from deephyper.benchmark import HpProblem
Problem = HpProblem()
Problem.add_dim('epochs', (5, 500))
Problem.add_dim('nunits_l1', (1, 1000))
Problem.add_dim('nunits_l2', (1, 1000))
Problem.add_dim('activation_l1', ['relu', 'elu', 'selu', 'tanh'])
Problem.add_dim('activation_l2', ['relu', 'elu', 'selu', 'tanh'])
Problem.add_dim('batch_size', (8, 1024))
Problem.add_dim('dropout_l1', (0.0, 1.0))
Problem.add_dim('dropout_l2', (0.0, 1.0))
Problem.add_starting_point(
epochs=5,
nunits_l1=1,
nunits_l2=2,
activation_l1='relu',
activation_l2='relu',
batch_size=8,
dropout_l1=0.0,
dropout_l2=0.0)
if __name__ == '__main__':
print(Problem)
from deephyper.benchmark import HpProblem
# import os
# import sys
# sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from torch_wrapper import use_knl # noqa
Problem = HpProblem()
Problem.add_dim("batch_size", (1, 32))
Problem.add_dim("image_size", [32])
Problem.add_dim("conv1_in_chan", [3])
Problem.add_dim("conv1_out_chan", (3, 64))
Problem.add_dim("conv1_kern", (3, 8))
# Problem.add_dim("pool_size", [2])
hp = Problem.add_dim("pool_size", [2])
print(hp)
Problem.add_dim("conv2_out_chan", (3, 64))
Problem.add_dim("conv2_kern", (3, 8))
# Problem.add_dim("fc1_out", (64, 256))
# Problem.add_dim("fc2_out", (32, 128))
Problem.add_dim("fc1_out", (64, 16384))
Problem.add_dim("fc2_out", (32, 16384))
Problem.add_dim("fc3_out", [10])
if use_knl:
# Problem.add_dim("omp_num_threads", (8, 64))
Problem.add_dim("omp_num_threads", [64])
Problem.add_starting_point(
batch_size=10,
image_size=32,
from deephyper.benchmark import HpProblem
Problem = HpProblem()
sp = {}
# Problem.add_dim("lr", (0.0009, 0.1))
for i in range(1, 18):
Problem.add_dim(f"u{i}", (32, 128))
Problem.add_dim(f"a{i}", [None, "relu", "sigmoid", "tanh"])
sp[f"u{i}"] = 64
sp[f"a{i}"] = "relu"
Problem.add_starting_point(**sp)
Problem.add_starting_point(
**{
"a1": None,
"a10": "relu",
"a11": "relu",
"a12": "tanh",
"a13": "tanh",
"a14": "tanh",
"a15": "relu",
"a16": "relu",
"a17": "tanh",
"a2": "sigmoid",
"a3": None,
"a4": "sigmoid",
"a5": "tanh",
"a6": None,
from deephyper.benchmark import HpProblem
Problem = HpProblem()
Problem.add_dim('epochs', (5,30), default=5)
Problem.add_dim('rnn_type', ['LSTM', 'GRU', 'SimpleRNN'], default='LSTM')
Problem.add_dim('nhidden', (1, 100), default=1)
#network parameters
Problem.add_dim('activation', ['relu', 'elu', 'selu', 'tanh'], default='relu')
Problem.add_dim('batch_size', (8, 1024), default=8)
Problem.add_dim('dropout', (0.0, 1.0), default=0.0)
Problem.add_dim('optimizer', ['sgd', 'rmsprop', 'adagrad', 'adadelta', 'adam', 'adamax', 'nadam'], default='sgd')
# common optimizer parameters
Problem.add_dim('learning_rate', (1e-04, 1e01), default=1e-4)
if __name__ == '__main__':
print(Problem)
from deephyper.benchmark import HpProblem
Problem = HpProblem()
Problem.add_dim('batch_size', (1, 8192))
Problem.add_dim('in_features', (128, 8192))
Problem.add_dim('out_features', (128, 8192))
Problem.add_dim('omp_num_threads', (8, 64))
Problem.add_dim('bias', [0, 1])
Problem.add_starting_point(batch_size=128,
in_features=1024,
out_features=512,
omp_num_threads=64,
bias=0)
if __name__ == '__main__':
print(Problem)
from deephyper.benchmark import HpProblem
# import os
# import sys
# sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from torch_wrapper import use_knl # noqa
Problem = HpProblem()
Problem.add_dim("batch_size", (1, 1024))
Problem.add_dim("image_size", (32, 64))
Problem.add_dim("in_channels", (2, 8))
Problem.add_dim("out_channels", (2, 16))
Problem.add_dim("kernel_size", (2, 4))
if use_knl:
# KGF: unlike 1D and 3D conv*/ problems, not restricted to 64 threads:
Problem.add_dim("omp_num_threads", (8, 64))
Problem.add_starting_point(
batch_size=128,
image_size=32,
in_channels=2,
out_channels=2,
kernel_size=2,
omp_num_threads=64,
)
else:
Problem.add_starting_point(batch_size=128,
image_size=32,
in_channels=2,
out_channels=2,
from deephyper.benchmark import HpProblem
# import os
# import sys
# sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from torch_wrapper import use_knl # noqa
Problem = HpProblem()
