#!/usr/bin/env python3
# Copyright (c) Meta Platforms, Inc. and affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
import functools
from copy import deepcopy
from random import randint
from typing import Any, Callable, Dict, List, Optional, Protocol, Tuple, Type, Union
import torch
from ax.models.model_utils import best_observed_point, get_observed
from ax.models.torch.utils import _to_inequality_constraints
from ax.models.torch_base import TorchModel
from ax.models.types import TConfig
from botorch.acquisition import get_acquisition_function
from botorch.acquisition.acquisition import AcquisitionFunction
from botorch.acquisition.fixed_feature import FixedFeatureAcquisitionFunction
from botorch.acquisition.objective import ConstrainedMCObjective, GenericMCObjective
from botorch.acquisition.utils import get_infeasible_cost
from botorch.exceptions.errors import UnsupportedError
from botorch.fit import fit_gpytorch_mll
from botorch.models.gp_regression import SingleTaskGP
from botorch.models.gp_regression_fidelity import SingleTaskMultiFidelityGP
from botorch.models.gpytorch import GPyTorchModel
from botorch.models.model import Model
from botorch.models.model_list_gp_regression import ModelListGP
from botorch.models.multitask import MultiTaskGP
from botorch.models.transforms.input import Warp
from botorch.optim.optimize import optimize_acqf
from botorch.utils import (
get_objective_weights_transform,
get_outcome_constraint_transforms,
)
from botorch.utils.multi_objective.scalarization import get_chebyshev_scalarization
from gpytorch.kernels import MaternKernel, ScaleKernel
from gpytorch.kernels.kernel import Kernel
from gpytorch.mlls.exact_marginal_log_likelihood import ExactMarginalLogLikelihood
from gpytorch.mlls.leave_one_out_pseudo_likelihood import LeaveOneOutPseudoLikelihood
from gpytorch.mlls.sum_marginal_log_likelihood import SumMarginalLogLikelihood
from gpytorch.priors import Prior
from gpytorch.priors.lkj_prior import LKJCovariancePrior
from gpytorch.priors.torch_priors import GammaPrior, LogNormalPrior
from torch import Tensor
MIN_OBSERVED_NOISE_LEVEL = 1e-6
NO_FEASIBLE_POINTS_MESSAGE = (
"There are no feasible observed points. This likely means that one "
"or more outcome constraints or objective thresholds is set too strictly. "
)
def _construct_model(
task_feature: Optional[int],
Xs: List[Tensor],
Ys: List[Tensor],
Yvars: List[Tensor],
fidelity_features: List[int],
metric_names: List[str],
use_input_warping: bool = False,
prior: Optional[Dict[str, Any]] = None,
*,
multitask_gp_ranks: Optional[Dict[str, Union[Prior, float]]] = None,
**kwargs: Any,
) -> GPyTorchModel:
"""
Figures out how to call `_get_model` depending on inputs. Used by
`get_and_fit_model`.
"""
if task_feature is None:
if len(Xs) == 1:
# Use single output, single task GP
return _get_model(
X=Xs[0],
Y=Ys[0],
Yvar=Yvars[0],
task_feature=task_feature,
fidelity_features=fidelity_features,
use_input_warping=use_input_warping,
prior=deepcopy(prior),
**kwargs,
)
if all(torch.equal(Xs[0], X) for X in Xs[1:]) and not use_input_warping:
# Use batched multioutput, single task GP
# Require using a ModelListGP if using input warping
Y = torch.cat(Ys, dim=-1)
Yvar = torch.cat(Yvars, dim=-1)
return _get_model(
X=Xs[0],
Y=Y,
Yvar=Yvar,
task_feature=task_feature,
fidelity_features=fidelity_features,
prior=deepcopy(prior),
**kwargs,
)
if task_feature is None:
models = [
_get_model(
X=X,
Y=Y,
Yvar=Yvar,
use_input_warping=use_input_warping,
prior=deepcopy(prior),
**kwargs,
)
for X, Y, Yvar in zip(Xs, Ys, Yvars)
]
else:
# use multi-task GP
mtgp_rank_dict = {} if multitask_gp_ranks is None else multitask_gp_ranks
# assembles list of ranks associated with each metric
if len({len(Xs), len(Ys), len(Yvars), len(metric_names)}) > 1:
raise ValueError(
"Lengths of Xs, Ys, Yvars, and metric_names must match. Your "
f"inputs have lengths {len(Xs)}, {len(Ys)}, {len(Yvars)}, and "
f"{len(metric_names)}, respectively."
