Add Fantasy Strategy for Variational GPs

examples/08_Advanced_Usage/index.rst

@@ -59,6 +59,14 @@ See the `1D derivatives GP example`_ or the `2D derivatives GP example`_ for exa
   Simple_Batch_Mode_GP_Regression.ipynb:
 
 
+Variational Fantasization
+----------------------------------
+We also include an example of how to perform fantasy modelling (e.g. efficient, closed form updates) for variational
+Gaussian process models, enabling their usage for lookahead optimization.
+
+.. _Variational fantasization:
+  SVGP_Model_Updating.ipynb
+
 Converting Models to TorchScript
 ----------------------------------
 
@@ -73,3 +81,4 @@ how to convert both an exact GP and a variational GP to a ScriptModule that can
 
    TorchScript_Exact_Models.ipynb
    TorchScript_Variational_Models.ipynb
+   SVGP_Model_Updating.ipynb

gpytorch/models/approximate_gp.py

@@ -75,6 +75,33 @@ def pyro_model(self, input, beta=1.0, name_prefix=""):
         """
         return super().pyro_model(input, beta=beta, name_prefix=name_prefix)
 
+    def get_fantasy_model(self, inputs, targets, **kwargs):
+        r"""
+        Returns a new GP model that incorporates the specified inputs and targets as new training data using
+        online variational conditioning (OVC).
+
+        This function first casts the inducing points and variational parameters into pseudo-points before
+        returning an equivalent ExactGP model with a specialized likelihood.
+
+        .. note::
+            If `targets` is a batch (e.g. `b x m`), then the GP returned from this method will be a batch mode GP.
+            If `inputs` is of the same (or lesser) dimension as `targets`, then it is assumed that the fantasy points
+            are the same for each target batch.
+
+        :param torch.Tensor inputs: (`b1 x ... x bk x m x d` or `f x b1 x ... x bk x m x d`) Locations of fantasy
+            observations.
+        :param torch.Tensor targets: (`b1 x ... x bk x m` or `f x b1 x ... x bk x m`) Labels of fantasy observations.
+        :return: An `ExactGP` model with `n + m` training examples, where the `m` fantasy examples have been added
+            and all test-time caches have been updated.
+        :rtype: ~gpytorch.models.ExactGP
+
+        Reference: "Conditioning Sparse Variational Gaussian Processes for Online Decision-Making,"
+            Maddox, Stanton, Wilson, NeurIPS, '21
+            https://papers.nips.cc/paper/2021/hash/325eaeac5bef34937cfdc1bd73034d17-Abstract.html
+
+        """
+        return self.variational_strategy.get_fantasy_model(inputs=inputs, targets=targets, **kwargs)
+
     def __call__(self, inputs, prior=False, **kwargs):
         if inputs.dim() == 1:
             inputs = inputs.unsqueeze(-1)

gpytorch/test/variational_test_case.py

@@ -95,6 +95,28 @@ def _eval_iter(self, model, batch_shape=torch.Size([]), cuda=False):
 
