y = batch size*seq length*input_length/output length
target = np.argmax(y, axis=2) #target = batch size*seq length *1
probs, outputs, rnn_state, image_summary = h_to_a_model.predict_and_return_state(
x, None,
summary=False) #output = seqlength*batch size* hiddenunits
#summary_writer.add_summary(image_summary)
prediction = np.argmax(
probs, axis=2) # prediction = batch size*seq length * 1
#print data_generator.xseqs
#print y
#print target[0]
#print prediction[0]
#print outputs[0:2]
correct_prediction = target == prediction #batch size *seq length * 1
for i in xrange(len(outputs)):
if (not rnn_model.is_stump(x[0][i])) and (not rnn_model.is_pad(
x[0][i])): #not stump nd pad
if correct_prediction[0][i]:
correct_prediction_outputs.append(outputs[i][0])
else:
wrong_prediction_outputs.append(outputs[i][0])
print 'num correct prediction traces', len(correct_prediction_outputs)
print 'num wrong prediction traces', len(wrong_prediction_outputs)
correct_prediction_svm = compute_svm(correct_prediction_outputs,
output_dir, 'correct_prediction')
#y_correct_predict = correct_prediction_svm.predict(correct_prediction_outputs)
wrong_prediction_svm = compute_svm(wrong_prediction_outputs,
output_dir, 'wrong_prediction')
def train():
# training parameters
hidden_units = 128
num_layers = 2
num_epochs = 1000
training_batch_size = 20
num_val_batches = 1
keep_prob = 1.0
learning_rate = 0.1
#model='lstm'
model='rnn'
# data_genrator
data_generator = DataGenerator(training_batch_size)
print len(data_generator.xseqs)
print data_generator.seq_length
print 'data number of batches', data_generator.num_batches
batches_per_epoch = data_generator.num_batches
seq_length = data_generator.seq_length
input_length = data_generator.encoder.size_x()
output_length = data_generator.encoder.size_y()
with tf.Session(config=config.get_tf_config()) as sess:
print "building model"
model = Seq2SeqModel(session=sess,
hidden_units=hidden_units,
model=model,
num_layers=num_layers,
seq_length=seq_length,
input_length=input_length,
output_length=output_length,
keep_prob=keep_prob,
learning_rate=learning_rate,
batch_size=training_batch_size,
scope="model")
#summary_writer = tf.train.SummaryWriter('output', sess.graph)
model.init_variables()
print "finished building model"
model.fit(data_generator,
num_epochs=num_epochs,
batches_per_epoch=batches_per_epoch,
num_val_batches=num_val_batches)
#model.debug(data_generator)
print "finished training"
def test(fileName=None, model_name=None):
# training parameters
hidden_units = 128
num_layers = 2
training_batch_size = 1 #100
num_val_batches = 1
model = 'rnn'
# data_genrator
data_generator = DataGenerator(training_batch_size, fileName)
seq_length = data_generator.seq_length
input_length = data_generator.encoder.size_x()
output_length = data_generator.encoder.size_y()
model_create_time = 0
model_load_time = 0
with tf.Session(config=config.get_tf_config()) as sess:
#print "building model"
start = time.time()
model = Seq2SeqModel(session=sess,
hidden_units=hidden_units,
model=model,
num_layers=num_layers,
seq_length=seq_length,
input_length=input_length,
output_length=output_length,
batch_size=training_batch_size,
scope="model")
end = time.time()
model_create_time = end - start
#model.load('vrep/version1/model.ckpt-967')
model.load(model_name)
start = time.time()
model_load_time = start - end
#print "finished building and loading model"
if fileName is None:
num_val_batches = data_generator.num_batches
for _ in xrange(num_val_batches):
val_set = [data_generator.next_batch(validation=True)]
