public
/
ottertune
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix multithread issue in DNN

This commit is contained in:
bohanjason 2019-12-16 22:27:14 -05:00 committed by Dana Van Aken
parent 2a7bc6145f
commit c76c8e7bfb
2 changed files with 144 additions and 112 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
client/driver/fabfile.py vendored
View File

@ -760,9 +760,21 @@ def integration_tests():
response = get_result(upload_code='ottertuneTestTuningDNN')
assert response['status'] == 'good'
# 2rd iteration Test DNN
upload_result(result_dir='./integrationTests/data/', prefix='1__',
upload_code='ottertuneTestTuningDNN')
response = get_result(upload_code='ottertuneTestTuningDNN')
assert response['status'] == 'good'
# Test GPR
LOG.info('Test GPR (gaussian process regression)')
upload_result(result_dir='./integrationTests/data/', prefix='0__',
upload_code='ottertuneTestTuningGPR')
response = get_result(upload_code='ottertuneTestTuningGPR')
assert response['status'] == 'good'
# 2rd iteration Test GPR
upload_result(result_dir='./integrationTests/data/', prefix='1__',
upload_code='ottertuneTestTuningGPR')
response = get_result(upload_code='ottertuneTestTuningGPR')
assert response['status'] == 'good'

56
server/analysis/nn_tf.py
View File

@ -46,6 +46,13 @@ class NeuralNet(object):
self.explore_iters = explore_iters
self.noise_scale_begin = noise_scale_begin
self.noise_scale_end = noise_scale_end
self.vars = {}
self.ops = {}
self.session = tf.Session()
self.graph = tf.get_default_graph()
with self.graph.as_default():
with self.session.as_default(): # pylint: disable=not-context-manager
self.optimizer = tf.train.AdamOptimizer(learning_rate=self.learning_rate)
# input X is placeholder, weights are variables.
self.model = keras.Sequential([
@ -57,15 +64,17 @@ class NeuralNet(object):
self.model.compile(loss='mean_squared_error',
optimizer=self.optimizer,
metrics=['mean_squared_error', 'mean_absolute_error'])
self.vars = {}
self.ops = {}
self.build_graph()
self._build_graph()
def save_weights(self, weights_file):
def save_weights_file(self, weights_file):
with self.graph.as_default():
with self.session.as_default(): # pylint: disable=not-context-manager
self.model.save_weights(weights_file)
def load_weights(self, weights_file):
def load_weights_file(self, weights_file):
try:
with self.graph.as_default():
with self.session.as_default(): # pylint: disable=not-context-manager
self.model.load_weights(weights_file)
if self.debug:
LOG.info('Neural Network Model weights file exists, load weights from the file')
@ -73,10 +82,15 @@ class NeuralNet(object):
LOG.info('Weights file does not match neural network model, train model from scratch')
def get_weights_bin(self):
return pickle.dumps(self.model.get_weights())
with self.graph.as_default():
with self.session.as_default(): # pylint: disable=not-context-manager
weights = self.model.get_weights()
return pickle.dumps(weights)
def set_weights_bin(self, weights):
try:
with self.graph.as_default():
with self.session.as_default(): # pylint: disable=not-context-manager
self.model.set_weights(pickle.loads(weights))
if self.debug:
LOG.info('Neural Network Model weights exists, load the existing weights')
@ -85,10 +99,10 @@ class NeuralNet(object):
# Build same neural network as self.model, But input X is variables,
# weights are placedholders. Find optimial X using gradient descent.
def build_graph(self):
def _build_graph(self):
batch_size = self.batch_size
self.graph = tf.Graph()
with self.graph.as_default():
with self.session.as_default(): # pylint: disable=not-context-manager
x_ = tf.Variable(tf.ones([batch_size, self.n_input]))
w1_ = tf.placeholder(tf.float32, [self.n_input, 64])
b1_ = tf.placeholder(tf.float32, [64])
@ -113,6 +127,8 @@ class NeuralNet(object):
self.ops['train_'] = train_
def fit(self, X_train, y_train, fit_epochs=500):
with self.graph.as_default():
with self.session.as_default(): # pylint: disable=not-context-manager
self.history = self.model.fit(
X_train, y_train, epochs=fit_epochs, verbose=0)
if self.debug:
@ -127,16 +143,18 @@ class NeuralNet(object):
size - 1, mse[size - 1])
def predict(self, X_pred):
with self.graph.as_default():
with self.session.as_default(): # pylint: disable=not-context-manager
return self.model.predict(X_pred)
# Reference: Parameter Space Noise for Exploration.ICLR 2018, https://arxiv.org/abs/1706.01905
def add_noise(self, weights):
scale = self.adaptive_noise_scale()
def _add_noise(self, weights):
scale = self._adaptive_noise_scale()
size = weights.shape[-1]
noise = scale * np.random.normal(size=size)
return weights + noise
def adaptive_noise_scale(self):
def _adaptive_noise_scale(self):
if self.recommend_iters > self.explore_iters:
scale = self.noise_scale_end
else:
@ -147,24 +165,26 @@ class NeuralNet(object):
def recommend(self, X_start, X_min=None, X_max=None, recommend_epochs=500, explore=False):
batch_size = len(X_start)
assert(batch_size == self.batch_size)
with self.graph.as_default():
with self.session.as_default() as sess: # pylint: disable=not-context-manager
w1, b1 = self.model.get_layer(index=0).get_weights()
w2, b2 = self.model.get_layer(index=2).get_weights()
w3, b3 = self.model.get_layer(index=3).get_weights()
if explore is True:
w1 = self.add_noise(w1)
b1 = self.add_noise(b1)
w2 = self.add_noise(w2)
b2 = self.add_noise(b2)
w3 = self.add_noise(w3)
b3 = self.add_noise(b3)
w1 = self._add_noise(w1)
b1 = self._add_noise(b1)
w2 = self._add_noise(w2)
b2 = self._add_noise(b2)
w3 = self._add_noise(w3)
b3 = self._add_noise(b3)
y_predict = self.predict(X_start)
if self.debug:
LOG.info("Recommend phase, y prediction: min %f, max %f, mean %f",
np.min(y_predict), np.max(y_predict), np.mean(y_predict))
with tf.Session(graph=self.graph) as sess:
init = tf.global_variables_initializer()
sess.run(init)
assign_x_op = self.vars['x_'].assign(X_start)