         return output
 
+    def _fantasy_iter(
+        self,
+        model,
+        likelihood,
+        batch_shape=torch.Size([]),
+        cuda=False,
+        num_fant=10,
+        covar_module=None,
+        mean_module=None,
+    ):
+        model.likelihood = likelihood
+        val_x = torch.randn(*batch_shape, num_fant, 2).clamp(-2.5, 2.5)
+        val_y = torch.linspace(-1, 1, num_fant)
+        val_y = val_y.view(num_fant, *([1] * (len(self.event_shape) - 1)))
+        val_y = val_y.expand(*batch_shape, num_fant, *self.event_shape[1:])
+        if cuda:
+            model = model.cuda()
+            val_x = val_x.cuda()
+            val_y = val_y.cuda()
+        updated_model = model.get_fantasy_model(val_x, val_y, covar_module=covar_module, mean_module=mean_module)
+        return updated_model
+
     @abstractproperty
     def batch_shape(self):
         raise NotImplementedError
@@ -272,3 +294,102 @@ def test_training_all_batch_zero_mean(self):
             expected_batch_shape=(torch.Size([3, 4]) + self.batch_shape),
             constant_mean=False,
         )
+
+    def test_fantasy_call(
+        self,
+        data_batch_shape=None,
+        inducing_batch_shape=None,
+        model_batch_shape=None,
+        expected_batch_shape=None,
+        constant_mean=True,
+    ):
+        # Batch shapes
+        model_batch_shape = model_batch_shape if model_batch_shape is not None else self.batch_shape
+        data_batch_shape = data_batch_shape if data_batch_shape is not None else self.batch_shape
+        inducing_batch_shape = inducing_batch_shape if inducing_batch_shape is not None else self.batch_shape
+        expected_batch_shape = expected_batch_shape if expected_batch_shape is not None else self.batch_shape
+
+        num_inducing = 16
+        num_fant = 10
+        # Make model and likelihood
+        model, likelihood = self._make_model_and_likelihood(
+            batch_shape=model_batch_shape,
+            inducing_batch_shape=inducing_batch_shape,
+            distribution_cls=self.distribution_cls,
+            strategy_cls=self.strategy_cls,
+            constant_mean=constant_mean,
+            num_inducing=num_inducing,
+        )
+
+        # we iterate through the covar and mean module possible settings
+        covar_mean_options = [
+            {"covar_module": None, "mean_module": None},
+            {"covar_module": gpytorch.kernels.MaternKernel(), "mean_module": gpytorch.means.ZeroMean()},
+        ]
+        for cm_dict in covar_mean_options:
+            fant_model = self._fantasy_iter(
+                model, likelihood, data_batch_shape, self.cuda, num_fant=num_fant, **cm_dict
+            )
+            self.assertTrue(isinstance(fant_model, gpytorch.models.ExactGP))
+
+            # we check to ensure setting the covar_module and mean_modules are okay
+            if cm_dict["covar_module"] is None:
+                self.assertEqual(type(fant_model.covar_module), type(model.covar_module))
+            else:
+                self.assertNotEqual(type(fant_model.covar_module), type(model.covar_module))
+            if cm_dict["mean_module"] is None:
+                self.assertEqual(type(fant_model.mean_module), type(model.mean_module))
+            else:
+                self.assertNotEqual(type(fant_model.mean_module), type(model.mean_module))
+
+            # now we check to ensure the shapes of the fantasy strategy are correct
+            self.assertTrue(fant_model.prediction_strategy is not None)
+            for key in fant_model.prediction_strategy._memoize_cache.keys():
+                if key[0] == "mean_cache":
+                    break
+            mean_cache = fant_model.prediction_strategy._memoize_cache[key]
+            self.assertEqual(mean_cache.shape, torch.Size([*expected_batch_shape, num_inducing + num_fant]))
+
+        # we remove the mean_module and covar_module and check for errors
+        del model.mean_module
+        with self.assertRaises(ModuleNotFoundError):
+            self._fantasy_iter(model, likelihood, data_batch_shape, self.cuda, num_fant=num_fant)
+
+        model.mean_module = gpytorch.means.ZeroMean()
+        del model.covar_module
+        with self.assertRaises(ModuleNotFoundError):
+            self._fantasy_iter(model, likelihood, data_batch_shape, self.cuda, num_fant=num_fant)
+
+        # finally we check to ensure failure for a non-gaussian likelihood
+        with self.assertRaises(NotImplementedError):
+            self._fantasy_iter(
+                model,
+                gpytorch.likelihoods.BernoulliLikelihood(),
+                data_batch_shape,
+                self.cuda,
+                num_fant=num_fant,
+            )
+
+    def test_fantasy_call_batch_inducing(self):
+        return self.test_fantasy_call(
+            model_batch_shape=(torch.Size([3]) + self.batch_shape),
+            data_batch_shape=self.batch_shape,
+            inducing_batch_shape=(torch.Size([3]) + self.batch_shape),
+            expected_batch_shape=(torch.Size([3]) + self.batch_shape),
+        )
+
+    def test_fantasy_call_batch_data(self):
+        return self.test_fantasy_call(
+            model_batch_shape=self.batch_shape,
+            inducing_batch_shape=self.batch_shape,
+            data_batch_shape=(torch.Size([3]) + self.batch_shape),
+            expected_batch_shape=(torch.Size([3]) + self.batch_shape),
+        )
+
+    def test_fantasy_call_batch_model(self):
+        return self.test_fantasy_call(
+            model_batch_shape=(torch.Size([3]) + self.batch_shape),
+            inducing_batch_shape=self.batch_shape,
+            data_batch_shape=self.batch_shape,
+            expected_batch_shape=(torch.Size([3]) + self.batch_shape),
+        )