start = time.time()
validate_loss = model.validate(val_set)
end = time.time()
print 'validate loss: ', validate_loss, ' time/sequence: {:.5f}'.format(
(end - start) / num_val_batches / training_batch_size)
else:
x, y = data_generator.next_batch(validation=True)
start = time.time()
probs = model.predict(x)
end = time.time()
#print probs.shape
probs_without_dummy_actions = [i[:-2] for i in probs[0]]
prediction = np.argmax([probs_without_dummy_actions], axis=2)
#prob_prediction = get_prob_prediction(probs_without_dummy_actions)
#print prediction[0]
print prediction[0][-2]
#if prob_prediction == -1:
# print prediction[0][-2]
#else:
# print prob_prediction
print ' '.join(str(p) for p in probs_without_dummy_actions[-2])
#print prob_prediction
print prediction[0][-2]
print prediction[0]
#print data_generator.seqs
#print data_generator.xseqs
print data_generator.yseqs
print model_name
print 'model create time : {:.5f}'.format(model_create_time)
print 'model load time : {:.5f}'.format(model_load_time)
print ' time to predict: {:.5f}'.format((end - start))
#print np.argmax(y, axis=2)[0]
'''
def test(model_name, svm_model_prefix, model_input, action='test'):
#generate training data for two svms
data_generator = rnn_model.DataGenerator(1, model_input)
num_val_batches = data_generator.num_batches
if action == 'testBatch':
print len(data_generator.xseqs)
print data_generator.seq_length
print 'data number of batches', data_generator.num_batches
#num_val_batches = 1
seq_length = data_generator.seq_length
with tf.Session(config=config.get_tf_config()) as sess:
h_to_a_model = load_model(model_name, sess, seq_length)
#summary_writer = tf.summary.FileWriter('output', sess.graph)
prediction_outputs = []
num_seen_predictions_correct_svm = 0
num_unseen_prediction_correct_svm = 0
num_seen_predictions_wrong_svm = 0
num_unseen_prediction_wrong_svm = 0
for _ in xrange(num_val_batches):
x, y = data_generator.next_batch(
) # x/y = batch size*seq length*input_length/output length
#target = np.argmax(y, axis=2) #target = batch size*seq length *1
probs, outputs, image_summary = h_to_a_model.predict_and_return_state(
x, summary=False) #output = seqlength*batch size* hiddenunits
#summary_writer.add_summary(image_summary)
prediction = np.argmax(
probs, axis=2) # prediction = batch size*seq length * 1
#print data_generator.xseqs
#print y
#print target[0]
#print prediction[0]
#print outputs[0:2]
for i in xrange(len(outputs)):
if (not rnn_model.is_stump(x[0][i])) and (not rnn_model.is_pad(
x[0][i])): #not stump nd pad
prediction_outputs.append(outputs[i][0])
if len(prediction_outputs) == 0: #Initial stump
print 1
print -1
else:
correct_prediction_svm = joblib.load(svm_model_prefix +
'correct_prediction_svm.pkl')
wrong_prediction_svm = joblib.load(svm_model_prefix +
'wrong_prediction_svm.pkl')
if action == 'test':
y_correct_predict = correct_prediction_svm.predict(
[prediction_outputs[-1]])
y_wrong_predict = wrong_prediction_svm.predict(
[prediction_outputs[-1]])
print y_correct_predict[-1]
print y_wrong_predict[-1]
#if(y_correct_predict[-1] == 1) and (y_wrong_predict[-1] == -1):
# print 1
#else:
# print 0
print y_correct_predict
print y_wrong_predict
#print data_generator.xseqs
if action == 'testBatch':
y_correct_predict = correct_prediction_svm.predict(
prediction_outputs)
y_wrong_predict = wrong_prediction_svm.predict(
prediction_outputs)
num_seen_predictions_correct_svm = num_seen_predictions_correct_svm + sum(
xx for xx in y_correct_predict if xx == 1)
num_unseen_prediction_correct_svm = num_unseen_prediction_correct_svm + sum(
