# Copyright 2020-2022 Huawei Technologies Co., Ltd
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ============================================================================
"""FTRL"""
from __future__ import absolute_import
from mindspore.ops import functional as F, composite as C, operations as P
from mindspore.ops.composite.multitype_ops.zeros_like_impl import zeros_like
from mindspore.common.api import jit
from mindspore import _checkparam as validator
from mindspore.nn.optim.optimizer import Optimizer
from mindspore.nn.optim.optimizer import opt_init_args_register
from mindspore.nn.optim._dist_optimizer_registry import _register_dist_optimizer
_ftrl_opt = C.MultitypeFuncGraph("ftrl_opt")
@_ftrl_opt.register("Function", "Function", "Number", "Number", "Number", "Tensor", "Tensor",
"RowTensor", "Tensor", "Tensor", "Bool", "Function", "Bool", "Function", "Bool")
def _tensor_run_opt_with_sparse_dist(opt, spars_opt, l1, l2, lr_power, learning_rate, linear,
gradient, weight, moment, cache_enable,
distributed_opt, use_flag, distributed_sparse_opt, use_sparse_flag):
"""Apply sparse ftrl optimizer to the weight parameter when the gradient is sparse."""
success = True
indices = gradient.indices
values = gradient.values
if use_sparse_flag:
success = F.depend(success, distributed_sparse_opt(weight, moment, linear, values, indices))
else:
success = F.depend(success, spars_opt(weight, moment, linear, values, indices))
return success
def _apply_map_tensor_ftrl(l1, l2, lr_power, learning_rate, linear, weight, moment, indices, values):
"""Apllpy ftrl optimizer for map parameter"""
success = True
linear_slice = linear.get(indices)
moment_slice = moment.get(indices)
weight_slice = weight.get(indices)
op_pow = P.Pow()
op_sign = P.Sign()
op_greater = P.Greater()
op_select = P.Select()
op_abs = P.Abs()
lr_power_val = -lr_power
accu_pow = op_pow(moment_slice, lr_power_val)
moment_slice = F.depend(moment_slice, accu_pow)
cur_accu = moment_slice + values * values
cur_accu_pow = op_pow(cur_accu, lr_power_val)
sigma = (cur_accu_pow - accu_pow) / learning_rate
linear_slice = linear_slice + values - sigma * weight_slice
update_weight_cond = op_greater(op_abs(linear_slice), l1)
updated_weight = (l1 * op_sign(linear_slice) - linear_slice) / (cur_accu_pow / learning_rate + 2 * l2)
zeros = zeros_like(weight_slice)
weight_slice = op_select(update_weight_cond, updated_weight, zeros)
moment_slice = cur_accu
moment.put(indices, moment_slice)
linear.put(indices, linear_slice)
weight.put(indices, weight_slice)
return success
@_ftrl_opt.register("Function", "Function", "Number", "Number", "Number", "Tensor", "MapTensor",
"MapTensor", "MapTensor", "MapTensor", "Bool", "Function", "Bool", "Function", "Bool")
def _run_map_tensor_opt_with_sparse_dist(opt, spars_opt, l1, l2, lr_power, learning_rate, linear,
gradient, weight, moment, cache_enable,
distributed_opt, use_flag, distributed_sparse_opt, use_sparse_flag):
"""Apply sparse ftrl optimizer to the weight parameter when the gradient is sparse."""
success = True
indices, values = gradient.get_data()
if use_sparse_flag:
# PS Mode.
success = F.depend(success, distributed_sparse_opt(weight, moment, linear, values, indices))
elif cache_enable:
# PS Cache mode.
_apply_map_tensor_ftrl(l1, l2, lr_power, learning_rate, linear, weight, moment, indices, values)
else:
raise Exception("Unexpected mode for distributed optimizer.")
return success
@_ftrl_opt.register("Function", "Function", "Number", "Number", "Number", "Tensor", "Tensor",
"Tensor", "Tensor", "Tensor", "Bool", "Function", "Bool", "Function", "Bool")
def _tensor_run_opt_dist(opt, spars_opt, l1, l2, lr_power, learning_rate, linear,
gradient, weight, moment, cache_enable,
distributed_opt, use_flag, distributed_sparse_opt, use_sparse_flag):
"""Apply ftrl optimizer to the weight parameter."""
success = True
if use_flag:
success = F.depend(success, distributed_opt(weight, moment, linear, gradient, learning_rate, l1, l2, lr_power))
else:
success = F.depend(success, opt(weight, moment, linear, gradient, learning_rate, l1, l2, lr_power))
return success
@_ftrl_opt.register("Function", "Function", "Number", "Number", "Number", "Tensor", "Tensor",
"RowTensor", "Tensor", "Tensor", "Bool")
def _tensor_run_opt_with_sparse(opt, spars_opt, l1, l2, lr_power, learning_rate, linear,
gradient, weight, moment, cache_enable):
"""Apply sparse ftrl optimizer to the weight parameter when the gradient is sparse."""
