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

improve dnn

This commit is contained in:
Bohan Zhang 2019-12-18 21:37:40 -08:00 committed by Dana Van Aken
parent 3a722df5e2
commit 9633b0e99c
3 changed files with 27 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

35
server/analysis/nn_tf.py
View File

@ -108,6 +108,12 @@ class NeuralNet(object):
with self.graph.as_default(): with self.graph.as_default():
with self.session.as_default(): # pylint: disable=not-context-manager with self.session.as_default(): # pylint: disable=not-context-manager
x_ = tf.Variable(tf.ones([batch_size, self.n_input])) x_ = tf.Variable(tf.ones([batch_size, self.n_input]))
X_min_ = tf.placeholder(tf.float32, [self.n_input])
X_max_ = tf.placeholder(tf.float32, [self.n_input])
x_bounded_ = tf.minimum(x_, X_max_)
x_bounded_ = tf.maximum(x_bounded_, X_min_)
x_bounded_ = tf.cast(x_bounded_, tf.float32)
w1_ = tf.placeholder(tf.float32, [self.n_input, 64]) w1_ = tf.placeholder(tf.float32, [self.n_input, 64])
b1_ = tf.placeholder(tf.float32, [64]) b1_ = tf.placeholder(tf.float32, [64])
w2_ = tf.placeholder(tf.float32, [64, 64]) w2_ = tf.placeholder(tf.float32, [64, 64])
@ -128,6 +134,9 @@ class NeuralNet(object):
self.vars['b1_'] = b1_ self.vars['b1_'] = b1_
self.vars['b2_'] = b2_ self.vars['b2_'] = b2_
self.vars['b3_'] = b3_ self.vars['b3_'] = b3_
self.vars['X_min_'] = X_min_
self.vars['X_max_'] = X_max_
self.vars['x_bounded_'] = x_bounded_
self.ops['train_'] = train_ self.ops['train_'] = train_
def fit(self, X_train, y_train, fit_epochs=500): def fit(self, X_train, y_train, fit_epochs=500):
@ -169,6 +178,10 @@ class NeuralNet(object):
def recommend(self, X_start, X_min=None, X_max=None, recommend_epochs=500, explore=False): def recommend(self, X_start, X_min=None, X_max=None, recommend_epochs=500, explore=False):
batch_size = len(X_start) batch_size = len(X_start)
assert(batch_size == self.batch_size) assert(batch_size == self.batch_size)
if X_min is None:
X_min = np.tile([-np.infty], self.n_input)
if X_max is None:
X_max = np.tile([np.infty], self.n_input)
with self.graph.as_default(): with self.graph.as_default():
with self.session.as_default() as sess: # pylint: disable=not-context-manager with self.session.as_default() as sess: # pylint: disable=not-context-manager
@ -196,7 +209,8 @@ class NeuralNet(object):
y_before = sess.run(self.vars['y_'], y_before = sess.run(self.vars['y_'],
feed_dict={self.vars['w1_']: w1, self.vars['w2_']: w2, feed_dict={self.vars['w1_']: w1, self.vars['w2_']: w2,
self.vars['w3_']: w3, self.vars['b1_']: b1, self.vars['w3_']: w3, self.vars['b1_']: b1,
self.vars['b2_']: b2, self.vars['b3_']: b3}) self.vars['b2_']: b2, self.vars['b3_']: b3,
self.vars['X_max_']: X_max, self.vars['X_min_']: X_min})
if self.debug: if self.debug:
LOG.info("Recommend phase, y before gradient descent: min %f, max %f, mean %f", LOG.info("Recommend phase, y before gradient descent: min %f, max %f, mean %f",
np.min(y_before), np.max(y_before), np.mean(y_before)) np.min(y_before), np.max(y_before), np.mean(y_before))
@ -205,29 +219,24 @@ class NeuralNet(object):
sess.run(self.ops['train_'], sess.run(self.ops['train_'],
feed_dict={self.vars['w1_']: w1, self.vars['w2_']: w2, feed_dict={self.vars['w1_']: w1, self.vars['w2_']: w2,
self.vars['w3_']: w3, self.vars['b1_']: b1, self.vars['w3_']: w3, self.vars['b1_']: b1,
self.vars['b2_']: b2, self.vars['b3_']: b3}) self.vars['b2_']: b2, self.vars['b3_']: b3,
self.vars['X_max_']: X_max, self.vars['X_min_']: X_min})
# constrain by X_min and X_max
if X_min is not None and X_max is not None:
X_train = sess.run(self.vars['x_'])
X_train = np.minimum(X_train, X_max)
X_train = np.maximum(X_train, X_min)
constraint_x_op = self.vars['x_'].assign(X_train)
sess.run(constraint_x_op)
if self.debug and i % self.debug_interval == 0: if self.debug and i % self.debug_interval == 0:
y_train = sess.run(self.vars['y_'], y_train = sess.run(self.vars['y_'],
feed_dict={self.vars['w1_']: w1, self.vars['w2_']: w2, feed_dict={self.vars['w1_']: w1, self.vars['w2_']: w2,
self.vars['w3_']: w3, self.vars['b1_']: b1, self.vars['w3_']: w3, self.vars['b1_']: b1,
self.vars['b2_']: b2, self.vars['b3_']: b3}) self.vars['b2_']: b2, self.vars['b3_']: b3,
self.vars['X_max_']: X_max, self.vars['X_min_']: X_min})
LOG.info("Recommend phase, epoch %d, y: min %f, max %f, mean %f", LOG.info("Recommend phase, epoch %d, y: min %f, max %f, mean %f",
i, np.min(y_train), np.max(y_train), np.mean(y_train)) i, np.min(y_train), np.max(y_train), np.mean(y_train))
y_recommend = sess.run(self.vars['y_'], y_recommend = sess.run(self.vars['y_'],
feed_dict={self.vars['w1_']: w1, self.vars['w2_']: w2, feed_dict={self.vars['w1_']: w1, self.vars['w2_']: w2,
self.vars['w3_']: w3, self.vars['b1_']: b1, self.vars['w3_']: w3, self.vars['b1_']: b1,
self.vars['b2_']: b2, self.vars['b3_']: b3}) self.vars['b2_']: b2, self.vars['b3_']: b3,
X_recommend = sess.run(self.vars['x_']) self.vars['X_max_']: X_max, self.vars['X_min_']: X_min})
X_recommend = sess.run(self.vars['x_bounded_'], feed_dict={self.vars['X_max_']: X_max, self.vars['X_min_']: X_min})
res = NeuralNetResult(minl=y_recommend, minl_conf=X_recommend) res = NeuralNetResult(minl=y_recommend, minl_conf=X_recommend)
if self.debug: if self.debug:

