39 lines
986 B
Python
39 lines
986 B
Python
import torch
|
|
|
|
from dassl.utils import check_isfile
|
|
from dassl.engine import TRAINER_REGISTRY, TrainerXU
|
|
|
|
|
|
@TRAINER_REGISTRY.register()
|
|
class AdaBN(TrainerXU):
|
|
"""Adaptive Batch Normalization.
|
|
|
|
https://arxiv.org/abs/1603.04779.
|
|
"""
|
|
|
|
def __init__(self, cfg):
|
|
super().__init__(cfg)
|
|
self.done_reset_bn_stats = False
|
|
|
|
def check_cfg(self, cfg):
|
|
assert check_isfile(
|
|
cfg.MODEL.INIT_WEIGHTS
|
|
), "The weights of source model must be provided"
|
|
|
|
def before_epoch(self):
|
|
if not self.done_reset_bn_stats:
|
|
for m in self.model.modules():
|
|
classname = m.__class__.__name__
|
|
if classname.find("BatchNorm") != -1:
|
|
m.reset_running_stats()
|
|
|
|
self.done_reset_bn_stats = True
|
|
|
|
def forward_backward(self, batch_x, batch_u):
|
|
input_u = batch_u["img"].to(self.device)
|
|
|
|
with torch.no_grad():
|
|
self.model(input_u)
|
|
|
|
return None
|