)
mtgp_rank_list = [mtgp_rank_dict.get(metric, None) for metric in metric_names]
models = [
_get_model(
X=X,
Y=Y,
Yvar=Yvar,
task_feature=task_feature,
rank=mtgp_rank,
use_input_warping=use_input_warping,
prior=deepcopy(prior),
**kwargs,
)
for X, Y, Yvar, mtgp_rank in zip(Xs, Ys, Yvars, mtgp_rank_list)
]
return ModelListGP(*models)
[docs]def get_and_fit_model(
Xs: List[Tensor],
Ys: List[Tensor],
Yvars: List[Tensor],
task_features: List[int],
fidelity_features: List[int],
metric_names: List[str],
state_dict: Optional[Dict[str, Tensor]] = None,
refit_model: bool = True,
use_input_warping: bool = False,
use_loocv_pseudo_likelihood: bool = False,
prior: Optional[Dict[str, Any]] = None,
*,
multitask_gp_ranks: Optional[Dict[str, Union[Prior, float]]] = None,
**kwargs: Any,
) -> GPyTorchModel:
r"""Instantiates and fits a botorch GPyTorchModel using the given data.
N.B. Currently, the logic for choosing ModelListGP vs other models is handled
using if-else statements in lines 96-137. In the future, this logic should be
taken care of by modular botorch.
Args:
Xs: List of X data, one tensor per outcome.
Ys: List of Y data, one tensor per outcome.
Yvars: List of observed variance of Ys.
task_features: List of columns of X that are tasks.
fidelity_features: List of columns of X that are fidelity parameters.
metric_names: Names of each outcome Y in Ys.
state_dict: If provided, will set model parameters to this state
dictionary. Otherwise, will fit the model.
refit_model: Flag for refitting model.
prior: Optional[Dict]. A dictionary that contains the specification of
GP model prior. Currently, the keys include:
- covar_module_prior: prior on covariance matrix e.g.
{"lengthscale_prior": GammaPrior(3.0, 6.0)}.
- type: type of prior on task covariance matrix e.g.`LKJCovariancePrior`.
- sd_prior: A scalar prior over nonnegative numbers, which is used for the
default LKJCovariancePrior task_covar_prior.
- eta: The eta parameter on the default LKJ task_covar_prior.
kwargs: Passed to `_get_model`.
Returns:
A fitted GPyTorchModel.
"""
if len(fidelity_features) > 0 and len(task_features) > 0:
raise NotImplementedError(
"Currently do not support MF-GP models with task_features!"
)
if len(fidelity_features) > 1:
raise NotImplementedError(
"Fidelity MF-GP models currently support only a single fidelity parameter!"
