def test_run_training_function_return_value():
from flaml import tune
# Test dict return value
def evaluate_config_dict(config):
metric = (round(config["x"]) - 85000)**2 - config["x"] / config["y"]
return {"metric": metric}
tune.run(
evaluate_config_dict,
config={
"x": tune.qloguniform(lower=1, upper=100000, q=1),
"y": tune.qrandint(lower=2, upper=100000, q=2),
},
metric="metric",
mode="max",
num_samples=100,
)
# Test scalar return value
def evaluate_config_scalar(config):
metric = (round(config["x"]) - 85000)**2 - config["x"] / config["y"]
return metric
tune.run(
evaluate_config_scalar,
config={
"x": tune.qloguniform(lower=1, upper=100000, q=1),
"y": tune.qlograndint(lower=2, upper=100000, q=2),
},
num_samples=100,
mode="max",
)
def search_space(cls, data_size, task):
space = {
'max_leaf': {
'domain': tune.qloguniform(lower=4, upper=data_size, q=1),
'init_value': 4
},
'n_iter': {
'domain': tune.qloguniform(lower=1, upper=data_size, q=1),
'init_value': 1
},
'n_tree_search': {
'domain': tune.qloguniform(lower=1, upper=32768, q=1),
'init_value': 1
},
'opt_interval': {
'domain': tune.qloguniform(lower=1, upper=10000, q=1),
'init_value': 100
},
'learning_rate': {
'domain': tune.loguniform(lower=0.01, upper=20.0)
},
'min_samples_leaf': {
'domain': tune.qloguniform(lower=1, upper=20, q=1),
'init_value': 20
},
}
return space
def search_space(cls, data_size, task):
upper = min(32768, int(data_size))
return {
'n_estimators': {
'domain': tune.qloguniform(lower=4, upper=upper, q=1),
'init_value': 4,
},
'max_leaves': {
'domain': tune.qloguniform(lower=4, upper=upper, q=1),
'init_value': 4,
},
}
def test_config_constraint():
from flaml import tune
# Test dict return value
def evaluate_config_dict(config):
metric = (round(config["x"]) - 85000)**2 - config["x"] / config["y"]
return {"metric": metric}
def config_constraint(config):
if config["y"] >= config["x"]:
return 1
else:
return 0
tune.run(
evaluate_config_dict,
config={
"x": tune.qloguniform(lower=1, upper=100000, q=1),
"y": tune.qrandint(lower=2, upper=100000, q=2),
},
config_constraints=[(config_constraint, ">", 0.5)],
metric="metric",
mode="max",
num_samples=100,
)
def test_define_by_run():
from flaml.tune.space import (
unflatten_hierarchical,
normalize,
indexof,
complete_config,
)
space = {
# Sample a float uniformly between -5.0 and -1.0
"uniform": tune.uniform(-5, -1),
# Sample a float uniformly between 3.2 and 5.4,
# rounding to increments of 0.2
"quniform": tune.quniform(3.2, 5.4, 0.2),
# Sample a float uniformly between 0.0001 and 0.01, while
# sampling in log space
"loguniform": tune.loguniform(1e-4, 1e-2),
# Sample a float uniformly between 0.0001 and 0.1, while
# sampling in log space and rounding to increments of 0.00005
"qloguniform": tune.qloguniform(1e-4, 1e-1, 5e-5),
# Sample a random float from a normal distribution with
# mean=10 and sd=2
# "randn": tune.randn(10, 2),
# Sample a random float from a normal distribution with
# mean=10 and sd=2, rounding to increments of 0.2
# "qrandn": tune.qrandn(10, 2, 0.2),
# Sample a integer uniformly between -9 (inclusive) and 15 (exclusive)
"randint": tune.randint(-9, 15),
# Sample a random uniformly between -21 (inclusive) and 12 (inclusive (!))
# rounding to increments of 3 (includes 12)
"qrandint": tune.qrandint(-21, 12, 3),
# Sample a integer uniformly between 1 (inclusive) and 10 (exclusive),
# while sampling in log space
"lograndint": tune.lograndint(1, 10),
# Sample a integer uniformly between 2 (inclusive) and 10 (inclusive (!)),
# while sampling in log space and rounding to increments of 2
"qlograndint": tune.qlograndint(2, 10, 2),
# Sample an option uniformly from the specified choices
"choice": tune.choice(["a", "b", "c"]),
"const": 5,
}
choice = {"nested": space}
bs = BlendSearch(
space={"c": tune.choice([choice])},
low_cost_partial_config={"c": choice},
metric="metric",
mode="max",
)
print(indexof(bs._gs.space["c"], choice))
print(indexof(bs._gs.space["c"], {"nested": {"const": 1}}))
config = bs._gs.suggest("t1")
print(config)
config = unflatten_hierarchical(config, bs._gs.space)[0]
print(config)
print(normalize({"c": [choice]}, bs._gs.space, config, {}, False))
space["randn"] = tune.randn(10, 2)
cfo = CFO(
space={"c": tune.choice([0, choice])},
metric="metric",
mode="max",
)
for i in range(5):
cfo.suggest(f"t{i}")
# print(normalize(config, bs._gs.space, config, {}, False))
print(complete_config({}, cfo._ls.space, cfo._ls))