gpytorch/variational/_variational_strategy.py

@@ -1,21 +1,47 @@
 #!/usr/bin/env python3
 
+import functools
 from abc import ABC, abstractproperty
+from copy import deepcopy
 
 import torch
 
 from .. import settings
 from ..distributions import Delta, MultivariateNormal
+from ..likelihoods import GaussianLikelihood
+from ..models import ExactGP
 from ..module import Module
 from ..utils.broadcasting import _mul_broadcast_shape
-from ..utils.memoize import cached, clear_cache_hook
+from ..utils.memoize import add_to_cache, cached, clear_cache_hook
+
+
+class _BaseExactGP(ExactGP):
+    def __init__(self, train_inputs, train_targets, likelihood, mean_module, covar_module):
+        super().__init__(train_inputs, train_targets, likelihood)
+        self.mean_module = mean_module
+        self.covar_module = covar_module
+
+    def forward(self, x):
+        mean = self.mean_module(x)
+        covar = self.covar_module(x)
+        return MultivariateNormal(mean, covar)
+
+
+def _add_cache_hook(tsr, pred_strat):
+    if tsr.grad_fn is not None:
+        wrapper = functools.partial(clear_cache_hook, pred_strat)
+        functools.update_wrapper(wrapper, clear_cache_hook)
+        tsr.grad_fn.register_hook(wrapper)
+    return tsr
 
 
 class _VariationalStrategy(Module, ABC):
     """
     Abstract base class for all Variational Strategies.
     """
 
+    has_fantasy_strategy = False
+
     def __init__(self, model, inducing_points, variational_distribution, learn_inducing_locations=True):
         super().__init__()
 
@@ -97,6 +123,144 @@ def kl_divergence(self):
             kl_divergence = torch.distributions.kl.kl_divergence(self.variational_distribution, self.prior_distribution)
         return kl_divergence
 
