https://github.com/GPflow/GPflow
Tip revision: ff13e612910cfcb6b83f2e8d4cc8627888bffe92 authored by Sergio Diaz on 21 March 2019, 13:41:09 UTC
fixing imports
fixing imports
Tip revision: ff13e61
util.py
import copy
import logging
from typing import List, Union, Callable, Tuple
import numpy as np
import tensorflow as tf
NoneType = type(None)
def create_logger(name=None):
return logging.getLogger('Temporary Logger Solution')
def default_jitter_eye(num_rows: int, num_columns: int = None, value: float = None) -> float:
value = default_jitter() if value is None else value
num_rows = int(num_rows)
num_columns = int(num_columns) if num_columns is not None else num_columns
return tf.eye(num_rows, num_columns=num_columns, dtype=default_float()) * value
def default_jitter() -> float:
return 1e-6
def default_float() -> float:
return np.float64
def default_int() -> int:
return np.int32
def leading_transpose(tensor: tf.Tensor, perm: List[Union[int, type(...)]]) -> tf.Tensor:
"""
Transposes tensors with leading dimensions. Leading dimensions in
permutation list represented via ellipsis `...`.
When leading dimensions are found, `transpose` method
considers them as a single grouped element indexed by 0 in `perm` list. So, passing
`perm=[-2, ..., -1]`, you assume that your input tensor has [..., A, B] shape,
and you want to move leading dims between A and B dimensions.
Dimension indices in permutation list can be negative or positive. Valid positive
indices start from 1 up to the tensor rank, viewing leading dimensions `...` as zero
index.
Example:
a = tf.random.normal((1, 2, 3, 4, 5, 6))
b = leading_transpose(a, [5, -3, ..., -2])
sess.run(b).shape
output> (6, 4, 1, 2, 3, 5)
:param tensor: TensorFlow tensor.
:param perm: List of permutation indices.
:returns: TensorFlow tensor.
:raises: ValueError when `...` cannot be found.
"""
perm = copy.copy(perm)
idx = perm.index(...)
perm[idx] = 0
rank = tf.rank(tensor)
perm_tf = perm % rank
leading_dims = tf.range(rank - len(perm) + 1)
perm = tf.concat([perm_tf[:idx], leading_dims, perm_tf[idx + 1:]], 0)
return tf.transpose(tensor, perm)
def set_trainable(model: tf.Module, flag: bool = False):
for variable in model.trainable_variables:
variable._trainable = flag
def training_loop(closure: Callable[..., tf.Tensor],
optimizer=tf.optimizers.Adam(),
var_list: List[tf.Variable] = None,
jit=True,
maxiter=1e3):
"""
Simple generic training loop. At each iteration uses a GradientTape to compute
the gradients of a loss function with respect to a set of variables.
:param closure: Callable that constructs a loss function based on data and model being trained
:param optimizer: tf.optimizers or tf.keras.optimizers that updates variables by applying the
corresponding loss gradients
:param var_list: List of model variables to be learnt during training
:param maxiter: Maximum number of
:return:
"""
def optimization_step():
with tf.GradientTape() as tape:
tape.watch(var_list)
loss = closure()
grads = tape.gradient(loss, var_list)
optimizer.apply_gradients(zip(grads, var_list))
if jit:
optimization_step = tf.function(optimization_step)
for _ in range(int(maxiter)):
optimization_step()