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

python style

This commit is contained in:
yangdsh 2019-11-21 04:11:34 +00:00 committed by Dana Van Aken
parent f0c6d7ef1f
commit f34e879cb1
3 changed files with 10 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
server/analysis/gpr/gpr_models.py
View File

@ -55,7 +55,7 @@ class BaseModel(object):
if optimize_hyperparameters: if optimize_hyperparameters:
opt = gpflow.train.AdamOptimizer(learning_rate) opt = gpflow.train.AdamOptimizer(learning_rate)
opt.minimize(m, maxiter=maxiter) opt.minimize(m, maxiter=maxiter)
self._model = m self.model = m
def _get_kernel_kwargs(self, **kwargs): def _get_kernel_kwargs(self, **kwargs):
return [] return []
@ -65,7 +65,7 @@ class BaseModel(object):
def get_hyperparameters(self): def get_hyperparameters(self):
return {k: float(v) if v.ndim == 0 else v.tolist() return {k: float(v) if v.ndim == 0 else v.tolist()
for k, v in self._model.read_values().items()} for k, v in self.model.read_values().items()}
def get_model_parameters(self): def get_model_parameters(self):
return { return {

13
server/analysis/simulation.py
View File

@ -179,7 +179,7 @@ def dnn(env, config, n_loops=100):
for entry in top10: for entry in top10:
X_samples = np.vstack((X_samples, np.array(entry[0]))) X_samples = np.vstack((X_samples, np.array(entry[0])))
tf.reset_default_graph() tf.reset_default_graph()
sess = tf.InteractiveSession() tf.InteractiveSession()
model_nn = NeuralNet(n_input=X_samples.shape[1], model_nn = NeuralNet(n_input=X_samples.shape[1],
batch_size=X_samples.shape[0], batch_size=X_samples.shape[0],
learning_rate=0.005, learning_rate=0.005,
@ -267,7 +267,7 @@ def run_optimize(X, y, X_samples, model_name, opt_kwargs, model_kwargs):
# Optimize the DBMS's configuration knobs # Optimize the DBMS's configuration knobs
timer.start() timer.start()
X_news, ypreds, yvar, loss = tf_optimize(m._model, X_samples, **opt_kwargs) X_news, ypreds, _, _ = tf_optimize(m.model, X_samples, **opt_kwargs)
timer.stop() timer.stop()
config_optimize_sec = timer.elapsed_seconds config_optimize_sec = timer.elapsed_seconds
@ -313,7 +313,8 @@ def gpr_new(env, config, n_loops=100):
actions, rewards = memory.get_all() actions, rewards = memory.get_all()
ucb_beta = config['beta'] ucb_beta = config['beta']
opt_kwargs['ucb_beta'] = ucb.get_ucb_beta(ucb_beta, scale=config['scale'], t=i + 1., ndim=env.knob_dim) opt_kwargs['ucb_beta'] = ucb.get_ucb_beta(ucb_beta, scale=config['scale'],
t=i + 1., ndim=env.knob_dim)
if model_opt_frequency > 0: if model_opt_frequency > 0:
optimize_hyperparams = i % model_opt_frequency == 0 optimize_hyperparams = i % model_opt_frequency == 0
if not optimize_hyperparams: if not optimize_hyperparams:
@ -397,15 +398,13 @@ def main():
title = 'dim=192' title = 'dim=192'
n_repeats = [1, 1, 1, 1, 1, 1] n_repeats = [1, 1, 1, 1, 1, 1]
n_loops = 200 n_loops = 200
configs = [ configs = [{'num_collections': 5, 'num_samples': 30, 'beta': 'get_beta_td', 'scale': 0.1},
{'num_collections': 5, 'num_samples': 30, 'beta': 'get_beta_td', 'scale': 0.1},
{'num_collections': 5, 'num_samples': 30, 'beta': 'get_beta_td', 'scale': 0.2}, {'num_collections': 5, 'num_samples': 30, 'beta': 'get_beta_td', 'scale': 0.2},
{'num_collections': 5, 'num_samples': 30, 'beta': 'get_beta_td', 'scale': 0.6}, {'num_collections': 5, 'num_samples': 30, 'beta': 'get_beta_td', 'scale': 0.6},
{'num_collections': 5, 'num_samples': 30}, {'num_collections': 5, 'num_samples': 30},
{'gamma': 0., 'c_lr': 0.001, 'a_lr': 0.02, 'num_collections': 1, 'n_epochs': 30, {'gamma': 0., 'c_lr': 0.001, 'a_lr': 0.02, 'num_collections': 1, 'n_epochs': 30,
'a_hidden_sizes': [128, 128, 64], 'c_hidden_sizes': [64, 128, 64]}, 'a_hidden_sizes': [128, 128, 64], 'c_hidden_sizes': [64, 128, 64]},
{'num_collections': 5, 'num_samples': 30} {'num_collections': 5, 'num_samples': 30}]
]
tuners = [gpr_new, gpr_new, gpr_new, gpr, ddpg, dnn] tuners = [gpr_new, gpr_new, gpr_new, gpr, ddpg, dnn]
labels = ['gpr_new_0.5', 'gpr_new_1', 'gpr_new_3', 'gpr', 'ddpg', 'dnn'] labels = ['gpr_new_0.5', 'gpr_new_1', 'gpr_new_3', 'gpr', 'ddpg', 'dnn']
run(tuners, configs, labels, title, env, n_loops, n_repeats) run(tuners, configs, labels, title, env, n_loops, n_repeats)