+    @cached(name="amortized_exact_gp")
+    def amortized_exact_gp(self, mean_module=None, covar_module=None):
+        mean_module = self.model.mean_module if mean_module is None else mean_module
+        covar_module = self.model.covar_module if covar_module is None else covar_module
+
+        with torch.no_grad():
+            # from here on down, we refer to the inducing points as pseudo_inputs
+            pseudo_target_covar, pseudo_target_mean = self.pseudo_points
+            pseudo_inputs = self.inducing_points.detach()
+            if pseudo_inputs.ndim < pseudo_target_mean.ndim:
+                pseudo_inputs = pseudo_inputs.expand(*pseudo_target_mean.shape[:-2], *pseudo_inputs.shape)
+            # TODO: add flag for conditioning into SGPR after building fantasy strategy for SGPR
+            new_covar_module = deepcopy(covar_module)
+
+            # update inducing mean if necessary
+            pseudo_target_mean = pseudo_target_mean.squeeze() + mean_module(pseudo_inputs)
+
+            inducing_exact_model = _BaseExactGP(
+                pseudo_inputs,
+                pseudo_target_mean,
+                mean_module=deepcopy(mean_module),
+                covar_module=new_covar_module,
+                likelihood=deepcopy(self.model.likelihood),
+            )
+
+            # now fantasize around this model
+            # as this model is new, we need to compute a posterior to construct the prediction strategy
+            # which uses the likelihood pseudo caches
+            faked_points = torch.randn(
+                *pseudo_target_mean.shape[:-2],
+                1,
+                pseudo_inputs.shape[-1],
+                device=pseudo_inputs.device,
+                dtype=pseudo_inputs.dtype,
+            )
+            inducing_exact_model.eval()
+            _ = inducing_exact_model(faked_points)
+
+            # then we overwrite the likelihood to take into account the multivariate normal term
+            pred_strat = inducing_exact_model.prediction_strategy
+            pred_strat._memoize_cache = {}
+            with torch.no_grad():
+                updated_lik_train_train_covar = pred_strat.train_prior_dist.lazy_covariance_matrix + pseudo_target_covar
+                pred_strat.lik_train_train_covar = updated_lik_train_train_covar
+
+            # do the mean cache because the mean cache doesn't solve against lik_train_train_covar
+            train_mean = inducing_exact_model.mean_module(*inducing_exact_model.train_inputs)
+            train_labels_offset = (inducing_exact_model.prediction_strategy.train_labels - train_mean).unsqueeze(-1)
+            mean_cache = updated_lik_train_train_covar.inv_matmul(train_labels_offset).squeeze(-1)
+            mean_cache = _add_cache_hook(mean_cache, inducing_exact_model.prediction_strategy)
+            add_to_cache(pred_strat, "mean_cache", mean_cache)
+            # TODO: check to see if we need to do the covar_cache?
+
+            inducing_exact_model.prediction_strategy = pred_strat
+        return inducing_exact_model
+
+    def pseudo_points(self):
+        raise NotImplementedError("Each variational strategy must implement its own pseudo points method")
+
+    def get_fantasy_model(
+        self,
+        inputs,
+        targets,
+        mean_module=None,
+        covar_module=None,
+        **kwargs,
+    ):
+        r"""
+        Performs the online variational conditioning (OVC) strategy of Maddox et al, '21 to return
+        an exact GP model that incorporates the inputs and targets alongside the variational model's inducing
+        points and targets.
+
+        Currently, instead of directly updating the variational parameters (and inducing points), we instead
+        return an ExactGP model rather than an updated variational GP model. This is done primarily for
+        numerical stability.
+
+        Unlike the ExactGP's call for get_fantasy_model, we enable options for mean_module and covar_module
+        that allow specification of the mean / covariance.
+
+        Reference: "Conditioning Sparse Variational Gaussian Processes for Online Decision-Making,"
+            Maddox, Stanton, Wilson, NeurIPS, '21
+            https://papers.nips.cc/paper/2021/hash/325eaeac5bef34937cfdc1bd73034d17-Abstract.html
+        """
+
+        # currently, we only support fantasization for CholeskyVariationalDistribution and
+        # whitened / unwhitened variational strategies
+        if not self.has_fantasy_strategy:
+            raise NotImplementedError(
+                "No fantasy model support for ",
+                self.__name__,
+                ". Only VariationalStrategy and UnwhitenedVariationalStrategy are currently supported.",
+            )
+        if not isinstance(self.model.likelihood, GaussianLikelihood):
+            raise NotImplementedError(
+                "No fantasy model support for ",
+                self.model.likelihood,
+                ". Only GaussianLikelihoods are currently supported.",
+            )
+        # we assume that either the user has given the model a mean_module and a covar_module
+        # or that it will be passed into the get_fantasy_model function. we check for these.
+        if mean_module is None:
+            mean_module = getattr(self.model, "mean_module", None)
+            if mean_module is None:
+                raise ModuleNotFoundError(
+                    "Either you must provide a mean_module as input to get_fantasy_model",
+                    "or it must be an attribute of the model.",
+                )
+        if covar_module is None:
+            covar_module = getattr(self.model, "covar_module", None)
+            if covar_module is None:
+                # raise an error
+                raise ModuleNotFoundError(
+                    "Either you must provide a covar_module as input to get_fantasy_model",
+                    "or it must be an attribute of the model.",
+                )
+
+        # first we construct an exact model over the inducing points with the inducing covariance
+        # matrix
+        inducing_exact_model = self.amortized_exact_gp(mean_module=mean_module, covar_module=covar_module)
+
+        # then we update this model by adding in the inputs and pseudo targets
+        # finally we fantasize wrt targets
+        fantasy_model = inducing_exact_model.get_fantasy_model(inputs, targets, **kwargs)
+        fant_pred_strat = fantasy_model.prediction_strategy
+
+        # first we update the lik_train_train_covar
+        # do the mean cache again because the mean cache resets the likelihood forward
+        train_mean = fantasy_model.mean_module(*fantasy_model.train_inputs)
+        train_labels_offset = (fant_pred_strat.train_labels - train_mean).unsqueeze(-1)
+        fantasy_lik_train_root_inv = fant_pred_strat.lik_train_train_covar.root_inv_decomposition()
+        mean_cache = fantasy_lik_train_root_inv.matmul(train_labels_offset).squeeze(-1)
+        mean_cache = _add_cache_hook(mean_cache, fant_pred_strat)
+        add_to_cache(fant_pred_strat, "mean_cache", mean_cache)
+        # TODO: should we update the covar_cache?
+
+        fantasy_model.prediction_strategy = fant_pred_strat
+        return fantasy_model
+
     def __call__(self, x, prior=False, **kwargs):
         # If we're in prior mode, then we're done!
         if prior: