Revision e24fd815cdfb8c654249da4576aeff6c2ce5a8ea authored by vdutor on 10 September 2020, 15:35 UTC, committed by vdutor on 10 September 2020, 15:35 UTC
1 parent 61b2e1c
Raw File
base.py
import functools
from enum import Enum
from typing import Any, List, Optional, Sequence, TYPE_CHECKING, Tuple, Union

import numpy as np
import tensorflow as tf
import tensorflow_probability as tfp
from tensorflow.python.ops import array_ops
from typing_extensions import Final

from .config import default_float, default_summary_fmt

if TYPE_CHECKING:
    from IPython.lib import pretty

DType = Union[np.dtype, tf.DType]
VariableData = Union[List, Tuple, np.ndarray, int, float]  # deprecated
Transform = Union[tfp.bijectors.Bijector]
Prior = Union[tfp.distributions.Distribution]


TensorType = Union[tf.Tensor, tf.Variable, "Parameter"]
"""
Type alias for tensor-like types that are supported by most TensorFlow and GPflow operations.

NOTE: Union types like this do not work with the `register` method of `multipledispatch`'s
`Dispatcher` class. Instead use `TensorLike`.
"""


# We've left this as object until we've tested the performance consequences of using the full set
# (np.ndarray, tf.Tensor, tf.Variable, Parameter), see https://github.com/GPflow/GPflow/issues/1434
TensorLike: Final[Tuple[type, ...]] = (object,)
"""
:var TensorLike: Collection of tensor-like types for registering implementations with
    `multipledispatch` dispatchers.
"""


_NativeScalar = Union[int, float]
_Array = Sequence[Any]  # a nested array of int, float, bool etc. kept simple for readability
TensorData = Union[_NativeScalar, _Array, TensorType]


def _IS_PARAMETER(o: object) -> bool:
    return isinstance(o, Parameter)


def _IS_TRAINABLE_PARAMETER(o: object) -> bool:
    return isinstance(o, Parameter) and o.trainable


class Module(tf.Module):
    @property
    def parameters(self) -> Tuple["Parameter", ...]:
        return tuple(self._flatten(predicate=_IS_PARAMETER))

    @property
    def trainable_parameters(self) -> Tuple["Parameter", ...]:
        return tuple(self._flatten(predicate=_IS_TRAINABLE_PARAMETER))

    def _representation_table(self, object_name: str, tablefmt: Optional[str]) -> str:
        from .utilities import leaf_components, tabulate_module_summary

        repr_components = [object_name]
        if leaf_components(self):
            repr_components.append(tabulate_module_summary(self, tablefmt=tablefmt))
        return "\n".join(repr_components)

    def _repr_html_(self) -> str:
        """ Nice representation of GPflow objects in IPython/Jupyter notebooks """
        from html import escape

        return self._representation_table(escape(repr(self)), "html")

    def _repr_pretty_(self, p: "pretty.RepresentationPrinter", cycle: bool) -> None:
        """ Nice representation of GPflow objects in the IPython shell """
        repr_str = self._representation_table(repr(self), default_summary_fmt())
        p.text(repr_str)


class PriorOn(Enum):
    CONSTRAINED = "constrained"
    UNCONSTRAINED = "unconstrained"


class Parameter(tfp.util.TransformedVariable):
    def __init__(
        self,
        value: TensorData,
        *,
        transform: Optional[Transform] = None,
        prior: Optional[Prior] = None,
        prior_on: Union[str, PriorOn] = PriorOn.CONSTRAINED,
        trainable: bool = True,
        dtype: Optional[DType] = None,
        name: Optional[str] = None,
    ):
        """
        A parameter retains both constrained and unconstrained
        representations. If no transform is provided, these two values will be the same.
        It is often challenging to operate with unconstrained parameters. For example, a variance cannot be negative,
        therefore we need a positive constraint and it is natural to use constrained values.
        A prior can be imposed either on the constrained version (default) or on the unconstrained version of the parameter.
        """
        if transform is None:
            transform = tfp.bijectors.Identity()

        value = _cast_to_dtype(value, dtype)
        _validate_unconstrained_value(value, transform, dtype)
        super().__init__(value, transform, dtype=value.dtype, trainable=trainable, name=name)

