mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-07 12:21:27 +01:00
6e2bb1c054
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/34211 Test Plan: Imported from OSS Differential Revision: D20248684 Pulled By: albanD fbshipit-source-id: 2294bfa41b82ff47f000bc98860780f59d7d4421
103 lines
4.1 KiB
Python
103 lines
4.1 KiB
Python
import torch
|
|
from .optimizer import Optimizer
|
|
|
|
|
|
class Adagrad(Optimizer):
|
|
"""Implements Adagrad algorithm.
|
|
|
|
It has been proposed in `Adaptive Subgradient Methods for Online Learning
|
|
and Stochastic Optimization`_.
|
|
|
|
Arguments:
|
|
params (iterable): iterable of parameters to optimize or dicts defining
|
|
parameter groups
|
|
lr (float, optional): learning rate (default: 1e-2)
|
|
lr_decay (float, optional): learning rate decay (default: 0)
|
|
weight_decay (float, optional): weight decay (L2 penalty) (default: 0)
|
|
eps (float, optional): term added to the denominator to improve
|
|
numerical stability (default: 1e-10)
|
|
|
|
.. _Adaptive Subgradient Methods for Online Learning and Stochastic
|
|
Optimization: http://jmlr.org/papers/v12/duchi11a.html
|
|
"""
|
|
|
|
def __init__(self, params, lr=1e-2, lr_decay=0, weight_decay=0, initial_accumulator_value=0, eps=1e-10):
|
|
if not 0.0 <= lr:
|
|
raise ValueError("Invalid learning rate: {}".format(lr))
|
|
if not 0.0 <= lr_decay:
|
|
raise ValueError("Invalid lr_decay value: {}".format(lr_decay))
|
|
if not 0.0 <= weight_decay:
|
|
raise ValueError("Invalid weight_decay value: {}".format(weight_decay))
|
|
if not 0.0 <= initial_accumulator_value:
|
|
raise ValueError("Invalid initial_accumulator_value value: {}".format(initial_accumulator_value))
|
|
if not 0.0 <= eps:
|
|
raise ValueError("Invalid epsilon value: {}".format(eps))
|
|
|
|
defaults = dict(lr=lr, lr_decay=lr_decay, eps=eps, weight_decay=weight_decay,
|
|
initial_accumulator_value=initial_accumulator_value)
|
|
super(Adagrad, self).__init__(params, defaults)
|
|
|
|
for group in self.param_groups:
|
|
for p in group['params']:
|
|
state = self.state[p]
|
|
state['step'] = 0
|
|
state['sum'] = torch.full_like(p, initial_accumulator_value, memory_format=torch.preserve_format)
|
|
|
|
def share_memory(self):
|
|
for group in self.param_groups:
|
|
for p in group['params']:
|
|
state = self.state[p]
|
|
state['sum'].share_memory_()
|
|
|
|
@torch.no_grad()
|
|
def step(self, closure=None):
|
|
"""Performs a single optimization step.
|
|
|
|
Arguments:
|
|
closure (callable, optional): A closure that reevaluates the model
|
|
and returns the loss.
|
|
"""
|
|
loss = None
|
|
if closure is not None:
|
|
with torch.enable_grad():
|
|
loss = closure()
|
|
|
|
for group in self.param_groups:
|
|
for p in group['params']:
|
|
if p.grad is None:
|
|
continue
|
|
|
|
grad = p.grad
|
|
state = self.state[p]
|
|
|
|
state['step'] += 1
|
|
|
|
if group['weight_decay'] != 0:
|
|
if p.grad.is_sparse:
|
|
raise RuntimeError("weight_decay option is not compatible with sparse gradients")
|
|
grad = grad.add(p, alpha=group['weight_decay'])
|
|
|
|
clr = group['lr'] / (1 + (state['step'] - 1) * group['lr_decay'])
|
|
|
|
if grad.is_sparse:
|
|
grad = grad.coalesce() # the update is non-linear so indices must be unique
|
|
grad_indices = grad._indices()
|
|
grad_values = grad._values()
|
|
size = grad.size()
|
|
|
|
def make_sparse(values):
|
|
constructor = grad.new
|
|
if grad_indices.dim() == 0 or values.dim() == 0:
|
|
return constructor().resize_as_(grad)
|
|
return constructor(grad_indices, values, size)
|
|
state['sum'].add_(make_sparse(grad_values.pow(2)))
|
|
std = state['sum'].sparse_mask(grad)
|
|
std_values = std._values().sqrt_().add_(group['eps'])
|
|
p.add_(make_sparse(grad_values / std_values), alpha=-clr)
|
|
else:
|
|
state['sum'].addcmul_(grad, grad, value=1)
|
|
std = state['sum'].sqrt().add_(group['eps'])
|
|
p.addcdiv_(grad, std, value=-clr)
|
|
|
|
return loss
|