--fix=update build loss api

mmseg/models/decode_heads/decode_head.py

@@ -8,9 +8,9 @@
 from mmengine.model import BaseModule
 from torch import Tensor
 
+from mmseg.registry import MODELS
 from mmseg.structures import build_pixel_sampler
 from mmseg.utils import ConfigType, SampleList
-from ..builder import build_loss
 from ..losses import accuracy
 from ..utils import resize
 
@@ -140,11 +140,11 @@ def __init__(self,
  self.threshold = threshold
 
  if isinstance(loss_decode, dict):
- self.loss_decode = build_loss(loss_decode)
+ self.loss_decode = MODELS.build(loss_decode)
  elif isinstance(loss_decode, (list, tuple)):
  self.loss_decode = nn.ModuleList()
  for loss in loss_decode:
- self.loss_decode.append(build_loss(loss))
+ self.loss_decode.append(MODELS.build(loss))
  else:
  raise TypeError(f'loss_decode must be a dict or sequence of dict,\
  but got {type(loss_decode)}')

mmseg/models/decode_heads/enc_head.py

@@ -9,7 +9,6 @@
 
 from mmseg.registry import MODELS
 from mmseg.utils import ConfigType, SampleList
-from ..builder import build_loss
 from ..utils import Encoding, resize
 from .decode_head import BaseDecodeHead
 
@@ -128,7 +127,7 @@ def __init__(self,
  norm_cfg=self.norm_cfg,
  act_cfg=self.act_cfg)
  if self.use_se_loss:
- self.loss_se_decode = build_loss(loss_se_decode)
+ self.loss_se_decode = MODELS.build(loss_se_decode)
  self.se_layer = nn.Linear(self.channels, self.num_classes)
 
  def forward(self, inputs):

mmseg/models/decode_heads/vpd_depth_head.py

@@ -10,7 +10,6 @@
 
 from mmseg.registry import MODELS
 from mmseg.utils import SampleList
-from ..builder import build_loss
 from ..utils import resize
 from .decode_head import BaseDecodeHead
 
@@ -184,11 +183,11 @@ def __init__(
 
  # build loss
  if isinstance(loss_decode, dict):
- self.loss_decode = build_loss(loss_decode)
+ self.loss_decode = MODELS.build(loss_decode)
  elif isinstance(loss_decode, (list, tuple)):
  self.loss_decode = nn.ModuleList()
  for loss in loss_decode:
- self.loss_decode.append(build_loss(loss))
+ self.loss_decode.append(MODELS.build(loss))
  else:
  raise TypeError(f'loss_decode must be a dict or sequence of dict,\
  but got {type(loss_decode)}')