success = True
indices = gradient.indices
values = gradient.values
success = F.depend(success, spars_opt(weight, moment, linear, values, indices))
return success
@_ftrl_opt.register("Function", "Function", "Number", "Number", "Number", "Tensor", "MapTensor",
"MapTensor", "MapTensor", "MapTensor", "Bool")
def _run_map_tensor_opt_with_sparse(opt, spars_opt, l1, l2, lr_power, learning_rate, linear,
gradient, weight, moment, cache_enable):
"""Apply sparse ftrl optimizer to the weight parameter when the gradient is sparse."""
success = True
indices, values = gradient.get_data()
_apply_map_tensor_ftrl(l1, l2, lr_power, learning_rate, linear, weight, moment, indices, values)
return success
@_ftrl_opt.register("Function", "Function", "Number", "Number", "Number", "Tensor", "Tensor",
"Tensor", "Tensor", "Tensor", "Bool")
def _tensor_run_opt(opt, spars_opt, l1, l2, lr_power, learning_rate, linear,
gradient, weight, moment, cache_enable):
"""Apply ftrl optimizer to the weight parameter."""
success = True
success = F.depend(success, opt(weight, moment, linear, gradient, learning_rate, l1, l2, lr_power))
return success
def _check_param(initial_accum, learning_rate, lr_power, l1, l2, use_locking, prim_name=None):
"""Check param."""
validator.check_value_type("initial_accum", initial_accum, [float], prim_name)
validator.check_number("initial_accum", initial_accum, 0.0, validator.GE, prim_name)
validator.check_value_type("learning_rate", learning_rate, [float], prim_name)
validator.check_positive_float(learning_rate, "learning_rate", prim_name)
validator.check_value_type("lr_power", lr_power, [float], prim_name)
validator.check_number("lr_power", lr_power, 0.0, validator.LE, prim_name)
validator.check_value_type("l1", l1, [float], prim_name)
validator.check_number("l1", l1, 0.0, validator.GE, prim_name)
validator.check_value_type("l2", l2, [float], prim_name)
validator.check_number("l2", l2, 0.0, validator.GE, prim_name)
validator.check_value_type("use_locking", use_locking, [bool], prim_name)
[docs]class FTRL(Optimizer):
r"""
Implements the FTRL algorithm.
FTRL is an online convex optimization algorithm that adaptively chooses its regularization function
based on the loss functions. Refer to paper `Adaptive Bound Optimization for Online Convex Optimization
<https://arxiv.org/abs/1002.4908>`_.
The updating formulas are as follows,
.. math::
\begin{array}{ll} \\
m_{t+1} = m_{t} + g^2 \\
u_{t+1} = u_{t} + g - \frac{m_{t+1}^\text{-p} - m_{t}^\text{-p}}{\alpha } * \omega_{t} \\
\omega_{t+1} =
\begin{cases}
\frac{(sign(u_{t+1}) * l1 - u_{t+1})}{\frac{m_{t+1}^\text{-p}}{\alpha } + 2 * l2 }
& \text{ if } |u_{t+1}| > l1 \\
0.0
& \text{ otherwise }
\end{cases}\\
\end{array}
:math:`m` represents accumulators, :math:`g` represents `grads`, :math:`t` represents the current step,
:math:`u` represents the linear coefficient to be updated, :math:`p` represents `lr_power`, :math:`\alpha`
represents `learning_rate`, :math:`\omega` represents `params`.
Note:
The sparse strategy is applied while the SparseGatherV2 operator is used for forward network. If the sparse
strategy wants to be executed on the host, set the target to the CPU.
The sparse feature is under continuous development.
If parameters are not grouped, the `weight_decay` in optimizer will be applied on the network parameters without
'beta' or 'gamma' in their names. Users can group parameters to change the strategy of decaying weight. When
parameters are grouped, each group can set `weight_decay`. If not, the `weight_decay` in optimizer will be
applied.
Args:
params (Union[list[Parameter], list[dict]]): Must be list of `Parameter` or list of `dict`. When the
`params` is a list of `dict`, the string "params", "weight_decay", "grad_centralization" and "order_params"
are the keys can be parsed.
- params: Required. Parameters in current group. The value must be a list of `Parameter`.
- lr: Using different learning rate by grouping parameters is currently not supported.
