def init_weights(self, pretrained=None):
"""Initiate the parameters either from existing checkpoint or from
scratch."""
if isinstance(pretrained, str):
if self.pretrained2d:
# Inflate 2D model into 3D model.
self.inflate_weights(pretrained)
else:
# Directly load 3D model.
load_checkpoint(self, pretrained, strict=False)
elif pretrained is None:
for m in self.modules():
if isinstance(m, nn.Conv3d):
kaiming_init(m)
elif isinstance(m, nn.BatchNorm3d):
constant_init(m, 1)
if self.zero_init_residual:
for m in self.modules():
if isinstance(m, Bottleneck3d):
constant_init(m.bn3, 0)
elif isinstance(m, BasicBlock3d):
constant_init(m.bn2, 0)
else:
raise TypeError('pretrained must be a str or None')
def init_weights(self):
if isinstance(self.pretrained, str):
load_checkpoint(self, self.pretrained)
elif self.pretrained is None:
self.fast_path.init_weights(self.pretrained)
self.slow_path.init_weights(self.pretrained)
else:
raise TypeError('pretrained must be a str of None')
def init_weights(self, pretrained=None):
if isinstance(pretrained, str):
load_checkpoint(self, pretrained, strict=False)
elif pretrained is None:
for m in self.modules():
if isinstance(m, nn.Conv2d):
normal_init(m, std=self.init_gain)
elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)):
normal_init(m, mean=1, std=self.init_gain)
else:
raise TypeError('pretrained must be a str or None')
def init_weights(self, pretrained=None):
if isinstance(pretrained, str):
load_checkpoint(self, pretrained, strict=False)
elif pretrained is None:
for m in self.modules():
if isinstance(m, nn.Conv2d):
kaiming_init(m)
elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)):
constant_init(m, 1)
else:
raise ValueError(f'pretrained should be str or None.')
def init_weights(self, pretrained=None):
if isinstance(pretrained, str):
load_checkpoint(self, pretrained, strict=False)
elif pretrained is None:
for m in self.modules():
if isinstance(m, nn.Conv2d):
n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
normal_init(m, mean=0, std=math.sqrt(2. / n))
elif isinstance(m, nn.BatchNorm2d):
constant_init(m, 1)
else:
raise TypeError('pretrained must be str or None.')
def init_weights(self, pretrained=None):
if isinstance(pretrained, str):
load_checkpoint(self, pretrained, strict=False)
elif pretrained is None:
for m in self.modules():
if isinstance(m, nn.Conv2d):
kaiming_init(m)
elif isinstance(m, nn.BatchNorm2d):
constant_init(m, 1)
else:
raise TypeError('pretrained must be a str or None')
def init_weights(self, pretrained=None):
if isinstance(pretrained, str):
load_checkpoint(self, pretrained, strict=False)
elif pretrained is None:
for m in self.modules():
if isinstance(m, nn.Conv2d):
kaiming_init(m)
elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)):
constant_init(m, 1)
if self.zero_init_residual:
for m in self.modules():
if isinstance(m, Bottleneck):
constant_init(m.bn3, 0)
elif isinstance(m, BasicBlock):
constant_init(m.bn2, 0)
else:
raise TypeError('pretrained must be a str or None')
def load_checkpoint(self, filename, map_location='cpu', strict=False):
self.log.info('load checkpoint from %s', filename)
if filename.endswith('npy'):
dtype = torch.cuda.FloatTensor
if hasattr(self.model, 'module'):
state_dict = load_npy_weights(self.model.module, filename,
dtype)
load_state_dict(self.model.module, state_dict, strict)
else:
state_dict = load_npy_weights(self.model, filename, dtype)
load_state_dict(self.model, state_dict, strict)
else:
return load_checkpoint(self.model, filename, map_location, strict)
def init_weights(self, pretrained=None):
if isinstance(pretrained, str):
load_checkpoint(self, pretrained, strict=False)
elif pretrained is not None:
raise TypeError('pretrained must be str or None.')
def load_checkpoint(self, filename, map_location='cpu', strict=False):
self.log.info('load checkpoint from %s', filename)
return load_checkpoint(self.model, filename, map_location, strict)
