!2277 Update _Loss to LossBase

Merge pull request !2277 from chenhaozhe/update_loss_base

official/cv/vit/src/cross_entropy.py

@@ -17,13 +17,7 @@
 from mindspore import nn
 from mindspore import Tensor
 from mindspore.common import dtype as mstype
-try:
-    from mindspore.nn.loss.loss import Loss
-except ImportError:
-    try:
-        from mindspore.nn.loss.loss import LossBase as Loss
-    except ImportError:
-        from mindspore.nn.loss.loss import _Loss as Loss
+from mindspore.nn.loss.loss import LossBase as Loss
 
 from mindspore.ops import functional as F
 from mindspore.ops import operations as P

research/cv/HRNetW48_cls/src/loss.py

@@ -16,12 +16,12 @@
 import mindspore.nn as nn
 from mindspore import Tensor
 from mindspore.common import dtype as mstype
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import functional as F
 from mindspore.ops import operations as P
 
 
-class CrossEntropySmooth(_Loss):
+class CrossEntropySmooth(LossBase):
     """CrossEntropy"""
     def __init__(self, sparse=True, reduction='mean', smooth_factor=0., num_classes=1000):
         super(CrossEntropySmooth, self).__init__()

research/cv/HRNetW48_seg/src/loss.py

@@ -18,10 +18,10 @@ import mindspore.ops.operations as P
 import mindspore.ops as F
 from mindspore.common.tensor import Tensor
 from mindspore import dtype as mstype
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 
 
-class CrossEntropyWithLogits(_Loss):
+class CrossEntropyWithLogits(LossBase):
     """
     Cross-entropy loss function for semantic segmentation,
     and different classes have the same weight.
@@ -61,7 +61,7 @@ class CrossEntropyWithLogits(_Loss):
         return loss
 
 
-class CrossEntropyWithWeights(_Loss):
+class CrossEntropyWithWeights(LossBase):
     """
     Cross-entropy loss function for semantic segmentation,
     and different classes have different weights.

research/cv/ICNet/Res50V1_PRE/src/CrossEntropySmooth.py

@@ -16,12 +16,12 @@
 import mindspore.nn as nn
 from mindspore import Tensor
 from mindspore.common import dtype as mstype
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import functional as F
 from mindspore.ops import operations as P
 
 
-class CrossEntropySmooth(_Loss):
+class CrossEntropySmooth(LossBase):
     """CrossEntropy"""
     def __init__(self, sparse=True, reduction='mean', smooth_factor=0., num_classes=1000):
         super(CrossEntropySmooth, self).__init__()

research/cv/NFNet/src/tools/criterion.py

@@ -17,12 +17,12 @@ import mindspore.nn as nn
 from mindspore import Tensor
 from mindspore import ops
 from mindspore.common import dtype as mstype
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import functional as F
 from mindspore.ops import operations as P
 
 
-class SoftTargetCrossEntropy(_Loss):
+class SoftTargetCrossEntropy(LossBase):
     """SoftTargetCrossEntropy for MixUp Augment"""
 
     def __init__(self):
@@ -38,7 +38,7 @@ class SoftTargetCrossEntropy(_Loss):
         return self.mean_ops(loss)
 
 
-class CrossEntropySmooth(_Loss):
+class CrossEntropySmooth(LossBase):
     """CrossEntropy"""
 
     def __init__(self, sparse=True, reduction='mean', smooth_factor=0., num_classes=1000):

research/cv/OCRNet/src/loss.py

@@ -19,7 +19,7 @@ import mindspore.ops.operations as P
 import mindspore.ops as F
 from mindspore.common.tensor import Tensor
 from mindspore import dtype as mstype
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 
 from src.config import config_hrnetv2_w48 as config
 
@@ -30,7 +30,7 @@ weights_list = [0.8373, 0.918, 0.866, 1.0345,
                 1.0865, 1.1529, 1.0507]
 
 
-class CrossEntropyWithLogits(_Loss):
+class CrossEntropyWithLogits(LossBase):
     """
     Cross-entropy loss function for semantic segmentation,
     and different classes have the same weight.
@@ -71,7 +71,7 @@ class CrossEntropyWithLogits(_Loss):
         return loss
 
 
-class CrossEntropyWithLogitsAndWeights(_Loss):
+class CrossEntropyWithLogitsAndWeights(LossBase):
     """
     Cross-entropy loss function for semantic segmentation,
     and different classes have different weights.

research/cv/Pix2Pix/src/models/loss.py

@@ -24,12 +24,12 @@ import mindspore.ops.operations as P
 from mindspore.parallel._utils import (_get_device_num, _get_gradients_mean, _get_parallel_mode)
 from mindspore.context import ParallelMode
 from mindspore.nn.wrap.grad_reducer import DistributedGradReducer
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from src.utils.config import get_args
 
 args = get_args()
 