- weight_decay: Optional. If "weight_decay" in the keys, the value of corresponding weight decay
will be used. If not, the `weight_decay` in the optimizer will be used. It should be noted that weight
decay can be a constant value or a Cell. It is a Cell only when dynamic weight decay is applied. Dynamic
weight decay is similar to dynamic learning rate, users need to customize a weight decay schedule only
with global step as input, and during training, the optimizer calls the instance of WeightDecaySchedule
to get the weight decay value of current step.
- grad_centralization: Optional. Must be Boolean. If "grad_centralization" is in the keys, the set value
will be used. If not, the `grad_centralization` is False by default. This configuration only works on the
convolution layer.
- order_params: Optional. When parameters is grouped, this usually is used to maintain the order of
parameters that appeared in the network to improve performance. The value should be parameters whose
order will be followed in optimizer.
If `order_params` in the keys, other keys will be ignored and the element of 'order_params' must be in
one group of `params`.
initial_accum (float): The starting value for accumulators `m`, must be zero or positive values.
Default: ``0.1`` .
learning_rate (float): The learning rate value, must be zero or positive, dynamic learning rate is currently
not supported. Default: ``0.001`` .
lr_power (float): Learning rate power controls how the learning rate decreases during training, must be less
than or equal to zero. Use fixed learning rate if lr_power is zero. Default: ``-0.5`` .
l1 (float): l1 regularization strength, must be greater than or equal to zero. Default: ``0.0`` .
l2 (float): l2 regularization strength, must be greater than or equal to zero. Default: ``0.0`` .
use_locking (bool): If true, use locks for updating operation. Default: ``False`` .
loss_scale (float): Value for the loss scale. It must be greater than 0.0. In general, use the default value.
Only when `FixedLossScaleManager` is used for training and the `drop_overflow_update` in
`FixedLossScaleManager` is set to ``False`` , then this value needs to be the same as the `loss_scale` in
`FixedLossScaleManager`. Refer to class :class:`mindspore.amp.FixedLossScaleManager` for more details.
Default: ``1.0`` .
weight_decay (Union[float, int, Cell]): Weight decay (L2 penalty). Default: ``0.0`` .
- float: The fixed weight decay value. Must be equal to or greater than 0.
- int: The fixed weight decay value. Must be equal to or greater than 0. It will be converted to float.
- Cell: Weight decay is dynamic. During training, the optimizer calls the instance of
the Cell with step as the input to get the weight decay value of current step.
Inputs:
- **grads** (tuple[Tensor]) - The gradients of `params` in the optimizer, the shape is the same as the `params`
in optimizer.
Outputs:
Tuple[Parameter], the updated parameters, the shape is the same as `params`.
Raises:
TypeError: If `initial_accum`, `learning_rate`, `lr_power`, `l1`, `l2` or `loss_scale` is not a float.
TypeError: If element of `parameters` is neither Parameter nor dict.
TypeError: If `weight_decay` is neither float nor int.
TypeError: If `use_nesterov` is not a bool.
ValueError: If `lr_power` is greater than 0.
ValueError: If `loss_scale` is less than or equal to 0.
ValueError: If `initial_accum`, `l1` or `l2` is less than 0.
Supported Platforms:
``Ascend`` ``GPU``
Examples:
>>> import mindspore as ms
>>> from mindspore import nn
>>>
>>> # Define the network structure of LeNet5. Refer to
>>> # https://gitee.com/mindspore/docs/blob/r2.3.1/docs/mindspore/code/lenet.py
>>> net = LeNet5()
>>> #1) All parameters use the same learning rate and weight decay
>>> optim = nn.FTRL(params=net.trainable_params())
>>>
>>> #2) Use parameter groups and set different values
>>> conv_params = list(filter(lambda x: 'conv' in x.name, net.trainable_params()))
>>> no_conv_params = list(filter(lambda x: 'conv' not in x.name, net.trainable_params()))
>>> group_params = [{'params': conv_params, 'weight_decay': 0.01, 'grad_centralization':True},
... {'params': no_conv_params},
... {'order_params': net.trainable_params()}]
>>> optim = nn.FTRL(group_params, learning_rate=0.1, weight_decay=0.0)
>>> # The conv_params's parameters will use default learning rate of 0.1 and weight decay of 0.01 and grad
>>> # centralization of True.
>>> # The no_conv_params's parameters will use default learning rate of 0.1 will use default weight decay
>>> # of 0.0 and grad centralization of False.
>>> # The final parameters order in which the optimizer will be followed is the value of 'order_params'.