        self.prior = prior
        self.prior_on = prior_on  # type: ignore  # see https://github.com/python/mypy/issues/3004

    def log_prior_density(self) -> tf.Tensor:
        """ Log of the prior probability density of the constrained variable. """

        if self.prior is None:
            return tf.convert_to_tensor(0.0, dtype=self.dtype)

        y = self

        if self.prior_on == PriorOn.CONSTRAINED:
            # evaluation is in same space as prior
            return tf.reduce_sum(self.prior.log_prob(y))

        else:
            # prior on unconstrained, but evaluating log-prior in constrained space
            x = self.unconstrained_variable
            log_p = tf.reduce_sum(self.prior.log_prob(x))

            if self.transform is not None:
                # need to include log|Jacobian| to account for coordinate transform
                log_det_jacobian = self.transform.inverse_log_det_jacobian(y, y.shape.ndims)
                log_p += tf.reduce_sum(log_det_jacobian)

            return log_p

    @property
    def prior_on(self) -> PriorOn:
        return self._prior_on

    @prior_on.setter
    def prior_on(self, value: Union[str, PriorOn]) -> None:
        self._prior_on = PriorOn(value)

    @property
    def unconstrained_variable(self) -> tf.Variable:
        return self._pretransformed_input

    @property
    def transform(self) -> Optional[Transform]:
        return self.bijector

    @property
    def trainable(self) -> bool:
        """
        `True` if this instance is trainable, else `False`.

        This attribute cannot be set directly. Use :func:`gpflow.set_trainable`.
        """
        return self.unconstrained_variable.trainable

    def assign(
        self,
        value: TensorData,
        use_locking: bool = False,
        name: Optional[str] = None,
        read_value: bool = True,
    ) -> tf.Tensor:
        """
        Assigns constrained `value` to the unconstrained parameter's variable.
        It passes constrained value through parameter's transform first.

        Example:
            ```
            a = Parameter(2.0, transform=tfp.bijectors.Softplus())
            b = Parameter(3.0)

            a.assign(4.0)               # `a` parameter to `2.0` value.
            a.assign(tf.constant(5.0))  # `a` parameter to `5.0` value.
            a.assign(b)                 # `a` parameter to constrained value of `b`.
            ```

        :param value: Constrained tensor-like value.
        :param use_locking: If `True`, use locking during the assignment.
        :param name: The name of the operation to be created.
        :param read_value: if True, will return something which evaluates to the new
            value of the variable; if False will return the assign op.
        """
        unconstrained_value = _validate_unconstrained_value(value, self.transform, self.dtype)
        return self.unconstrained_variable.assign(
            unconstrained_value, use_locking=use_locking, name=name, read_value=read_value
        )


def _cast_to_dtype(
    value: TensorData, dtype: Optional[DType] = None
) -> Union[tf.Tensor, tf.Variable]:
    if dtype is None:
        dtype = default_float()

    if tf.is_tensor(value):
        # NOTE(awav) TF2.2 resolves issue with cast.
        # From TF2.2, `tf.cast` can be used alone instead of this auxiliary function.
        # workaround for https://github.com/tensorflow/tensorflow/issues/35938
        return tf.cast(value, dtype)
    else:
        return tf.convert_to_tensor(value, dtype=dtype)


def _validate_unconstrained_value(
    value: TensorData, transform: tfp.bijectors.Bijector, dtype: DType
) -> tf.Tensor:
    value = _cast_to_dtype(value, dtype)
    unconstrained_value = _to_unconstrained(value, transform)
    message = (
        "gpflow.Parameter: the value to be assigned is incompatible with this parameter's "
        "transform (the corresponding unconstrained value has NaN or Inf) and hence cannot be "
        "assigned."
    )
    return tf.debugging.assert_all_finite(unconstrained_value, message=message)


def _to_constrained(value: TensorType, transform: Optional[Transform]) -> TensorType:
    if transform is not None:
        return transform.forward(value)
    return value


def _to_unconstrained(value: TensorType, transform: Optional[Transform]) -> TensorType:
    if transform is not None:
        return transform.inverse(value)
    return value
back to top