Problem.add_dim("batch_size", (1, 8192))
Problem.add_dim("in_features", (128, 8192))
Problem.add_dim("out_features", (128, 8192))
Problem.add_dim("bias", [0, 1])
if use_knl:
Problem.add_dim("omp_num_threads", (8, 64))
Problem.add_starting_point(batch_size=128,
in_features=1024,
out_features=512,
bias=0,
omp_num_threads=64)
else:
Problem.add_starting_point(batch_size=128,
in_features=1024,
out_features=512,
bias=0)
if __name__ == "__main__":
print(Problem)
from deephyper.benchmark import HpProblem
Problem = HpProblem()
Problem.add_dim('log2_batch_size', (5, 10), 7)
Problem.add_dim('nunits_1', (10, 100), 100)
Problem.add_dim('nunits_2', (10, 30), 20)
Problem.add_dim('dropout_1', (0.0, 1.0), 0.2)
Problem.add_dim('dropout_2', (0.0, 1.0), 0.2)
Problem.add_dim('optimizer_type', ['RMSprop', 'Adam'], 'RMSprop')
from deephyper.benchmark import HpProblem
Problem = HpProblem()
Problem.add_dim('epochs', (5, 500), 5)
# benchmark specific parameters
Problem.add_dim('rnn_type', ['LSTM', 'GRU', 'SimpleRNN'], 'LSTM')
Problem.add_dim('embed_hidden_size', (1, 100), 1)
Problem.add_dim('sent_hidden_size', (1, 100), 1)
Problem.add_dim('query_hidden_size', (1, 100), 1)
# network parameters
Problem.add_dim('activation', ['relu', 'elu', 'selu', 'tanh'], 'relu')
Problem.add_dim('batch_size', (8, 1024), 8)
Problem.add_dim('dropout', (0.0, 1.0), 0.0)
Problem.add_dim(
'optimizer',
['sgd', 'rmsprop', 'adagrad', 'adadelta', 'adam', 'adamax', 'nadam'],
'sgd')
# common optimizer parameters
#space['clipnorm'] = (1e-04, 1e01)
#space['clipvalue'] = (1e-04, 1e01)
# optimizer parameters
Problem.add_dim('learning_rate', (1e-04, 1e01), 1e-04)
#space['momentum'] = (0, 1e01)
#space['decay'] = (0, 1e01)
#space['nesterov'] = [False, True]
#space['rho'] = (1e-04, 1e01)
#space['epsilon'] = (1e-08, 1e01)
#space['beta1'] = (1e-04, 1e01)
#space['beta2'] = (1e-04, 1e01)
if __name__ == '__main__':
from deephyper.benchmark import HpProblem
# import os
# import sys
# sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from torch_wrapper import use_knl # noqa
Problem = HpProblem()
Problem.add_dim("batch_size", (1, 1024))
Problem.add_dim("seq_length", (128, 1024))
Problem.add_dim("in_features", (128, 1024))
Problem.add_dim("hidden_units", (128, 1024))
Problem.add_dim("num_layers", (1, 2))
# Problem.add_dim("out_features", (128, 8192))
Problem.add_dim("bias", [0, 1])
if use_knl:
Problem.add_dim("omp_num_threads", (8, 64))
Problem.add_starting_point(
batch_size=128,
seq_length=512,
in_features=1024,
hidden_units=512,
num_layers=1,
bias=0,
omp_num_threads=64,
)
else:
Problem.add_starting_point(
batch_size=128,
from deephyper.benchmark import HpProblem
Problem = HpProblem()
Problem.add_dim('batch_size', (1, 1024))
Problem.add_dim('image_size', (128, 1024))
Problem.add_dim('in_channels', (2, 64))
Problem.add_dim('out_channels', (2, 64))
Problem.add_dim('kernel_size', (2, 16))
Problem.add_dim('omp_num_threads', (8, 64))
Problem.add_starting_point(batch_size=10,
image_size=128,
in_channels=2,
out_channels=2,
kernel_size=2,
omp_num_threads=64)
if __name__ == '__main__':
print(Problem)
from deephyper.benchmark import HpProblem
# import os
# import sys
# sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from torch_wrapper import use_knl # noqa
Problem = HpProblem()
Problem.add_dim("batch_size", (1, 512))
Problem.add_dim("height", (128, 1024))
Problem.add_dim("width", (128, 1024))
Problem.add_dim("in_channels", (2, 64))
Problem.add_dim("out_channels", (2, 64))
Problem.add_dim("kernel_size", (2, 16))
if use_knl:
# KGF: unlike 1D and 3D conv*/ problems, not restricted to 64 threads:
Problem.add_dim("omp_num_threads", (8, 64))
Problem.add_starting_point(
batch_size=128,
height=512,
width=512,
in_channels=3,
out_channels=64,
kernel_size=3,
omp_num_threads=64,
)
else:
Problem.add_starting_point(
batch_size=128,
from deephyper.benchmark import HpProblem
Problem = HpProblem()
Problem.add_dim('units', (1, 100))
Problem.add_dim('activation', [None, 'relu', 'sigmoid', 'tanh'])
Problem.add_dim('lr', (0.0001, 1.))
Problem.add_starting_point(units=10, activation=None, lr=0.01)
if __name__ == '__main__':
print(Problem)
from deephyper.benchmark import HpProblem
Problem = HpProblem()
Problem.add_dim('batch_size', (1, 4096))
Problem.add_dim('image_size', (128, 8192))
Problem.add_dim('in_channels', (2, 1024))
Problem.add_dim('out_channels', (2, 1024))
Problem.add_dim('kernel_size', (2, 64))
Problem.add_dim('omp_num_threads', [64])
Problem.add_starting_point(batch_size=10,
image_size=128,
in_channels=2,
out_channels=2,
kernel_size=2,
omp_num_threads=64)
if __name__ == '__main__':
print(Problem)
from deephyper.benchmark import HpProblem
Problem = HpProblem()
Problem.add_dim('batch_size', (1, 512))
Problem.add_dim('height', (128, 1024))
Problem.add_dim('width', (128, 1024))
Problem.add_dim('in_channels', (2, 64))
Problem.add_dim('out_channels', (2, 64))
Problem.add_dim('kernel_size', (2, 16))
Problem.add_dim('omp_num_threads', (8, 64))
Problem.add_starting_point(batch_size=128,
height=512,
width=512,
in_channels=3,
out_channels=64,
kernel_size=3,
omp_num_threads=64)
if __name__ == '__main__':
print(Problem)
from deephyper.benchmark import HpProblem
Problem = HpProblem()
Problem.add_dim('epochs', (5, 500), 5)
# benchmark specific parameters
Problem.add_dim('nhidden', (1, 100), 1)
Problem.add_dim('nunits', (1, 1000), 1)
# network parameters
Problem.add_dim('activation', ['relu', 'elu', 'selu', 'tanh'], 'relu')
Problem.add_dim('batch_size', (8, 1024), 8)
Problem.add_dim('dropout', (0.0, 1.0), 0.0)
Problem.add_dim(
'optimizer',
['sgd', 'rmsprop', 'adagrad', 'adadelta', 'adam', 'adamax', 'nadam'],
'sgd')
# common optimizer parameters
#Problem.add_dim(['clipnorm'] = (1e-04, 1e01)
#Problem.add_dim(['clipvalue'] = (1e-04, 1e01)
# optimizer parameters
Problem.add_dim('learning_rate', (1e-04, 1e01), 1e-04)
#Problem.add_dim(['momentum'] = (0, 1e01)
#Problem.add_dim(['decay'] = (0, 1e01)
#Problem.add_dim(['nesterov'] = [False, True]
#Problem.add_dim(['rho'] = (1e-04, 1e01)
#Problem.add_dim(['epsilon'] = (1e-08, 1e01)
#Problem.add_dim(['beta1'] = (1e-04, 1e01)
#Problem.add_dim(['beta2'] = (1e-04, 1e01)
from deephyper.benchmark import HpProblem
# import os
# import sys
# sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from torch_wrapper import use_knl # noqa
Problem = HpProblem()
Problem.add_dim("batch_size", (1, 64))
Problem.add_dim("image_size", (16, 128))
Problem.add_dim("in_channels", (2, 64))
Problem.add_dim("out_channels", (2, 64))
Problem.add_dim("kernel_size", (2, 10))
if use_knl:
Problem.add_dim("omp_num_threads", [64])
# Problem.add_dim("omp_num_threads", (8, 64))
Problem.add_starting_point(
batch_size=10,
image_size=28,
in_channels=2,
out_channels=2,
kernel_size=2,
omp_num_threads=64,
)
else:
Problem.add_starting_point(
batch_size=10, image_size=28, in_channels=2, out_channels=2, kernel_size=2
)
# some_file.py
import sys
# insert at 1, 0 is the script path (or '' in REPL)
sys.path.insert(1, '/home/yzamora/perf_pred/deephyper_repo/deephyper/')
from deephyper.benchmark import HpProblem
Problem = HpProblem()
#Width of hidden layers
Problem.add_dim('nunits', (1, 1000))
Problem.add_dim('depth', (1,20))
#Problem.add_dim('nunits_l2', (1, 1000))
Problem.add_dim('activation', ['relu', 'elu', 'selu', 'tanh'])
Problem.add_dim('batch_size', (8, 100))
#Problem.add_dim('dropout_l1', (0.0, 1.0))
#Problem.add_dim('dropout_l2', (0.0, 1.0))
Problem.add_starting_point(
nunits=1,
activation='relu',
batch_size=8,
depth=1
)
if __name__ == '__main__':
print(Problem)
from deephyper.benchmark import HpProblem
# import os
# import sys
# sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from torch_wrapper import use_knl # noqa
Problem = HpProblem()
Problem.add_dim("batch_size", (1, 4096))
Problem.add_dim("image_size", (128, 8192))
Problem.add_dim("in_channels", (2, 1024))
Problem.add_dim("out_channels", (2, 1024))
Problem.add_dim("kernel_size", (2, 64))
if use_knl:
# Problem.add_dim("omp_num_threads", (8, 64))
Problem.add_dim("omp_num_threads", [64])
Problem.add_starting_point(
batch_size=10,
image_size=128,
in_channels=2,
out_channels=2,
kernel_size=2,
omp_num_threads=64,
)
else:
Problem.add_starting_point(batch_size=10,
image_size=128,
in_channels=2,
out_channels=2,
import os
import numpy as np
from deephyper.benchmark import HpProblem
from deephyper.benchmark.benchmark_functions_wrappers import polynome_2
# Problem definition
Problem = HpProblem()
num_dim = 10
for i in range(num_dim):
Problem.add_dim(f'e{i}', (-10.0, 10.0))
Problem.add_starting_point(**{f'e{i}': 10.0 for i in range(num_dim)})
# Definition of the function which runs the model
def run(param_dict):
f, _, _ = polynome_2()
num_dim = 10
x = np.array([param_dict[f'e{i}'] for i in range(num_dim)])
return f(x) # the objective
if __name__ == '__main__':
print(Problem)
from deephyper.benchmark import HpProblem
Problem = HpProblem()
Problem.add_dim('batch_size', (1, 32))
Problem.add_dim('image_size', [32])
Problem.add_dim('conv1_in_chan', [3])
Problem.add_dim('conv1_out_chan', (3, 64))
Problem.add_dim('conv1_kern', (3, 8))
Problem.add_dim('pool_size', [2])
Problem.add_dim('conv2_out_chan', (3, 64))
Problem.add_dim('conv2_kern', (3, 8))
Problem.add_dim('fc1_out', (64, 16384))
Problem.add_dim('fc2_out', (32, 16384))
Problem.add_dim('fc3_out', [10])
Problem.add_dim('omp_num_threads', [64])
Problem.add_starting_point(batch_size=10,
image_size=32,
conv1_in_chan=3,
conv1_out_chan=16,
conv1_kern=5,
pool_size=2,
conv2_out_chan=16,
conv2_kern=5,
fc1_out=128,
fc2_out=84,
fc3_out=10,
omp_num_threads=64)
if __name__ == '__main__':
print(Problem)
import random
from deephyper.benchmark import HpProblem
NDIM = 2
Problem = HpProblem()
def_values = [random.uniform(-3.0, 4.0) for i in range(NDIM)]
for i, startVal in enumerate(def_values, 1):
dim = f"x{i}"
Problem.add_dim(dim, (-3.0, 4.0), default=startVal)
from deephyper.benchmark import HpProblem
Problem = HpProblem()
Problem.add_dim('batch_size', (1, 64))
Problem.add_dim('image_size', (16, 128))
Problem.add_dim('in_channels', (2, 64))
Problem.add_dim('out_channels', (2, 64))
Problem.add_dim('kernel_size', (2, 10))
Problem.add_dim('omp_num_threads', [64])
Problem.add_starting_point(batch_size=10,
image_size=28,
in_channels=2,
out_channels=2,
kernel_size=2,
omp_num_threads=64)
if __name__ == '__main__':
print(Problem)
import os
import numpy as np
from deephyper.benchmark import HpProblem
from deephyper.benchmark.benchmark_functions_wrappers import polynome_2
# Problem definition
Problem = HpProblem()
num_dim = 10
for i in range(num_dim):
Problem.add_dim(f'e{i}', (-10, 10), i)
# Definition of the function which runs the model
def run(param_dict):
f, (a, b), _ = polynome_2()
num_dim = 10
x = np.array([param_dict[f'e{i}'] for i in range(num_dim)])
return f(x) # the objective
if __name__ == '__main__':
print(Problem)