>>>
>>> loss = nn.SoftmaxCrossEntropyWithLogits()
>>> model = ms.train.Model(net, loss_fn=loss, optimizer=optim)
"""
@opt_init_args_register
def __init__(self, params, initial_accum=0.1, learning_rate=0.001, lr_power=-0.5, l1=0.0, l2=0.0,
use_locking=False, loss_scale=1.0, weight_decay=0.0):
super(FTRL, self).__init__(learning_rate, params, weight_decay, loss_scale=loss_scale)
if self.dynamic_lr or self.is_group_lr:
raise ValueError(f"For 'FTRL', dynamic learning rate and group learning rate are currently not supported "
f"in FTRL, they should all be false, but got dynamic learning rate {self.dynamic_lr} and"
f" group learning rate {self.is_group_lr}.")
_check_param(initial_accum, learning_rate, lr_power, l1, l2, use_locking, self.cls_name)
self.moments = self._parameters.clone(prefix="moments", init=initial_accum)
self.linear = self._parameters.clone(prefix="linear", init='zeros')
self.l1 = l1
self.l2 = l2
self.lr = learning_rate
self.lr_power = lr_power
if not self.is_group:
self.decay_flags = tuple((lambda: True)() for x in self._parameters)
self.opt = P.ApplyFtrl(use_locking=use_locking)
self.use_locking = use_locking
self.sparse_opt = P.SparseApplyFtrl(learning_rate, l1, l2, lr_power, use_locking=use_locking)
self._init_distributed_opts(use_locking, learning_rate, l1, l2, lr_power)
@jit
def construct(self, grads):
params = self._parameters
moments = self.moments
linear = self.linear
grads = self.flatten_gradients(grads)
grads = self.decay_weight(grads)
grads = self.gradients_centralization(grads)
grads = self.scale_grad(grads)
grads = self._grad_sparse_indices_deduplicate(grads)
lr = self.get_lr()
self.assignadd(self.global_step, self.global_step_increase_tensor)
if self.use_dist_optimizer:
success = self.map_(F.partial(_ftrl_opt, self.opt, self.sparse_opt, self.l1, self.l2, self.lr_power, lr),
linear, grads, params, moments, self.cache_enable,
self.distributed_opts, self.use_distributed_opt_flags,
self.distributed_sparse_opts, self.use_distributed_sparse_opt_flags)
else:
success = self.map_(F.partial(_ftrl_opt, self.opt, self.sparse_opt, self.l1, self.l2, self.lr_power, lr),
linear, grads, params, moments, self.cache_enable)
return success
@Optimizer.target.setter
def target(self, value):
"""
If the input value is set to "CPU", the parameters will be updated on the host using the Fused
optimizer operation.
"""
if not isinstance(value, str):
raise TypeError(f"For 'FTRL', the property 'target' must be string type, "
f"but got type {type(value)}.")
if value not in ('CPU', 'Ascend', 'GPU'):
raise ValueError(f"For 'FTRL', the property 'target' must be 'CPU', 'Ascend' or 'GPU', "
f"but got {value}.")
if value == 'CPU':
self.sparse_opt = P.FusedSparseFtrl(self.lr, self.l1, self.l2, self.lr_power, self.use_locking)
self.sparse_opt.set_device("CPU")
else:
self.sparse_opt = P.SparseApplyFtrl(self.lr, self.l1, self.l2, self.lr_power, self.use_locking)
self._target = value
def _init_distributed_opts(self, use_locking, learning_rate, l1, l2, lr_power):
self.use_dist_optimizer = self._use_distibuted_optimizer()
self.distributed_opts, self.use_distributed_opt_flags =\
self._get_distributed_optimizer_list("ftrl", use_locking=use_locking)
self.distributed_sparse_opts, self.use_distributed_sparse_opt_flags =\
self._get_distributed_optimizer_list("fused_sparse_ftrl", learning_rate,
l1, l2, lr_power, use_locking=use_locking)
def create_distributed_ftrl(*args, **kwargs):
"""
Create the distributed ApplyFtrl op.
"""
ftrl = P.ApplyFtrl(*args, **kwargs)
ftrl.add_prim_attr("gradient_type", "dense_gradient")
ftrl.add_prim_attr("parameter_input_index", 0)
ftrl.add_prim_attr("gradient_input_index", 3)
return ftrl
def create_distributed_fused_sparse_ftrl(*args, **kwargs):
"""
Create the distributed FusedSparseFtrl op.
"""
sparse_ftrl = P.FusedSparseFtrl(*args, **kwargs)
sparse_ftrl.add_prim_attr("gradient_type", "sparse_gradient")
sparse_ftrl.add_prim_attr("parameter_input_index", 0)
sparse_ftrl.add_prim_attr("gradient_input_index", 3)
sparse_ftrl.add_prim_attr("indices_input_index", 4)
return sparse_ftrl
_register_dist_optimizer("ftrl", create_distributed_ftrl)
_register_dist_optimizer("fused_sparse_ftrl", create_distributed_fused_sparse_ftrl)