yujunjun/autogen
1
0
Fork 0
mirror of https://github.com/microsoft/autogen.git synced 2025-11-09 14:24:05 +00:00
Code Issues Packages Projects Releases Wiki Activity

update

Browse Source
This commit is contained in:
skzhang1 2023-01-17 06:49:59 -08:00
parent b7c0c24269
commit 3a68da8774
2 changed files with 63 additions and 65 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

90
flaml/tune/searcher/flow2.py
View File

@ -373,25 +373,27 @@ class FLOW2(Searcher):
k_values = np.array(self._histories[k_metric]) k_values = np.array(self._histories[k_metric])
feasible_value = k_values.take(feasible_index) feasible_value = k_values.take(feasible_index)
self._f_best[k_metric] = np.min(feasible_value) self._f_best[k_metric] = np.min(feasible_value)
if not isinstance(self.lexico_objectives["tolerances"][k_metric], str):
tolerance_bound = (
self._f_best[k_metric]
+ self.lexico_objectives["tolerances"][k_metric]
)
else:
assert (
self.lexico_objectives["tolerances"][k_metric][-1] == "%"
), "String tolerance of {} should use %% as the suffix".format(k_metric)
tolerance_bound = self._f_best[k_metric] * (
1
+ 0.01
* float(
self.lexico_objectives["tolerances"][k_metric].replace("%", "")
)
)
feasible_index_filter = np.where( feasible_index_filter = np.where(
feasible_value feasible_value
<= max( <= max(
[ [
self._f_best[k_metric] tolerance_bound,
+ self.lexico_objectives["tolerances"][k_metric]
if not isinstance(
self.lexico_objectives["tolerances"][k_metric], str
)
else self._f_best[k_metric]
* (
1
+ 0.01
* float(
self.lexico_objectives["tolerances"][k_metric].replace(
"%", ""
)
)
),
self.lexico_objectives["targets"][k_metric], self.lexico_objectives["targets"][k_metric],
] ]
) )
@ -417,47 +419,31 @@ class FLOW2(Searcher):
if k_mode == "min" if k_mode == "min"
else -self.lexico_objectives["targets"][k_metric] else -self.lexico_objectives["targets"][k_metric]
) )
if ( if not isinstance(self.lexico_objectives["tolerances"][k_metric], str):
result[k_metric] tolerance_bound = (
< max( self._f_best[k_metric]
[ + self.lexico_objectives["tolerances"][k_metric]
self._f_best[k_metric]
+ self.lexico_objectives["tolerances"][k_metric]
if not isinstance(
self.lexico_objectives["tolerances"][k_metric], str
)
else self._f_best[k_metric]
* (
1
+ 0.01
* float(
self.lexico_objectives["tolerances"][
k_metric
].replace("%", "")
)
),
k_target,
]
) )
) and ( else:
assert (
self.lexico_objectives["tolerances"][k_metric][-1] == "%"
), "String tolerance of {} should use %% as the suffix".format(
k_metric
)
tolerance_bound = self._f_best[k_metric] * (
1
+ 0.01
* float(
self.lexico_objectives["tolerances"][k_metric].replace(
"%", ""
)
)
)
if (result[k_metric] < max([tolerance_bound, k_target])) and (
self.best_obj[k_metric] self.best_obj[k_metric]
< max( < max(
[ [
self._f_best[k_metric] tolerance_bound,
+ self.lexico_objectives["tolerances"][k_metric]
if not isinstance(
self.lexico_objectives["tolerances"][k_metric], str
)
else self._f_best[k_metric]
* (
1
+ 0.01
* float(
self.lexico_objectives["tolerances"][
k_metric
].replace("%", "")
)
),
k_target, k_target,
] ]
) )

38
test/tune/test_lexiflow.py
View File

@ -32,7 +32,7 @@ def _BraninCurrin(config):
return {"brain": brain_result, "currin": currin_result} return {"brain": brain_result, "currin": currin_result}
def test_lexiflow(mode="absolute"): def test_lexiflow():
train_dataset = torchvision.datasets.FashionMNIST( train_dataset = torchvision.datasets.FashionMNIST(
"test/data", "test/data",
train=True, train=True,
@ -105,12 +105,6 @@ def test_lexiflow(mode="absolute"):
lexico_objectives = {} lexico_objectives = {}
lexico_objectives["metrics"] = ["error_rate", "flops"] lexico_objectives["metrics"] = ["error_rate", "flops"]
if mode == "absolute":
lexico_objectives["tolerances"] = {"error_rate": 0.02, "flops": 0.0}
else:
lexico_objectives["tolerances"] = {"error_rate": "10%", "flops": "0%"}
lexico_objectives["targets"] = {"error_rate": 0.0, "flops": 0.0}
lexico_objectives["modes"] = ["min", "min"]
search_space = { search_space = {
"n_layers": tune.randint(lower=1, upper=3), "n_layers": tune.randint(lower=1, upper=3),
@ -132,7 +126,27 @@ def test_lexiflow(mode="absolute"):
"n_epoch": 1, "n_epoch": 1,
} }
# Non lexico tune
analysis = tune.run(
evaluate_function,
metric="error_rate",
mode="min",
num_samples=5,
config=search_space,
use_ray=False,
lexico_objectives=None,
low_cost_partial_config=low_cost_partial_config,
)
print(analysis.best_trial)
print(analysis.best_config)
print(analysis.best_result)
# lexico tune # lexico tune
lexico_objectives["targets"] = {"error_rate": 0.0, "flops": 0.0}
lexico_objectives["modes"] = ["min", "min"]
# 1. lexico tune: absoute tune
lexico_objectives["tolerances"] = {"error_rate": 0.02, "flops": 0.0}
analysis = tune.run( analysis = tune.run(
evaluate_function, evaluate_function,
num_samples=5, num_samples=5,
@ -145,15 +159,14 @@ def test_lexiflow(mode="absolute"):
print(analysis.best_config) print(analysis.best_config)
print(analysis.best_result) print(analysis.best_result)
# Non lexico tune # 2. lexico tune: percentage tolerance
lexico_objectives["tolerances"] = {"error_rate": "10%", "flops": "0%"}
analysis = tune.run( analysis = tune.run(
evaluate_function, evaluate_function,
metric="error_rate",
mode="min",
num_samples=5, num_samples=5,
config=search_space, config=search_space,
use_ray=False, use_ray=False,
lexico_objectives=None, lexico_objectives=lexico_objectives,
low_cost_partial_config=low_cost_partial_config, low_cost_partial_config=low_cost_partial_config,
) )
print(analysis.best_trial) print(analysis.best_trial)
@ -191,6 +204,5 @@ def test_lexiflow_performance():
if __name__ == "__main__": if __name__ == "__main__":
test_lexiflow(mode="absolute") test_lexiflow()
test_lexiflow(mode="percentage")
test_lexiflow_performance() test_lexiflow_performance()