)
if len(task_features) > 1:
raise NotImplementedError(
f"This model only supports 1 task feature (got {task_features})"
)
elif len(task_features) == 1:
task_feature = task_features[0]
else:
task_feature = None
model = _construct_model(
task_feature=task_feature,
Xs=Xs,
Ys=Ys,
Yvars=Yvars,
fidelity_features=fidelity_features,
metric_names=metric_names,
use_input_warping=use_input_warping,
prior=prior,
multitask_gp_ranks=multitask_gp_ranks,
**kwargs,
)
# TODO: Better logic for deciding when to use a ModelListGP. Currently the
# logic is unclear. The two cases in which ModelListGP is used are
# (i) the training inputs (Xs) are not the same for the different outcomes, and
# (ii) a multi-task model is used
model.to(Xs[0])
if state_dict is not None:
model.load_state_dict(state_dict)
if state_dict is None or refit_model:
# TODO: Add bounds for optimization stability - requires revamp upstream
bounds = {}
if use_loocv_pseudo_likelihood:
mll_cls = LeaveOneOutPseudoLikelihood
else:
mll_cls = ExactMarginalLogLikelihood
if isinstance(model, ModelListGP):
mll = SumMarginalLogLikelihood(model.likelihood, model, mll_cls=mll_cls)
else:
mll = mll_cls(model.likelihood, model)
mll = fit_gpytorch_mll(mll, optimizer_kwargs={"bounds": bounds})
return model
[docs]class TAcqfConstructor(Protocol):
def __call__(
self, # making this a static method makes Pyre unhappy, better to keep `self`
model: Model,
objective_weights: Tensor,
outcome_constraints: Optional[Tuple[Tensor, Tensor]] = None,
X_observed: Optional[Tensor] = None,
X_pending: Optional[Tensor] = None,
**kwargs: Any,
) -> AcquisitionFunction:
... # pragma: no cover
[docs]def get_acqf(
acquisition_function_name: str,
) -> Callable[[Callable[[], None]], TAcqfConstructor]:
"""Returns a decorator whose wrapper function instantiates an acquisition function.
NOTE: This is a decorator factory instead of a simple factory as serialization
of Botorch model kwargs requires callables to be have module-level paths, and
closures created by a simple factory do not have such paths. We solve this by
wrapping "empty" module-level functions with this decorator, we ensure that they
are serialized correctly, in addition to reducing code duplication.
Example:
>>> @get_acqf("qEI")
... def get_qEI() -> None:
... pass
>>> acqf = get_qEI(
... model=model,
... objective_weights=objective_weights,
... outcome_constraints=outcome_constraints,
... X_observed=X_observed,
... X_pending=X_pending,
... **kwargs,
... )
>>> type(acqf)
... botorch.acquisition.monte_carlo.qExpectedImprovement
Args:
acquisition_function_name: The name of the acquisition function to be
instantiated by the returned function.
Returns:
A decorator whose wrapper function is a TAcqfConstructor, i.e. it requires a
`model`, `objective_weights`, and optional `outcome_constraints`, `X_observed`,
and `X_pending` as inputs, as well as `kwargs`, and returns an
`AcquisitionFunction` instance that corresponds to `acquisition_function_name`.
"""
def decorator(empty_acqf_getter: Callable[[], None]) -> TAcqfConstructor:
# `wraps` allows the function to keep its original, module-level name, enabling
# serialization via `callable_to_reference`. `empty_acqf_getter` is otherwise
# not used in the wrapper.
@functools.wraps(empty_acqf_getter)
def wrapper(
model: Model,
objective_weights: Tensor,
outcome_constraints: Optional[Tuple[Tensor, Tensor]] = None,
X_observed: Optional[Tensor] = None,
X_pending: Optional[Tensor] = None,
**kwargs: Any,
) -> AcquisitionFunction:
return _get_acquisition_func(
model=model,
acquisition_function_name=acquisition_function_name,
objective_weights=objective_weights,
outcome_constraints=outcome_constraints,
X_observed=X_observed,
X_pending=X_pending,
**kwargs,
)
return wrapper
return decorator
[docs]@get_acqf("qEI")
def get_qEI() -> None:
"""A TAcqfConstructor to instantiate a qEI acquisition function. The function body
is filled in by the decorator function `get_acqf` to simultaneously reduce code
duplication and allow serialization in Ax. TODO: Deprecate with legacy Ax model.
"""
[docs]@get_acqf("qLogEI")
def get_qLogEI() -> None:
"""TAcqfConstructor instantiating qLogEI. See docstring of get_qEI for details."""
[docs]@get_acqf("qNEI")
def get_NEI() -> None: # no "q" in method name for backward compatibility
"""TAcqfConstructor instantiating qNEI. See docstring of get_qEI for details."""