-class SigmoidCrossEntropyWithLogits(_Loss):
+class SigmoidCrossEntropyWithLogits(LossBase):
     def __init__(self):
         super(SigmoidCrossEntropyWithLogits, self).__init__()
         self.cross_entropy = P.SigmoidCrossEntropyWithLogits()
@@ -38,7 +38,7 @@ class SigmoidCrossEntropyWithLogits(_Loss):
         x = self.cross_entropy(data, label)
         return self.get_loss(x)
 
-class D_Loss(_Loss):
+class D_Loss(LossBase):
     """
         Define Dloss.
     """
@@ -70,7 +70,7 @@ class D_WithLossCell(nn.Cell):
         pred0 = self.netD(realA, fakeB)
         return self._loss_fn(pred1, pred0)
 
-class G_Loss(_Loss):
+class G_Loss(LossBase):
     """
         Define Gloss.
     """

research/cv/ProtoNet/src/PrototypicalLoss.py

@@ -18,11 +18,11 @@ loss function script.
 import mindspore.ops as ops
 import mindspore.nn as nn
 from mindspore import Tensor
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 import mindspore as ms
 import numpy as np
 
-class PrototypicalLoss(_Loss):
+class PrototypicalLoss(LossBase):
     '''
     Loss class deriving from Module for the prototypical loss function defined below
     '''

research/cv/ResNeSt50/src/crossentropy.py

@@ -15,14 +15,14 @@
 """
 define loss function for network.
 """
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import operations as P
 from mindspore.ops import functional as F
 from mindspore import Tensor
 from mindspore.common import dtype as mstype
 import mindspore.nn as nn
 
-class CrossEntropy(_Loss):
+class CrossEntropy(LossBase):
     """
     the redefined loss function with SoftmaxCrossEntropyWithLogits.
     """

research/cv/SE_ResNeXt50/eval.py

@@ -24,7 +24,7 @@ from mindspore.train.serialization import load_checkpoint, load_param_into_net
 from mindspore.common import set_seed
 from mindspore import Tensor
 from mindspore.common import dtype as mstype
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import functional as F
 from mindspore.ops import operations as P
 
@@ -41,7 +41,7 @@ parser.add_argument('--checkpoint_path', type=str, default='./ckpt_0', help='Che
 args_opt = parser.parse_args()
 
 
-class CrossEntropySmooth(_Loss):
+class CrossEntropySmooth(LossBase):
     """CrossEntropy"""
     def __init__(self, sparse=True, reduction='mean', smooth_factor=0., num_classes=1000):
         super(CrossEntropySmooth, self).__init__()

research/cv/SE_ResNeXt50/train.py

@@ -34,7 +34,7 @@ from mindspore.context import ParallelMode
 from mindspore.train.serialization import load_checkpoint, load_param_into_net
 from mindspore.common import set_seed
 from mindspore.common import dtype as mstype
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import functional as F
 from mindspore.ops import operations as P
 
@@ -152,7 +152,7 @@ def warmup_cosine_annealing_lr(lr5, steps_per_epoch, warmup_epochs, max_epoch, T
     return np.array(lr_each_step).astype(np.float32)
 
 
-class CrossEntropySmooth(_Loss):
+class CrossEntropySmooth(LossBase):
     """CrossEntropy"""
     def __init__(self, sparse=True, reduction='mean', smooth_factor=0., num_classes=1000):
         super(CrossEntropySmooth, self).__init__()

research/cv/TNT/src/tools/criterion.py

@@ -17,12 +17,12 @@ import mindspore.nn as nn
 from mindspore import Tensor
 from mindspore import ops
 from mindspore.common import dtype as mstype
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import functional as F
 from mindspore.ops import operations as P
 
 
-class SoftTargetCrossEntropy(_Loss):
+class SoftTargetCrossEntropy(LossBase):
     """SoftTargetCrossEntropy for MixUp Augment"""
 
     def __init__(self):
@@ -38,7 +38,7 @@ class SoftTargetCrossEntropy(_Loss):
         return self.mean_ops(loss)
 
 
-class CrossEntropySmooth(_Loss):
+class CrossEntropySmooth(LossBase):
     """CrossEntropy"""
 
     def __init__(self, sparse=True, reduction='mean', smooth_factor=0., num_classes=1000):

research/cv/VehicleNet/src/loss.py

@@ -13,14 +13,14 @@
 # limitations under the License.
 # ============================================================================
 """loss function"""
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import operations as P
 from mindspore.ops import functional as F
 from mindspore import Tensor
 from mindspore.common import dtype as mstype
 import mindspore.nn as nn
 
