Files
insightface/recognition/partial_fc/pytorch/partial_classifier.py
2020-11-11 10:18:30 +08:00

96 lines
3.7 KiB
Python

"""
Author: {Yang Xiao, Xiang An, XuHan Zhu} in DeepGlint,
Partial FC: Training 10 Million Identities on a Single Machine
See the original paper:
https://arxiv.org/abs/2010.05222
"""
import math
from typing import Any
import torch
import torch.distributed as dist
import torch.nn.functional as F
from torch.nn import Module
from torch.nn import init
from torch.nn.parameter import Parameter
from config import config as cfg
class DistSampleClassifier(Module):
def _forward_unimplemented(self, *input: Any) -> None:
pass
@torch.no_grad()
def __init__(self, rank, local_rank, world_size):
super(DistSampleClassifier, self).__init__()
self.sample_rate = cfg.sample_rate
self.num_local = cfg.num_classes // world_size + int(
rank < cfg.num_classes % world_size)
self.class_start = cfg.num_classes // world_size * rank + min(
rank, cfg.num_classes % world_size)
self.num_sample = int(self.sample_rate * self.num_local)
self.local_rank = local_rank
self.world_size = world_size
self.weight = torch.empty(size=(self.num_local, cfg.embedding_size),
device=local_rank)
self.weight_mom = torch.zeros_like(self.weight)
self.stream = torch.cuda.Stream(local_rank)
init.kaiming_uniform_(self.weight, a=math.sqrt(5))
self.index = None
if int(self.sample_rate) == 1:
self.update = lambda: 0
self.sub_weight = Parameter(self.weight)
self.sub_weight_mom = self.weight_mom
else:
self.sub_weight = Parameter(torch.empty((0, 0)).cuda(local_rank))
@torch.no_grad()
def sample(self, total_label):
P = (self.class_start <=
total_label) & (total_label < self.class_start + self.num_local)
total_label[~P] = -1
total_label[P] -= self.class_start
if int(self.sample_rate) != 1:
positive = torch.unique(total_label[P], sorted=True)
if self.num_sample - positive.size(0) >= 0:
perm = torch.rand(self.num_local,device=cfg.local_rank)
perm[positive] = 2.0
index = torch.topk(perm,k=self.num_sample)[1]
index = index.sort()[0]
else:
index = positive
self.index = index
total_label[P] = torch.searchsorted(index, total_label[P])
self.sub_weight = Parameter(self.weight[index])
self.sub_weight_mom = self.weight_mom[index]
def forward(self, total_features, norm_weight):
torch.cuda.current_stream().wait_stream(self.stream)
logits = F.linear(total_features, norm_weight)
return logits
@torch.no_grad()
def update(self, ):
self.weight_mom[self.index] = self.sub_weight_mom
self.weight[self.index] = self.sub_weight
def prepare(self, label, optimizer):
with torch.cuda.stream(self.stream):
total_label = torch.zeros(label.size()[0] * self.world_size,
device=self.local_rank,
dtype=torch.long)
dist.all_gather(list(total_label.chunk(self.world_size, dim=0)),
label)
self.sample(total_label)
optimizer.state.pop(optimizer.param_groups[-1]['params'][0], None)
optimizer.param_groups[-1]['params'][0] = self.sub_weight
optimizer.state[
self.sub_weight]['momentum_buffer'] = self.sub_weight_mom
norm_weight = F.normalize(self.sub_weight)
return total_label, norm_weight