[docs]@get_acqf("qLogNEI")
def get_qLogNEI() -> None:
"""TAcqfConstructor instantiating qLogNEI. See docstring of get_qEI for details."""
def _get_acquisition_func(
model: Model,
acquisition_function_name: str,
objective_weights: Tensor,
outcome_constraints: Optional[Tuple[Tensor, Tensor]] = None,
X_observed: Optional[Tensor] = None,
X_pending: Optional[Tensor] = None,
mc_objective: Type[GenericMCObjective] = GenericMCObjective,
constrained_mc_objective: Optional[
Type[ConstrainedMCObjective]
] = ConstrainedMCObjective,
# pyre-fixme[24]: Generic type `dict` expects 2 type parameters, use
# `typing.Dict` to avoid runtime subscripting errors.
mc_objective_kwargs: Optional[Dict] = None,
*,
chebyshev_scalarization: bool = False,
prune_baseline: bool = True,
mc_samples: int = 512,
marginalize_dim: Optional[int] = None,
) -> AcquisitionFunction:
r"""Instantiates a acquisition function.
Args:
model: The underlying model which the acqusition function uses
to estimate acquisition values of candidates.
acquisition_function_name: Name of the acquisition function.
objective_weights: The objective is to maximize a weighted sum of
the columns of f(x). These are the weights.
outcome_constraints: A tuple of (A, b). For k outcome constraints
and m outputs at f(x), A is (k x m) and b is (k x 1) such that
A f(x) <= b. (Not used by single task models)
X_observed: A tensor containing points observed for all objective
outcomes and outcomes that appear in the outcome constraints (if
there are any).
X_pending: A tensor containing points whose evaluation is pending (i.e.
that have been submitted for evaluation) present for all objective
outcomes and outcomes that appear in the outcome constraints (if
there are any).
mc_objective: GenericMCObjective class, used for constructing a
MC-objective. If constructing a penalized MC-objective, pass in
PenalizedMCObjective together with mc_objective_kwargs .
constrained_mc_objective: ConstrainedMCObjective class, used when
applying constraints on the outcomes.
mc_objective_kwargs: kwargs for constructing MC-objective.
For GenericMCObjective, leave it as None. For PenalizedMCObjective,
it needs to be specified in the format of kwargs.
mc_samples: The number of MC samples to use (default: 512).
prune_baseline: If True, prune the baseline points for NEI (default: True).
chebyshev_scalarization: Use augmented Chebyshev scalarization.
Returns:
The instantiated acquisition function.
"""
if acquisition_function_name not in [
"qSR",
"qEI",
"qLogEI",
"qPI",
"qNEI",
"qLogNEI",
]:
raise NotImplementedError(f"{acquisition_function_name=} not implemented yet.")
if X_observed is None:
raise ValueError(NO_FEASIBLE_POINTS_MESSAGE)
# construct Objective module
if chebyshev_scalarization:
with torch.no_grad():
Y = model.posterior(X_observed).mean # pyre-ignore [16]
obj_tf = get_chebyshev_scalarization(weights=objective_weights, Y=Y)
else:
obj_tf = get_objective_weights_transform(objective_weights)
# pyre-fixme[53]: Captured variable `obj_tf` is not annotated.
def objective(samples: Tensor, X: Optional[Tensor] = None) -> Tensor:
return obj_tf(samples)
mc_objective_kwargs = {} if mc_objective_kwargs is None else mc_objective_kwargs
objective = mc_objective(objective=objective, **mc_objective_kwargs)
if outcome_constraints is None:
con_tfs = None
else:
con_tfs = get_outcome_constraint_transforms(outcome_constraints)
# All acquisition functions registered in BoTorch's `get_acquisition_function`
# except qSR and qUCB support a principled treatment of the constraints by
# directly passing them to the constructor.
if acquisition_function_name == "qSR":
if constrained_mc_objective is None:
raise ValueError(
"constrained_mc_objective cannot be set to None "
"when applying outcome constraints."