-class CrossEntropy(_Loss):
+class CrossEntropy(LossBase):
     """CrossEntropy"""
     def __init__(self, smooth_factor=0., num_classes=1000):
         super(CrossEntropy, self).__init__()

research/cv/aecrnet/src/contras_loss.py

@@ -16,7 +16,7 @@
 import os
 
 import mindspore.ops as ops
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore import nn
 from mindspore import load_checkpoint, load_param_into_net
 from mindspore.ops.functional import stop_gradient
@@ -82,7 +82,7 @@ class Vgg19(nn.Cell):
         return out
 
 
-class ContrastLoss(_Loss):
+class ContrastLoss(LossBase):
     """[ContrastLoss]
 
     Args:

research/cv/cct/src/tools/criterion.py

@@ -17,12 +17,12 @@ import mindspore.nn as nn
 from mindspore import Tensor
 from mindspore import ops
 from mindspore.common import dtype as mstype
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import functional as F
 from mindspore.ops import operations as P
 
 
-class SoftTargetCrossEntropy(_Loss):
+class SoftTargetCrossEntropy(LossBase):
     """SoftTargetCrossEntropy for MixUp Augment"""
 
     def __init__(self):
@@ -38,7 +38,7 @@ class SoftTargetCrossEntropy(_Loss):
         return self.mean_ops(loss)
 
 
-class CrossEntropySmooth(_Loss):
+class CrossEntropySmooth(LossBase):
     """CrossEntropy"""
 
     def __init__(self, sparse=True, reduction='mean', smooth_factor=0., num_classes=1000):

research/cv/csd/src/contras_loss.py

@@ -16,7 +16,7 @@
 import os
 
 import mindspore.ops as ops
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore import nn
 from mindspore import load_checkpoint, load_param_into_net
 from mindspore.ops.functional import stop_gradient
@@ -108,7 +108,7 @@ class Vgg19(nn.Cell):
         return out
 
 
-class ContrastLoss(_Loss):
+class ContrastLoss(LossBase):
     """[ContrastLoss]
 
     Args:

research/cv/efficientnet-b0/src/loss.py

@@ -16,12 +16,12 @@
 import mindspore.nn as nn
 from mindspore import Tensor
 from mindspore.common import dtype as mstype
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import functional as F
 from mindspore.ops import operations as P
 
 
-class CrossEntropySmooth(_Loss):
+class CrossEntropySmooth(LossBase):
     """CrossEntropy"""
     def __init__(self, sparse=True, reduction='mean', smooth_factor=0., num_classes=1000):
         super(CrossEntropySmooth, self).__init__()

research/cv/efficientnet-b1/src/loss.py

@@ -13,7 +13,7 @@
 # limitations under the License.
 # ============================================================================
 """define loss function for network."""
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import operations as P
 from mindspore.ops import functional as F
 from mindspore.common import dtype as mstype
@@ -21,7 +21,7 @@ from mindspore import Tensor
 import mindspore.nn as nn
 
 
-class LabelSmoothingCrossEntropy(_Loss):
+class LabelSmoothingCrossEntropy(LossBase):
     """LabelSmoothingCrossEntropy"""
     def __init__(self, smooth_factor=0.1, num_classes=1000):
         super(LabelSmoothingCrossEntropy, self).__init__()
@@ -38,7 +38,7 @@ class LabelSmoothingCrossEntropy(_Loss):
         return loss_logit
 
 
-class CrossEntropySmooth(_Loss):
+class CrossEntropySmooth(LossBase):
     """CrossEntropy"""
     def __init__(self, sparse=True, reduction='mean', smooth_factor=0., num_classes=1000):
         super(CrossEntropySmooth, self).__init__()

research/cv/efficientnet-b3/src/loss.py

@@ -16,12 +16,12 @@
 import mindspore.nn as nn
 from mindspore import Tensor
 from mindspore.common import dtype as mstype
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import functional as F
 from mindspore.ops import operations as P
 
 
-class CrossEntropySmooth(_Loss):
+class CrossEntropySmooth(LossBase):
     """CrossEntropy"""
     def __init__(self, sparse=True, reduction='mean', smooth_factor=0., num_classes=1000):
         super(CrossEntropySmooth, self).__init__()

research/cv/fishnet99/eval.py

@@ -24,7 +24,7 @@ from mindspore.train.serialization import load_checkpoint, load_param_into_net
 from mindspore.common import set_seed
 from mindspore import Tensor
 from mindspore.common import dtype as mstype
-from mindspore.nn.loss.loss import _Loss
+from mindspore.nn.loss.loss import LossBase
 from mindspore.ops import functional as F
 from mindspore.ops import operations as P
 
@@ -43,7 +43,7 @@ parser.add_argument('--checkpoint_path', type=str, default='./ckpt_0', help='Che
 args_opt = parser.parse_args()
 
 
-class CrossEntropySmooth(_Loss):
+class CrossEntropySmooth(LossBase):
     """CrossEntropy"""
     def __init__(self, sparse=True, reduction='mean', smooth_factor=0., num_classes=1000):
         super(CrossEntropySmooth, self).__init__()