import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.nn import Parameter
import math
class LinearSequential(nn.Module):
def __init__(self, in_features, layers=[128, 2], dropout_probs=[0.5, 0.]):
super(LinearSequential, self).__init__()
self.in_features = in_features
self.sequential = self.__make_layers__(layers, dropout_probs)
def __make_layers__(self, layer_sizes, dropout_probs):
layers = []
input_size = self.in_features
for layer_size, dropout_p in zip(layer_sizes, dropout_probs):
output_size = layer_size
layers.append(nn.Linear(input_size, output_size))
layers.append(nn.ReLU(inplace=True))
if dropout_p > 0 and dropout_p < 1:
layers.append(nn.Dropout(p=dropout_p))
input_size = output_size
return nn.Sequential(*layers)
def forward(self, input, target=None):
return self.sequential(input)
class ArcMarginProduct(nn.Module):
def __init__(self, in_features, out_features, s=30.0, m=0.50, easy_margin=True):
super(ArcMarginProduct, self).__init__()
self.in_features = in_features
self.out_features = out_features
self.s = s
self.m = m
self.weight = Parameter(torch.FloatTensor(out_features, in_features))
nn.init.xavier_uniform_(self.weight)
self.easy_margin = easy_margin
self.cos_m = math.cos(m)
self.sin_m = math.sin(m)
self.th = math.cos(math.pi - m)
self.mm = math.sin(math.pi - m) * m
def forward(self, input, label=None):
if self.training:
# --------------------------- cos(theta) & phi(theta) ---------------------------
cosine = F.linear(F.normalize(input), F.normalize(self.weight))
sine = torch.sqrt(1.0 - torch.pow(cosine, 2))
phi = cosine * self.cos_m - sine * self.sin_m
if self.easy_margin:
phi = torch.where(cosine > 0, phi, cosine)
else:
phi = torch.where(cosine > self.th, phi, cosine - self.mm)
# --------------------------- convert label to one-hot ---------------------------
# one_hot = torch.zeros(cosine.size(), requires_grad=True, device='cuda')
one_hot = torch.zeros(cosine.size(), device='cuda')
one_hot.scatter_(1, label.view(-1, 1).long(), 1)
# -------------torch.where(out_i = {x_i if condition_i else y_i) -------------
output = (one_hot * phi) + ((1.0 - one_hot) * cosine) # you can use torch.where if your torch.__version__ is 0.4
output *= self.s
else:
cosine = F.linear(F.normalize(input), F.normalize(self.weight))
output = self.s * cosine
return output
class ArcMarginProduct_v2(nn.Module):
def __init__(self, in_features, out_features, m=0.50, easy_margin=True):
super(ArcMarginProduct_v2, self).__init__()
self.in_features = in_features
self.out_features = out_features
self.m = m
self.weight = Parameter(torch.FloatTensor(out_features, in_features))
nn.init.xavier_uniform_(self.weight)
self.easy_margin = easy_margin
self.cos_m = math.cos(m)
self.sin_m = math.sin(m)
self.th = math.cos(math.pi - m)
self.mm = math.sin(math.pi - m) * m
def forward(self, input, label=None, alpha=0.1):
cosine = F.linear(F.normalize(input), F.normalize(self.weight))
input_norm = input.norm(2, 1, True).clamp(min=1e-12)
crossentropy_output = F.linear(input, self.weight)
if self.training:
# --------------------------- cos(theta) & phi(theta) ---------------------------
sine = torch.sqrt(1.0 - torch.pow(cosine, 2))
phi = cosine * self.cos_m - sine * self.sin_m
if self.easy_margin:
phi = torch.where(cosine > 0, phi, cosine)
else:
phi = torch.where(cosine > self.th, phi, cosine - self.mm)
# --------------------------- convert label to one-hot ---------------------------
# one_hot = torch.zeros(cosine.size(), requires_grad=True, device='cuda')
one_hot = torch.zeros(cosine.size(), device='cuda')
one_hot.scatter_(1, label.view(-1, 1).long(), 1)
# -------------torch.where(out_i = {x_i if condition_i else y_i) -------------
arcmargin_output = (one_hot * phi) + ((1.0 - one_hot) * cosine)
else:
arcmargin_output = cosine
arcmargin_output = arcmargin_output * input_norm.expand_as(arcmargin_output)
output = alpha * arcmargin_output + (1 - alpha) * crossentropy_output
return output
class ArcMarginProduct_v3(nn.Module):
def __init__(self, in_features, out_features, m=0.01, easy_margin=True):
super(ArcMarginProduct_v3, self).__init__()
self.in_features = in_features
self.out_features = out_features
self.m = m
self.weight = Parameter(torch.FloatTensor(out_features, in_features))
nn.init.xavier_uniform_(self.weight)
self.easy_margin = easy_margin
self.cos_m = math.cos(m)
self.sin_m = math.sin(m)
self.th = math.cos(math.pi - m)
self.mm = math.sin(math.pi - m) * m
def forward(self, input, label=None, alpha=0.1):
cosine = F.linear(F.normalize(input), F.normalize(self.weight))
input_norm = input.norm(2, 1, True).clamp(min=1e-12)
crossentropy_output = F.linear(input, self.weight)
if self.training:
# --------------------------- cos(theta) & phi(theta) ---------------------------
sine = torch.sqrt(1.0 - torch.pow(cosine, 2))
phi = cosine * self.cos_m - sine * self.sin_m
if self.easy_margin:
phi = torch.where(cosine > 0, phi, cosine)
else:
phi = torch.where(cosine > self.th, phi, cosine - self.mm)
# --------------------------- convert label to one-hot ---------------------------
# one_hot = torch.zeros(cosine.size(), requires_grad=True, device='cuda')
one_hot = torch.zeros(cosine.size(), device='cuda')
# Only for positive class
one_hot[:,1] = label
# -------------torch.where(out_i = {x_i if condition_i else y_i) -------------
arcmargin_output = (one_hot * phi) + ((1.0 - one_hot) * cosine)
else:
arcmargin_output = cosine
arcmargin_output = arcmargin_output * input_norm.expand_as(arcmargin_output)
output = alpha * arcmargin_output + (1 - alpha) * crossentropy_output
return output