)
inf_cost = get_infeasible_cost(
X=X_observed, model=model, objective=objective
)
objective = constrained_mc_objective(
objective=objective, constraints=con_tfs or [], infeasible_cost=inf_cost
)
return get_acquisition_function(
acquisition_function_name=acquisition_function_name,
model=model,
objective=objective,
X_observed=X_observed,
X_pending=X_pending,
prune_baseline=prune_baseline,
mc_samples=mc_samples,
seed=randint(1, 10000),
marginalize_dim=marginalize_dim,
constraints=con_tfs,
)
[docs]def scipy_optimizer(
acq_function: AcquisitionFunction,
bounds: Tensor,
n: int,
inequality_constraints: Optional[List[Tuple[Tensor, Tensor, float]]] = None,
equality_constraints: Optional[List[Tuple[Tensor, Tensor, float]]] = None,
fixed_features: Optional[Dict[int, float]] = None,
rounding_func: Optional[Callable[[Tensor], Tensor]] = None,
*,
num_restarts: int = 20,
raw_samples: Optional[int] = None,
joint_optimization: bool = False,
options: Optional[Dict[str, Union[bool, float, int, str]]] = None,
) -> Tuple[Tensor, Tensor]:
r"""Optimizer using scipy's minimize module on a numpy-adpator.
Args:
acq_function: A botorch AcquisitionFunction.
bounds: A `2 x d`-dim tensor, where `bounds[0]` (`bounds[1]`) are the
lower (upper) bounds of the feasible hyperrectangle.
n: The number of candidates to generate.
inequality constraints: A list of tuples (indices, coefficients, rhs),
with each tuple encoding an inequality constraint of the form
`\sum_i (X[indices[i]] * coefficients[i]) >= rhs`
equality constraints: A list of tuples (indices, coefficients, rhs),
with each tuple encoding an equality constraint of the form
`\sum_i (X[indices[i]] * coefficients[i]) == rhs`
fixed_features: A map {feature_index: value} for features that should
be fixed to a particular value during generation.
rounding_func: A function that rounds an optimization result
appropriately (i.e., according to `round-trip` transformations).
Returns:
2-element tuple containing
- A `n x d`-dim tensor of generated candidates.
- In the case of joint optimization, a scalar tensor containing
the joint acquisition value of the `n` points. In the case of
sequential optimization, a `n`-dim tensor of conditional acquisition
values, where `i`-th element is the expected acquisition value
conditional on having observed candidates `0,1,...,i-1`.
"""
sequential = not joint_optimization
optimize_acqf_options: Dict[str, Union[bool, float, int, str]] = {
"batch_limit": 5,
"init_batch_limit": 32,
}
if options is not None:
optimize_acqf_options.update(options)
X, expected_acquisition_value = optimize_acqf(
acq_function=acq_function,
bounds=bounds,
q=n,
num_restarts=num_restarts,
raw_samples=50 * num_restarts if raw_samples is None else raw_samples,
options=optimize_acqf_options,
inequality_constraints=inequality_constraints,
equality_constraints=equality_constraints,
fixed_features=fixed_features,
sequential=sequential,
post_processing_func=rounding_func,
)
return X, expected_acquisition_value
[docs]def recommend_best_observed_point(
model: TorchModel,
bounds: List[Tuple[float, float]],
objective_weights: Tensor,
outcome_constraints: Optional[Tuple[Tensor, Tensor]] = None,
linear_constraints: Optional[Tuple[Tensor, Tensor]] = None,
fixed_features: Optional[Dict[int, float]] = None,
model_gen_options: Optional[TConfig] = None,
target_fidelities: Optional[Dict[int, float]] = None,
) -> Optional[Tensor]:
"""
A wrapper around `ax.models.model_utils.best_observed_point` for TorchModel
that recommends a best point from previously observed points using either a
"max_utility" or "feasible_threshold" strategy.
