# Copyright 2023 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.
# ============================================================================
"""radam"""
from __future__ import absolute_import
from mindspore.ops import functional as F, composite as C, operations as P
from mindspore.common import Tensor, Parameter
import mindspore.common.dtype as mstype
from mindspore import _checkparam as validator
from mindspore.experimental.optim.optimizer import Optimizer, check_not_less_than, check_not_less_than_without_equal
from mindspore import jit
_radam_opt = C.MultitypeFuncGraph("radam_opt")
op_pow = P.Pow()
op_sqrt = P.Sqrt()
op_cast = P.Cast()
@_radam_opt.register("Number", "Number", "Number", "Tensor", "Number", "Tensor", "Tensor", "Tensor", "Tensor", "Tensor",
"Tensor", "Tensor")
def _tensor_run_opt(beta1, beta2, eps, lr, rho_inf, rho_t, bias_correction1, bias_correction2, param, grad, exp_avg,
exp_avg_sq):
"""Apply radam optimizer to the weight parameter."""
F.assign(exp_avg, exp_avg * beta1 + grad * (1 - beta1))
F.assign(exp_avg_sq, exp_avg_sq * beta2 + grad * grad * (1 - beta2))
bias_corrected_exp_avg = exp_avg / bias_correction1
if rho_t > 5.0:
rect = op_sqrt((rho_t - 4) * (rho_t - 2) * rho_inf / ((rho_inf - 4) * (rho_inf - 2) * rho_t))
exp_avg_sq_sqrt = op_sqrt(exp_avg_sq) + eps
adaptive_lr = op_sqrt(bias_correction2) / exp_avg_sq_sqrt
F.assign(param, param - bias_corrected_exp_avg * lr * adaptive_lr * rect)
else:
F.assign(param, param - bias_corrected_exp_avg * lr)
return True
[文档]class RAdam(Optimizer):
r"""
Implements RAdam algorithm.
.. math::
\begin{aligned}
&\rule{110mm}{0.4pt} \\
&\textbf{input} : \gamma \text{ (lr)}, \: \beta_1, \beta_2
\text{ (betas)}, \: \theta_0 \text{ (params)}, \:f(\theta) \text{ (objective)}, \:
\lambda \text{ (weightdecay)}, \\
&\hspace{13mm} \epsilon \text{ (epsilon)} \\
&\textbf{initialize} : m_0 \leftarrow 0 \text{ ( first moment)},
v_0 \leftarrow 0 \text{ ( second moment)}, \\
&\hspace{18mm} \rho_{\infty} \leftarrow 2/(1-\beta_2) -1 \\[-1.ex]
&\rule{110mm}{0.4pt} \\
&\textbf{for} \: t=1 \: \textbf{to} \: \ldots \: \textbf{do} \\
&\hspace{6mm}g_t \leftarrow \nabla_{\theta} f_t (\theta_{t-1}) \\
&\hspace{5mm} \textbf{if} \: \lambda \neq 0 \\
&\hspace{10mm} g_t \leftarrow g_t + \lambda \theta_{t-1} \\
&\hspace{6mm}m_t \leftarrow \beta_1 m_{t-1} + (1 - \beta_1) g_t \\
&\hspace{6mm}v_t \leftarrow \beta_2 v_{t-1} + (1-\beta_2) g^2_t \\
&\hspace{6mm}\widehat{m_t} \leftarrow m_t/\big(1-\beta_1^t \big) \\
&\hspace{6mm}\rho_t \leftarrow \rho_{\infty} -
2 t \beta^t_2 /\big(1-\beta_2^t \big) \\[-1.ex]
&\hspace{6mm}\textbf{if} \: \rho_t > 5 \\
&\hspace{12mm} l_t \leftarrow \frac{\sqrt{ (1-\beta^t_2) }}{ \sqrt{v_t} +\epsilon } \\
&\hspace{12mm} r_t \leftarrow
\sqrt{\frac{(\rho_t-4)(\rho_t-2)\rho_{\infty}}{(\rho_{\infty}-4)(\rho_{\infty}-2) \rho_t}} \\
&\hspace{12mm}\theta_t \leftarrow \theta_{t-1} - \gamma \widehat{m_t} r_t l_t \\
&\hspace{6mm}\textbf{else} \\
&\hspace{12mm}\theta_t \leftarrow \theta_{t-1} - \gamma \widehat{m_t} \\
&\rule{110mm}{0.4pt} \\[-1.ex]
&\bf{return} \: \theta_t \\[-1.ex]
&\rule{110mm}{0.4pt} \\[-1.ex]
\end{aligned}
.. warning::
This is an experimental optimizer API that is subject to change.