3
server/website/website/settings/constants.py
View File

@ -84,6 +84,9 @@ HP_LEARNING_RATE = 0.001
# ---GRADIENT DESCENT FOR DNN--- # ---GRADIENT DESCENT FOR DNN---
DNN_TRAIN_ITER = 100 DNN_TRAIN_ITER = 100
# Gradient Descent iteration for recommendation
DNN_GD_ITER = 100
DNN_EXPLORE = False DNN_EXPLORE = False
DNN_EXPLORE_ITER = 500 DNN_EXPLORE_ITER = 500

4
server/website/website/tasks/async_tasks.py
View File

@ -44,7 +44,7 @@ from website.settings import (USE_GPFLOW, DEFAULT_LENGTH_SCALE, DEFAULT_MAGNITUD
DNN_TRAIN_ITER, DNN_EXPLORE, DNN_EXPLORE_ITER, DNN_TRAIN_ITER, DNN_EXPLORE, DNN_EXPLORE_ITER,
DNN_NOISE_SCALE_BEGIN, DNN_NOISE_SCALE_END, DNN_NOISE_SCALE_BEGIN, DNN_NOISE_SCALE_END,
DNN_DEBUG, DNN_DEBUG_INTERVAL, GPR_DEBUG, UCB_BETA, DNN_DEBUG, DNN_DEBUG_INTERVAL, GPR_DEBUG, UCB_BETA,
GPR_MODEL_NAME, ENABLE_DUMMY_ENCODER) GPR_MODEL_NAME, ENABLE_DUMMY_ENCODER, DNN_GD_ITER)
from website.settings import INIT_FLIP_PROB, FLIP_PROB_DECAY from website.settings import INIT_FLIP_PROB, FLIP_PROB_DECAY
from website.types import VarType from website.types import VarType
@ -700,7 +700,7 @@ def configuration_recommendation(recommendation_input):
model_nn.set_weights_bin(session.dnn_model) model_nn.set_weights_bin(session.dnn_model)
model_nn.fit(X_scaled, y_scaled, fit_epochs=DNN_TRAIN_ITER) model_nn.fit(X_scaled, y_scaled, fit_epochs=DNN_TRAIN_ITER)
res = model_nn.recommend(X_samples, X_min, X_max, res = model_nn.recommend(X_samples, X_min, X_max,
explore=DNN_EXPLORE, recommend_epochs=MAX_ITER) explore=DNN_EXPLORE, recommend_epochs=DNN_GD_ITER)
session.dnn_model = model_nn.get_weights_bin() session.dnn_model = model_nn.get_weights_bin()
session.save() session.save()