tf.keras.mixed_precision.experimental.LossScaleOptimizer

TensorFlow 1 version View source on GitHub

An optimizer that applies loss scaling.

Inherits From: Optimizer

tf.keras.mixed_precision.experimental.LossScaleOptimizer(
    optimizer, loss_scale
)

Used in the notebooks

Used in the guide

Loss scaling is a process that multiplies the loss by a multiplier called the loss scale, and divides each gradient by the same multiplier. The pseudocode for this process is:

loss = ...
loss *= loss_scale
grads = gradients(loss, vars)
grads /= loss_scale

Mathematically, loss scaling has no effect, but can help avoid numerical underflow in intermediate gradients when float16 tensors are used. By multiplying the loss, each intermediate gradient will have the same multiplier applied.

The loss scale can either be a fixed constant, chosen by the user, or be dynamically determined. Dynamically determining the loss scale is convenient as a loss scale does not have to be explicitly chosen. However it reduces performance.

This optimizer wraps another optimizer and applies loss scaling to it via a LossScale. Loss scaling is applied whenever gradients are computed, either through minimize() or get_gradients(). The loss scale is updated via LossScale.update() whenever gradients are applied, either through minimize() or apply_gradients(). For example:

opt = tf.keras.optimizers.SGD(0.25)
opt = tf.keras.mixed_precision.experimental.LossScaleOptimizer(opt,
                                                               "dynamic")
var = tf.Variable(1.)
loss_fn = lambda: var ** 2
# 'minimize' applies loss scaling to the loss and updates the loss sale.
opt.minimize(loss_fn, var_list=var)
var.numpy()
0.5

If a tf.GradientTape is used to compute gradients instead of LossScaleOptimizer.minimize or LossScaleOptimizer.get_gradients, the loss and gradients must be scaled manually. This can be done by calling LossScaleOptimizer.get_scaled_loss before passing the loss to tf.GradientTape, and LossScaleOptimizer.get_unscaled_gradients after computing the gradients with tf.GradientTape. For example:

with tf.GradientTape() as tape:
  loss = loss_fn()
  scaled_loss = opt.get_scaled_loss(loss)
scaled_grad = tape.gradient(scaled_loss, var)
(grad,) = opt.get_unscaled_gradients([scaled_grad])
opt.apply_gradients([(grad, var)])  # Loss scale is updated here
var.numpy()
0.25

optimizer The Optimizer instance to wrap.
loss_scale The loss scale to scale the loss and gradients. This can either be an int/float to use a fixed loss scale, the string "dynamic" to use dynamic loss scaling, or an instance of a LossScale. The string "dynamic" equivalent to passing DynamicLossScale(), and passing an int/float is equivalent to passing a FixedLossScale with the given loss scale.

iterations Variable. The number of training steps this Optimizer has run.
learning_rate

loss_scale The LossScale instance associated with this optimizer.
lr

weights Returns variables of this Optimizer based on the order created.

Methods

add_slot

View source

add_slot(
    var, slot_name, initializer='zeros'
)

Add a new slot variable for var.

add_weight

View source

add_weight(
    name, shape, dtype=None, initializer='zeros', trainable=None,
    synchronization=tf.VariableSynchronization.AUTO,
    aggregation=tf.compat.v1.VariableAggregation.NONE
)

apply_gradients

View source

apply_gradients(
    grads_and_vars, name=None, experimental_aggregate_gradients=True
)

Apply gradients to variables.

This is the second part of minimize(). It returns an Operation that applies gradients.

The method sums gradients from all replicas in the presence of tf.distribute.Strategy by default. You can aggregate gradients yourself by passing experimental_aggregate_gradients=False.

Example:

grads = tape.gradient(loss, vars)
grads = tf.distribute.get_replica_context().all_reduce('sum', grads)
# Processing aggregated gradients.
optimizer.apply_gradients(zip(grads, vars),
    experimental_aggregate_gradients=False)

Args
grads_and_vars List of (gradient, variable) pairs.
name Optional name for the returned operation. Default to the name passed to the Optimizer constructor.
experimental_aggregate_gradients Whether to sum gradients from different replicas in the presense of tf.distribute.Strategy. If False, it's user responsibility to aggregate the gradients. Default to True.

Returns
An Operation that applies the specified gradients. The iterations will be automatically increased by 1.

Raises
TypeError If grads_and_vars is malformed.
ValueError If none of the variables have gradients.

from_config

View source

@classmethod
from_config(
    config, custom_objects=None
)

Creates an optimizer from its config.

This method is the reverse of get_config, capable of instantiating the same optimizer from the config dictionary.

Arguments
config A Python dictionary, typically the output of get_config.
custom_objects A Python dictionary mapping names to additional Python objects used to create this optimizer, such as a function used for a hyperparameter.

Returns
An optimizer instance.

get_config

View source

get_config()

Returns the config of the optimizer.

An optimizer config is a Python dictionary (serializable) containing the configuration of an optimizer. The same optimizer can be reinstantiated later (without any saved state) from this configuration.

Returns
Python dictionary.

get_gradients

View source

get_gradients(
    loss, params
)

Returns gradients of loss with respect to params.

Arguments
loss Loss tensor.
params List of variables.

Returns
List of gradient tensors.

Raises
ValueError In case any gradient cannot be computed (e.g. if gradient function not implemented).

get_scaled_loss

View source

get_scaled_loss(
    loss
)

Scales the loss by the loss scale.

This method is only needed if you compute gradients manually, e.g. with tf.GradientTape. In that case, call this method to scale the loss before passing the loss to tf.GradientTape. If you use LossScaleOptimizer.minimize or LossScaleOptimizer.get_gradients, loss scaling is automatically applied and this method is unneeded.

If this method is called, get_unscaled_gradients should also be called. See the tf.keras.mixed_precision.experimental.LossScaleOptimizer doc for an example.

Args
loss The loss, which will be multiplied by the loss scale. Can either be a tensor or a callable returning a tensor.

Returns
loss multiplied by LossScaleOptimizer.loss_scale().

get_slot

View source

get_slot(
    var, slot_name
)

get_slot_names

View source

get_slot_names()

A list of names for this optimizer's slots.

get_unscaled_gradients

View source

get_unscaled_gradients(
    grads
)

Unscales the gradients by the loss scale.

This method is only needed if you compute gradients manually, e.g. with tf.GradientTape. In that case, call this method to unscale the gradients after computing them with tf.GradientTape. If you use LossScaleOptimizer.minimize or LossScaleOptimizer.get_gradients, loss scaling is automatically applied and this method is unneeded.

If this method is called, get_scaled_loss should also be called. See the tf.keras.mixed_precision.experimental.LossScaleOptimizer doc for an example.

Args
grads A list of tensors, each which will be divided by the loss scale. Can have None values, which are ignored.

Returns
A new list the same size as grads, where every non-None value in grads is divided by LossScaleOptimizer.loss_scale().

get_updates

View source

get_updates(
    loss, params
)

get_weights

View source

get_weights()

Returns the current weights of the optimizer.

The weights of an optimizer are its state (ie, variables). This function returns the weight values associated with this optimizer as a list of Numpy arrays. The first value is always the iterations count of the optimizer, followed by the optimizer's state variables in the order they were created. The returned list can in turn be used to load state into similarly parameterized optimizers.

For example, the RMSprop optimizer for this simple model returns a list of three values-- the iteration count, followed by the root-mean-square value of the kernel and bias of the single Dense layer:

opt = tf.keras.optimizers.RMSprop()
m = tf.keras.models.Sequential([tf.keras.layers.Dense(10)])
<