Args:
model: A TorchModel.
bounds: A list of (lower, upper) tuples for each column of X.
objective_weights: The objective is to maximize a weighted sum of
the columns of f(x). These are the weights.
outcome_constraints: A tuple of (A, b). For k outcome constraints
and m outputs at f(x), A is (k x m) and b is (k x 1) such that
A f(x) <= b.
linear_constraints: A tuple of (A, b). For k linear constraints on
d-dimensional x, A is (k x d) and b is (k x 1) such that
A x <= b.
fixed_features: A map {feature_index: value} for features that
should be fixed to a particular value in the best point.
model_gen_options: A config dictionary that can contain
model-specific options. See `TorchOptConfig` for details.
target_fidelities: A map {feature_index: value} of fidelity feature
column indices to their respective target fidelities. Used for
multi-fidelity optimization.
Returns:
A d-array of the best point, or None if no feasible point was observed.
"""
if target_fidelities:
raise NotImplementedError(
"target_fidelities not implemented for base BotorchModel"
)
x_best = best_observed_point(
model=model,
bounds=bounds,
objective_weights=objective_weights,
outcome_constraints=outcome_constraints,
linear_constraints=linear_constraints,
fixed_features=fixed_features,
options=model_gen_options,
)
if x_best is None:
return None
return x_best.to(dtype=model.dtype, device=torch.device("cpu"))
[docs]def recommend_best_out_of_sample_point(
model: TorchModel,
bounds: List[Tuple[float, float]],
objective_weights: Tensor,
outcome_constraints: Optional[Tuple[Tensor, Tensor]] = None,
linear_constraints: Optional[Tuple[Tensor, Tensor]] = None,
fixed_features: Optional[Dict[int, float]] = None,
model_gen_options: Optional[TConfig] = None,
target_fidelities: Optional[Dict[int, float]] = None,
) -> Optional[Tensor]:
"""
Identify the current best point by optimizing the posterior mean of the model.
This is "out-of-sample" because it considers un-observed designs as well.
Return None if no such point can be identified.
Args:
model: A TorchModel.
bounds: A list of (lower, upper) tuples for each column of X.
objective_weights: The objective is to maximize a weighted sum of
the columns of f(x). These are the weights.
outcome_constraints: A tuple of (A, b). For k outcome constraints
and m outputs at f(x), A is (k x m) and b is (k x 1) such that
A f(x) <= b.
linear_constraints: A tuple of (A, b). For k linear constraints on
d-dimensional x, A is (k x d) and b is (k x 1) such that
A x <= b.
fixed_features: A map {feature_index: value} for features that
should be fixed to a particular value in the best point.
model_gen_options: A config dictionary that can contain
model-specific options. See `TorchOptConfig` for details.
target_fidelities: A map {feature_index: value} of fidelity feature
column indices to their respective target fidelities. Used for
multi-fidelity optimization.
Returns:
A d-array of the best point, or None if no feasible point exists.
"""
options = model_gen_options or {}
fixed_features = fixed_features or {}
acf_options = options.get("acquisition_function_kwargs", {})
optimizer_options = options.get("optimizer_kwargs", {})
X_observed = get_observed(
Xs=model.Xs, # pyre-ignore: [16]
objective_weights=objective_weights,
outcome_constraints=outcome_constraints,
)
if hasattr(model, "_get_best_point_acqf"):
acq_function, non_fixed_idcs = model._get_best_point_acqf( # pyre-ignore: [16]
X_observed=X_observed,
objective_weights=objective_weights,
mc_samples=acf_options.get("mc_samples", 512),
fixed_features=fixed_features,
target_fidelities=target_fidelities,
outcome_constraints=outcome_constraints,
seed_inner=acf_options.get("seed_inner", None),
qmc=acf_options.get("qmc", True),
)
else:
raise RuntimeError("The model should implement _get_best_point_acqf.")
