Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Introduce common base class for BotorchSurrogates and EntingSurrogates #283

Draft
wants to merge 2 commits into
base: main
Choose a base branch
from
Draft

Introduce common base class for BotorchSurrogates and EntingSurrogates #283

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
968a481
add surrogates.py
jduerholt Sep 11, 2023
9b62237
remove type attribute
jduerholt Sep 11, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
93 changes: 2 additions & 91 deletions bofire/data_models/surrogates/botorch_surrogates.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,11 +1,5 @@
import itertools
from typing import List, Union

from pydantic import validator

from bofire.data_models.base import BaseModel
from bofire.data_models.domain.api import Inputs, Outputs
from bofire.data_models.features.api import TInputTransformSpecs
from bofire.data_models.surrogates.empirical import EmpiricalSurrogate
from bofire.data_models.surrogates.fully_bayesian import SaasSingleTaskGPSurrogate
from bofire.data_models.surrogates.mixed_single_task_gp import (
Expand All @@ -14,6 +8,7 @@
from bofire.data_models.surrogates.mlp import MLPEnsemble
from bofire.data_models.surrogates.random_forest import RandomForestSurrogate
from bofire.data_models.surrogates.single_task_gp import SingleTaskGPSurrogate
from bofire.data_models.surrogates.surrogates import Surrogates
from bofire.data_models.surrogates.tanimoto_gp import TanimotoGPSurrogate

AnyBotorchSurrogate = Union[
Expand All @@ -27,93 +22,9 @@
]


class BotorchSurrogates(BaseModel):
class BotorchSurrogates(Surrogates):
""" "List of botorch surrogates.

Behaves similar to a Surrogate."""

surrogates: List[AnyBotorchSurrogate]

@property
def input_preprocessing_specs(self) -> TInputTransformSpecs:
return {
key: value
for model in self.surrogates
for key, value in model.input_preprocessing_specs.items()
}

@property
def outputs(self) -> Outputs:
return Outputs(
features=list(
itertools.chain.from_iterable(
[model.outputs.get() for model in self.surrogates] # type: ignore
)
)
)

def _check_compability(self, inputs: Inputs, outputs: Outputs):
# TODO: add sync option
used_output_feature_keys = self.outputs.get_keys()
if sorted(used_output_feature_keys) != sorted(outputs.get_keys()):
raise ValueError("Output features do not match.")
used_feature_keys = []
for i, model in enumerate(self.surrogates):
if len(model.inputs) > len(inputs):
raise ValueError(
f"Model with index {i} has more features than acceptable."
)
for feat in model.inputs:
try:
other_feat = inputs.get_by_key(feat.key)
except KeyError:
raise ValueError(f"Feature {feat.key} not found.")
# now compare the features
# TODO: make more sohisticated comparisons based on the type
# has to to be implemented in features, for the start
# we go with __eq__
if feat != other_feat:
raise ValueError(f"Features with key {feat.key} are incompatible.")
if feat.key not in used_feature_keys:
used_feature_keys.append(feat.key)
if len(used_feature_keys) != len(inputs):
raise ValueError("Unused features are present.")

@validator("surrogates")
def validate_surrogates(cls, v, values):
# validate that all surrogates are single output surrogates
# TODO: this restriction has to be removed at some point
for model in v:
if len(model.outputs) != 1:
raise ValueError("Only single output surrogates allowed.")
# check that the output feature keys are distinctw
used_output_feature_keys = list(
itertools.chain.from_iterable([model.outputs.get_keys() for model in v])
)
if len(set(used_output_feature_keys)) != len(used_output_feature_keys):
raise ValueError("Output feature keys are not unique across surrogates.")
# get the feature keys present in all surrogates
used_feature_keys = []
for model in v:
for key in model.inputs.get_keys():
if key not in used_feature_keys:
used_feature_keys.append(key)
# check that the features and preprocessing steps are equal trough the surrogates
for key in used_feature_keys:
features = [
model.inputs.get_by_key(key)
for model in v
if key in model.inputs.get_keys()
]
preproccessing = [
model.input_preprocessing_specs[key]
for model in v
if key in model.input_preprocessing_specs
]
if all(features) is False:
raise ValueError(f"Features with key {key} are incompatible.")
if all(i == preproccessing[0] for i in preproccessing) is False:
raise ValueError(
f"Preprocessing steps for features with {key} are incompatible."
)
return v
95 changes: 95 additions & 0 deletions bofire/data_models/surrogates/surrogates.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,95 @@
import itertools
from typing import Annotated, List

from pydantic import Field, validator

from bofire.data_models.base import BaseModel
from bofire.data_models.domain.api import Inputs, Outputs
from bofire.data_models.features.api import TInputTransformSpecs


class Surrogates(BaseModel):
type: str
surrogates: Annotated[List, Field(min_items=1)]

@property
def input_preprocessing_specs(self) -> TInputTransformSpecs:
return {
key: value
for model in self.surrogates
for key, value in model.input_preprocessing_specs.items()
}

@property
def outputs(self) -> Outputs:
return Outputs(
features=list(
itertools.chain.from_iterable(
[model.outputs.get() for model in self.surrogates] # type: ignore
)
)
)

@validator("surrogates")
def validate_surrogates(cls, v, values):
# validate that all surrogates are single output surrogates
for model in v:
if len(model.outputs) != 1:
raise ValueError("Only single output surrogates allowed.")
# check that the output feature keys are distinct
used_output_feature_keys = list(
itertools.chain.from_iterable([model.outputs.get_keys() for model in v])
)
if len(set(used_output_feature_keys)) != len(used_output_feature_keys):
raise ValueError("Output feature keys are not unique across surrogates.")
# get the feature keys present in all surrogates
used_feature_keys = []
for model in v:
for key in model.inputs.get_keys():
if key not in used_feature_keys:
used_feature_keys.append(key)
# check that the features and preprocessing steps are equal trough the surrogates
for key in used_feature_keys:
features = [
model.inputs.get_by_key(key)
for model in v
if key in model.inputs.get_keys()
]
preproccessing = [
model.input_preprocessing_specs[key]
for model in v
if key in model.input_preprocessing_specs
]
if all(features) is False:
raise ValueError(f"Features with key {key} are incompatible.")
if all(i == preproccessing[0] for i in preproccessing) is False:
raise ValueError(
f"Preprocessing steps for features with {key} are incompatible."
)
return v

def _check_compability(self, inputs: Inputs, outputs: Outputs):
used_output_feature_keys = self.outputs.get_keys()
if sorted(used_output_feature_keys) != sorted(outputs.get_keys()):
raise ValueError("Output features do not match.")
used_feature_keys = []
for i, model in enumerate(self.surrogates):
if len(model.inputs) > len(inputs):
raise ValueError(
f"Model with index {i} has more features than acceptable."
)
for feat in model.inputs:
try:
other_feat = inputs.get_by_key(feat.key)
except KeyError:
raise ValueError(f"Feature {feat.key} not found.")
# now compare the features
# TODO: make more sohisticated comparisons based on the type
# has to to be implemented in features, for the start
# we go with __eq__
if feat != other_feat:
raise ValueError(f"Features with key {feat.key} are incompatible.")
if feat.key not in used_feature_keys:
used_feature_keys.append(feat.key)
if len(used_feature_keys) != len(inputs):
raise ValueError("Unused features are present.")