xx for xx in y_correct_predict if xx == -1)
num_seen_predictions_wrong_svm = num_seen_predictions_wrong_svm + sum(
xx for xx in y_wrong_predict if xx == 1)
num_unseen_prediction_wrong_svm = num_unseen_prediction_wrong_svm + sum(
xx for xx in y_wrong_predict if xx == -1)
if action == 'test':
print data_generator.yseqs
print prediction[0]
if action == 'testBatch':
print "Num seen predictions (correct svm, wrong svm):" + repr(
num_seen_predictions_correct_svm) + "," + repr(
num_seen_predictions_wrong_svm)
print "Num unseen predictions (corect svm, wrong svm):" + repr(
num_unseen_prediction_correct_svm) + "," + repr(
num_unseen_prediction_wrong_svm)
def debug(model_name, model_input):
hidden_units = 128
model = 'rnn'
num_layers = 2
prob_config = config.get_problem_config(rnn_model.PROBLEM_NAME)
observation_length = prob_config['input_length']
action_length = prob_config['output_length']
encoder = rnn_model.Encoder(action_length, observation_length)
input_length = encoder.size_x()
output_length = encoder.size_y()
batch_size = 3
seq_length = 1
#data_generator = rnn_model.DataGenerator(batch_size, model_input)
#seq_length = data_generator.seq_length
#print len(data_generator.xseqs)
#print data_generator.seq_length
#print 'data number of batches', data_generator.num_batches
x1 = np.ones((batch_size, seq_length, input_length), dtype=np.int8)
x2 = 2 * x1
x = np.concatenate((x1, x2), axis=1)
print x1.shape
print x2.shape
print x.shape
with tf.Session(config=config.get_tf_config()) as sess:
#h_to_a_model = load_model(model_name, sess, seq_length, data_generator.batch_size)
print "Before creating model"
h_to_a_model_1 = Seq2SeqModelExt(session=sess,
hidden_units=hidden_units,
model=model,
num_layers=num_layers,
seq_length=seq_length,
input_length=input_length,
output_length=output_length,
batch_size=batch_size,
scope="model")
print "model_rceated"
h_to_a_model_1.init_variables()
tf.train.Saver(max_to_keep=0).save(sess, "output/" + model_name)
#x, y = data_generator.next_batch()
#target = np.argmax(y, axis=2) #target = batch size*seq length *1
#probs, outputs, image_summary = h_to_a_model.predict_and_return_state(x, summary = False) #output = seqlength*batch size* hiddenunits
#print probs
state_size, rnn_state_1 = h_to_a_model_1.get_rnn_state_info(
x1) #output = seqlength*batch size* hiddenunits
print np.array(rnn_state_1).shape
print rnn_state_1
print "Querying again"
state_size, rnn_state_2 = h_to_a_model_1.get_rnn_state_info(
x2, rnn_state_1)
#print state_size
print np.array(rnn_state_2).shape
print rnn_state_2
#for i in tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES, scope='model'):
# print i
tf.reset_default_graph()
#for i in tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES, scope='model'):
# print i
with tf.Session(config=config.get_tf_config()) as sess:
#h_to_a_model = load_model(model_name, sess, seq_length, data_generator.batch_size)
print "Before creating model"
h_to_a_model_2 = Seq2SeqModelExt(session=sess,
hidden_units=hidden_units,
model=model,
num_layers=num_layers,
seq_length=2 * seq_length,
input_length=input_length,
output_length=output_length,
batch_size=batch_size,
scope="model")
h_to_a_model_2.load(model_name)
state_size, rnn_state_3 = h_to_a_model_2.get_rnn_state_info(x)
print rnn_state_3
assert np.array_equal(rnn_state_3, rnn_state_2)
def get_learning_model(model_name, problem_name):
rnn_model.PROBLEM_NAME = problem_name
sess = tf.Session(config=config.get_tf_config())
h_to_a_model = load_model(model_name, sess, 1, 1)
return h_to_a_model