inequality_constraints = _to_inequality_constraints(linear_constraints)
# TODO: update optimizers to handle inequality_constraints
# (including transforming constraints b/c of fixed features)
if inequality_constraints is not None:
raise UnsupportedError("Inequality constraints are not supported!")
return_best_only = optimizer_options.get("return_best_only", True)
bounds_ = torch.tensor(bounds, dtype=model.dtype, device=model.device)
bounds_ = bounds_.transpose(-1, -2)
if non_fixed_idcs is not None:
bounds_ = bounds_[..., non_fixed_idcs]
opt_options: Dict[str, Union[bool, float, int, str]] = {
"batch_limit": 8,
"maxiter": 200,
"method": "L-BFGS-B",
"nonnegative": False,
}
opt_options.update(optimizer_options.get("options", {}))
candidates, _ = optimize_acqf(
acq_function=acq_function,
bounds=bounds_,
q=1,
num_restarts=optimizer_options.get("num_restarts", 60),
raw_samples=optimizer_options.get("raw_samples", 1024),
inequality_constraints=inequality_constraints,
fixed_features=None, # handled inside the acquisition function
options=opt_options,
return_best_only=return_best_only,
)
rec_point = candidates.detach().cpu()
if isinstance(acq_function, FixedFeatureAcquisitionFunction):
rec_point = acq_function._construct_X_full(rec_point)
if return_best_only:
rec_point = rec_point.view(-1)
return rec_point
def _get_model(
X: Tensor,
Y: Tensor,
Yvar: Tensor,
task_feature: Optional[int] = None,
fidelity_features: Optional[List[int]] = None,
use_input_warping: bool = False,
covar_module: Optional[Kernel] = None,
prior: Optional[Dict[str, Any]] = None,
**kwargs: Any,
) -> GPyTorchModel:
"""Instantiate a model of type depending on the input data.
Args:
X: A `n x d` tensor of input features.
Y: A `n x m` tensor of input observations.
Yvar: A `n x m` tensor of input variances (NaN if unobserved).
task_feature: The index of the column pertaining to the task feature
(if present).
fidelity_features: List of columns of X that are fidelity parameters.
covar_module: Optional. A data kernel of GP model.
prior: Optional[Dict]. A dictionary that contains the specification of
GP model prior. Currently, the keys include:
- covar_module_prior: prior on covariance matrix e.g.
{"lengthscale_prior": GammaPrior(3.0, 6.0)}.
- type: type of prior on task covariance matrix e.g.`LKJCovariancePrior`.
- sd_prior: A scalar prior over nonnegative numbers, which is used for the
default LKJCovariancePrior task_covar_prior.
- eta: The eta parameter on the default LKJ task_covar_prior.
Returns:
A GPyTorchModel (unfitted).
"""
Yvar = Yvar.clamp_min(MIN_OBSERVED_NOISE_LEVEL)
is_nan = torch.isnan(Yvar)
any_nan_Yvar = torch.any(is_nan)
all_nan_Yvar = torch.all(is_nan)
if any_nan_Yvar and not all_nan_Yvar:
if task_feature:
# TODO (jej): Replace with inferred noise before making perf judgements.
Yvar[Yvar != Yvar] = MIN_OBSERVED_NOISE_LEVEL
else:
raise ValueError(
"Mix of known and unknown variances indicates valuation function "
"errors. Variances should all be specified, or none should be."
)
if use_input_warping:
if Y.shape[-1] > 1 and X.ndim > 2:
raise UnsupportedError(
"Input warping is not supported for batched multi output models."
