"""
Optimizers not provided by PyTorch natively.
"""
import math
from typing import Any, Callable, Dict, Iterable, Optional, Tuple, Union
import torch
from torch.optim.optimizer import Optimizer
Params = Union[Iterable[torch.Tensor], Iterable[dict]]
LossClosure = Callable[[], float]
OptLossClosure = Optional[LossClosure]
Betas2 = Tuple[float, float]
State = Dict[str, Any]
OptFloat = Optional[float]
Nus2 = Tuple[float, float]
[docs]class Ranger(Optimizer):
"""
Implements Ranger optimization algorithm (Lookahead with RAdam).
Implementation is modified version from ``pytorch-ranger`` package which build upon
its `original implementation <https://github.com/lessw2020/Ranger-Deep-Learning-Optimizer>`_.
Ranger seems to be benefiting most models.
Args:
params: iterable of parameters to optimize or dicts defining
parameter groups
lr: learning rate (default: 1e-3)
alpha: linear interpolation factor. 1.0 recovers the inner optimizer.
(default: 0.5)
k: number of lookahead steps (default: 6)
N_sma_threshhold: Maximum length of the simple moving average (SMA)
betas: coefficients used for computing
running averages of gradient and its square (default: (0.95, 0))
eps: term added to the denominator to improve
numerical stability (default: 1e-8)
weight_decay: weight decay (L2 penalty) (default: 0)
Example:
>>> from pytorch_forecasting.optim import Ranger
>>> optimizer = Ranger(model.parameters(), lr=0.1)
>>> optimizer.zero_grad()
>>> loss_fn(model(input), target).backward()
>>> scheduler = StepLR(optimizer, step_size=1, gamma=0.7)
>>> optimizer.step()
>>> scheduler.step()
"""
def __init__(
self,
params: Params,
lr: float = 1e-3,
alpha: float = 0.5,
k: int = 6,
N_sma_threshhold: int = 5,
betas: Betas2 = (0.95, 0.999),
eps: float = 1e-5,
weight_decay: float = 0,
):
# parameter checks
if not 0.0 <= alpha <= 1.0:
raise ValueError("Invalid slow update rate: {}".format(alpha))
if not 1 <= k:
raise ValueError("Invalid lookahead steps: {}".format(k))
if not lr > 0:
raise ValueError("Invalid Learning Rate: {}".format(lr))
if not eps > 0:
raise ValueError("Invalid eps: {}".format(eps))
# parameter comments:
# beta1 (momentum) of .95 seems to work better than .90...
# N_sma_threshold of 5 seems better in testing than 4.
# In both cases, worth testing on your dataset (.90 vs .95, 4 vs 5) to
# make sure which works best for you.
# prep defaults and init torch.optim base
defaults = dict(
lr=lr,
alpha=alpha,
k=k,
step_counter=0,
betas=betas,
N_sma_threshhold=N_sma_threshhold,
eps=eps,
weight_decay=weight_decay,
)
super().__init__(params, defaults)
# adjustable threshold
self.N_sma_threshhold = N_sma_threshhold
# now we can get to work...
# removed as we now use step from RAdam...no need for
# duplicate step counting
# for group in self.param_groups:
# group["step_counter"] = 0
# print("group step counter init")
# look ahead params
self.alpha = alpha
self.k = k
# radam buffer for state
self.radam_buffer = [[None, None, None] for ind in range(10)]
# self.first_run_check=0
# lookahead weights
# 9/2/19 - lookahead param tensors have been moved to state storage.
# This should resolve issues with load/save where weights were left in
# GPU memory from first load, slowing down future runs.
# self.slow_weights = [[p.clone().detach() for p in group['params']]
# for group in self.param_groups]
# don't use grad for lookahead weights
# for w in it.chain(*self.slow_weights):
# w.requires_grad = False
def __setstate__(self, state: dict) -> None:
super().__setstate__(state)
[docs] def step(self, closure: OptLossClosure = None) -> OptFloat:
r"""Performs a single optimization step.
Arguments:
closure: A closure that reevaluates the model and returns the loss.
"""
_ = closure()
loss = None
# note - below is commented out b/c I have other work that passes back
# the loss as a float, and thus not a callable closure.
# Uncomment if you need to use the actual closure...
# if closure is not None:
# loss = closure()
# Evaluate averages and grad, update param tensors
for group in self.param_groups:
for p in group["params"]:
if p.grad is None:
continue
grad = p.grad.data.float()
if grad.is_sparse:
raise RuntimeError("Ranger optimizer does not support " "sparse gradients")
p_data_fp32 = p.data.float()
state = self.state[p] # get state dict for this param
if len(state) == 0: # if first time to run...init dictionary
# with our desired entries
# if self.first_run_check==0:
# self.first_run_check=1
# print("Initializing slow buffer...should not see this
# at load from saved model!")
state["step"] = 0
state["exp_avg"] = torch.zeros_like(p_data_fp32)
state["exp_avg_sq"] = torch.zeros_like(p_data_fp32)
# look ahead weight storage now in state dict
state["slow_buffer"] = torch.empty_like(p.data)
state["slow_buffer"].copy_(p.data)
else:
state["exp_avg"] = state["exp_avg"].type_as(p_data_fp32)
state["exp_avg_sq"] = state["exp_avg_sq"].type_as(p_data_fp32)
# begin computations
exp_avg, exp_avg_sq = state["exp_avg"], state["exp_avg_sq"]
beta1, beta2 = group["betas"]
# compute variance mov avg
exp_avg_sq.mul_(beta2).addcmul_(grad, grad, value=1 - beta2)
# compute mean moving avg
exp_avg.mul_(beta1).add_(grad, alpha=1 - beta1)
state["step"] += 1
buffered = self.radam_buffer[int(state["step"] % 10)]
if state["step"] == buffered[0]:
N_sma, step_size = buffered[1], buffered[2]
else:
buffered[0] = state["step"]
beta2_t = beta2 ** state["step"]
N_sma_max = 2 / (1 - beta2) - 1
N_sma = N_sma_max - 2 * state["step"] * beta2_t / (1 - beta2_t)
buffered[1] = N_sma
if N_sma > self.N_sma_threshhold:
step_size = math.sqrt(
(1 - beta2_t)
* (N_sma - 4)
/ (N_sma_max - 4)
* (N_sma - 2)
/ N_sma
* N_sma_max
/ (N_sma_max - 2)
) / (1 - beta1 ** state["step"])
else:
step_size = 1.0 / (1 - beta1 ** state["step"])
buffered[2] = step_size
if group["weight_decay"] != 0:
p_data_fp32.add_(p_data_fp32, alpha=-group["weight_decay"] * group["lr"])
if N_sma > self.N_sma_threshhold:
denom = exp_avg_sq.sqrt().add_(group["eps"])
p_data_fp32.addcdiv_(exp_avg, denom, value=-step_size * group["lr"])
else:
p_data_fp32.add_(exp_avg, alpha=-step_size * group["lr"])
p.data.copy_(p_data_fp32)
# integrated look ahead...
# we do it at the param level instead of group level
if state["step"] % group["k"] == 0:
slow_p = state["slow_buffer"] # get access to slow param tensor
slow_p.add_(p.data - slow_p, alpha=self.alpha) # (fast weights - slow weights) * alpha
p.data.copy_(slow_p) # copy interpolated weights to RAdam param tensor
return loss