This module must be used with lr scheduler module in `LRScheduler Class
<https://www.mindspore.cn/docs/en/r2.4.1/api_python/mindspore.experimental.html#lrscheduler-class>`_ .
Args:
params (Union[list(Parameter), list(dict)]): list of parameters to optimize or dicts defining
parameter groups.
lr (Union[int, float, Tensor], optional): learning rate. Default: ``1e-3``.
betas (Tuple[float, float], optional): coefficients used for computing
running averages of gradient and its square. Default: ``(0.9, 0.999)``.
eps (float, optional): term added to the denominator to improve
numerical stability. Default: ``1e-8``.
weight_decay (float, optional): weight decay (L2 penalty). Default: ``0.0``.
Inputs:
- **gradients** (tuple[Tensor]) - The gradients of `params`.
Raises:
ValueError: If the learning rate is not int, float or Tensor.
ValueError: If the learning rate is less than 0.
ValueError: If the `eps` is less than 0.0.
ValueError: If the `weight_decay` is less than 0.
ValueError: If elements of `betas` not in the range of [0, 1).
Supported Platforms:
``Ascend`` ``GPU`` ``CPU``
Examples:
>>> import mindspore
>>> from mindspore import nn
>>> from mindspore.experimental import optim
>>> # Define the network structure of LeNet5. Refer to
>>> # https://gitee.com/mindspore/docs/blob/r2.4.1/docs/mindspore/code/lenet.py
>>> net = LeNet5()
>>> loss_fn = nn.SoftmaxCrossEntropyWithLogits(sparse=True)
>>> optimizer = optim.RAdam(net.trainable_params(), lr=0.1)
>>> def forward_fn(data, label):
... logits = net(data)
... loss = loss_fn(logits, label)
... return loss, logits
>>> grad_fn = mindspore.value_and_grad(forward_fn, None, optimizer.parameters, has_aux=True)
>>> def train_step(data, label):
... (loss, _), grads = grad_fn(data, label)
... optimizer(grads)
... return loss
"""
def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weight_decay=0.0):
check_not_less_than_without_equal(lr, "lr", self.cls_name)
check_not_less_than(weight_decay, "weight_decay", self.cls_name)
check_not_less_than_without_equal(eps, "eps", self.cls_name)
validator.check_float_range(betas[0], 0., 1., validator.INC_LEFT, "betas[0]", self.cls_name)
validator.check_float_range(betas[1], 0., 1., validator.INC_LEFT, "betas[1]", self.cls_name)
defaults = dict(
lr=lr,
betas=betas,
eps=eps,
weight_decay=weight_decay,
)
super(RAdam, self).__init__(params, defaults)
self.step_t = Parameter(Tensor(0, mstype.int32), "step_t")
self.exp_avg = self.parameters.clone(prefix="exp_avg", init='zeros')
self.exp_avg_sq = self.parameters.clone(prefix="exp_avg_sq", init='zeros')
self.increase_tensor = Tensor(1, mstype.int32)
self.assignadd = P.AssignAdd()
@jit
def implementation(self, lr, beta1, beta2, weight_decay, eps, start_id, end_id, gradients):
"""Extract the common computing part for acceleration"""
params = self.parameters[start_id: end_id]
grads = gradients[start_id: end_id]
grads = self._decay_weight(weight_decay, params, grads)
exp_avg = self.exp_avg[start_id: end_id]
exp_avg_sq = self.exp_avg_sq[start_id: end_id]
bias_correction1 = 1 - op_pow(beta1, self.step_t.value())
bias_correction2 = 1 - op_pow(beta2, self.step_t.value())
rho_inf = 2 / (1 - beta2) - 1
beta2_pow = op_pow(beta2, self.step_t.value())
right = 2 * self.step_t.value() * beta2_pow / bias_correction2
rho_t = rho_inf - right
self.hyper_map(F.partial(_radam_opt, beta1, beta2, eps, lr, rho_inf, rho_t, bias_correction1, bias_correction2),
params, grads, exp_avg, exp_avg_sq)
return True
def construct(self, gradients):
self.assignadd(self.step_t, self.increase_tensor)
for group_id, group in enumerate(self.param_groups):
lr = self.lrs[group_id]
if isinstance(group.get("lr"), float):
lr = op_cast(group.get("lr"), mstype.float32)
beta1, beta2 = group["betas"]
start_id = self.group_start_id[group_id]
end_id = self.group_start_id[group_id + 1]
weight_decay = group["weight_decay"]
eps = group["eps"]
self.implementation(lr, beta1, beta2, weight_decay, eps, start_id, end_id, gradients)
return True