)
warp_tf = get_warping_transform(
d=X.shape[-1],
task_feature=task_feature,
batch_shape=X.shape[:-2],
)
else:
warp_tf = None
if fidelity_features is None:
fidelity_features = []
if len(fidelity_features) == 0:
# only pass linear_truncated arg if there are fidelities
kwargs = {k: v for k, v in kwargs.items() if k != "linear_truncated"}
# construct kernel based on customized prior if covar_module is None
prior_dict = prior or {}
covar_module_prior_dict = prior_dict.pop("covar_module_prior", None)
if (covar_module_prior_dict is not None) and (covar_module is None):
covar_module = _get_customized_covar_module(
covar_module_prior_dict=covar_module_prior_dict,
ard_num_dims=X.shape[-1],
aug_batch_shape=_get_aug_batch_shape(X, Y),
task_feature=task_feature,
)
if len(fidelity_features) > 0:
if task_feature:
raise NotImplementedError(
"multi-task multi-fidelity models not yet available"
)
# at this point we can assume that there is only a single fidelity parameter
gp = SingleTaskMultiFidelityGP(
train_X=X,
train_Y=Y,
data_fidelity=fidelity_features[0],
input_transform=warp_tf,
**kwargs,
)
elif task_feature is None:
gp = SingleTaskGP(
train_X=X,
train_Y=Y,
train_Yvar=None if all_nan_Yvar else Yvar,
covar_module=covar_module,
input_transform=warp_tf,
**kwargs,
)
else:
# instantiate multitask GP
all_tasks, _, _ = MultiTaskGP.get_all_tasks(X, task_feature)
num_tasks = len(all_tasks)
task_covar_prior = None
if len(prior_dict) > 0:
prior_type = prior_dict.get("type", None)
if issubclass(prior_type, LKJCovariancePrior):
sd_prior = prior_dict.get("sd_prior", GammaPrior(1.0, 0.15))
sd_prior._event_shape = torch.Size([num_tasks])
eta = prior_dict.get("eta", 0.5)
if not isinstance(eta, float) and not isinstance(eta, int):
raise ValueError(f"eta must be a real number, your eta was {eta}")
task_covar_prior = LKJCovariancePrior(num_tasks, eta, sd_prior)
else:
raise NotImplementedError(
"Currently only LKJ prior is supported,"
f"your prior type was {prior_type}."
)
gp = MultiTaskGP(
train_X=X,
train_Y=Y,
train_Yvar=None if all_nan_Yvar else Yvar,
task_feature=task_feature,
covar_module=covar_module,
rank=kwargs.get("rank"),
task_covar_prior=task_covar_prior,
input_transform=warp_tf,
)
return gp
def _get_customized_covar_module(
covar_module_prior_dict: Dict[str, Prior],
ard_num_dims: int,
aug_batch_shape: torch.Size,
task_feature: Optional[int] = None,
) -> Kernel:
"""Construct a GP kernel based on customized prior dict.
Args:
covar_module_prior_dict: Dict. The keys are the names of the prior and values
are the priors. e.g. {"lengthscale_prior": GammaPrior(3.0, 6.0)}.
ard_num_dims: The dimension of the inputs, including task features.
aug_batch_shape: The output dimension augmented batch shape of the model
(different from the batch shape for batched multi-output models).
task_feature: The index of the task feature.
"""
# TODO: add more checks of covar_module_prior_dict
if task_feature is not None:
ard_num_dims -= 1
return ScaleKernel(
MaternKernel(
nu=2.5,
ard_num_dims=ard_num_dims,
batch_shape=aug_batch_shape,
lengthscale_prior=covar_module_prior_dict.get(
"lengthscale_prior", GammaPrior(3.0, 6.0)
),
),
batch_shape=aug_batch_shape,
outputscale_prior=covar_module_prior_dict.get(
"outputscale_prior", GammaPrior(2.0, 0.15)
),
)
def _get_aug_batch_shape(X: Tensor, Y: Tensor) -> torch.Size:
"""Obtain the output-augmented batch shape of GP model.
Args:
X: A `(input_batch_shape) x n x d` tensor of input features.
Y: A `n x m` tensor of input observations.
Returns:
The output-augmented batch shape: `input_batch_shape x (m)`
"""
batch_shape = X.shape[:-2]
num_outputs = Y.shape[-1]
if num_outputs > 1:
batch_shape += torch.Size([num_outputs]) # pyre-ignore
return batch_shape # pyre-ignore
