File size: 2,883 Bytes
fe8202e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 |
import torch
import torch.nn as nn
import numpy as np
class DeepSupervisionWrapper(nn.Module):
def __init__(self, loss, weight_factors=None):
"""
Wraps a loss function so that it can be applied to multiple outputs. Forward accepts an arbitrary number of
inputs. Each input is expected to be a tuple/list. Each tuple/list must have the same length. The loss is then
applied to each entry like this:
l = w0 * loss(input0[0], input1[0], ...) + w1 * loss(input0[1], input1[1], ...) + ...
If weights are None, all w will be 1.
"""
super(DeepSupervisionWrapper, self).__init__()
self.weight_factors = weight_factors
self.loss = loss
def forward(self, *args):
for i in args:
assert isinstance(i, (tuple, list)), "all args must be either tuple or list, got %s" % type(i)
# we could check for equal lengths here as well but we really shouldn't overdo it with checks because
# this code is executed a lot of times!
if self.weight_factors is None:
weights = [1] * len(args[0])
else:
weights = self.weight_factors
# we initialize the loss like this instead of 0 to ensure it sits on the correct device, not sure if that's
# really necessary
l = weights[0] * self.loss(*[j[0] for j in args])
for i, inputs in enumerate(zip(*args)):
if i == 0:
continue
l += weights[i] * self.loss(*inputs)
return l
class AutoDeepSupervision(nn.Module):
def __init__(self, loss, label_scale) -> None:
super().__init__()
weights = np.array([1 / (2 ** i) for i in range(len(label_scale))])
weights[-1] = 0
# we don't use the lowest 2 outputs. Normalize weights so that they sum to 1
weights = weights / weights.sum()
print(f"loss weights is {weights}")
self.warpper = DeepSupervisionWrapper(loss, weights)
self.label_scale = label_scale
def forward(self, preds, label):
pred_len = len(preds)
assert pred_len == len(self.label_scale)
labels = []
for scale in self.label_scale:
labels.append(torch.nn.functional.interpolate(label, scale_factor=scale, mode="nearest"))
# label_1_2 = torch.nn.functional.interpolate(label, scale_factor=self.label_scale[1], mode="nearest")
# label_1_4 = torch.nn.functional.interpolate(label, scale_factor=self.label_scale[2], mode="nearest")
# label_1_8 = torch.nn.functional.interpolate(label, scale_factor=self.label_scale[3], mode="nearest")
# label_1_16 = torch.nn.functional.interpolate(label, scale_factor=self.label_scale[4], mode="nearest")
# labels = [label, label_1_2, label_1_4, label_1_8, label_1_16]
return self.warpper(preds, labels) |