vilc
/
mindspore2022
forked from huawei/mindspore2022
0
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix comments for LossBase.get_loss

This commit is contained in:
chenhaozhe 2021-08-16 11:01:38 +08:00
parent 44268e9ad8
commit 209b2f1e04
1 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
mindspore/nn/loss/loss.py
View File

@ -76,8 +76,8 @@ class LossBase(Cell):
Args:
weights (Union[float, Tensor]): Optional `Tensor` whose rank is either 0, or the same rank as inputs,
and must be broadcastable to inputs (i.e., all dimensions must be either `1`,
or the same as the corresponding inputs dimension).
and must be broadcastable to inputs (i.e., all dimensions must be either `1`,
or the same as the corresponding inputs dimension).
"""
input_dtype = x.dtype
x = self.cast(x, mstype.float32)
@ -1282,10 +1282,10 @@ class FocalLoss(LossBase):
convert_weight = self.squeeze(convert_weight)
log_probability = log_probability * convert_weight
weight = F.pows(-probability + 1.0, self.gamma)
weight = F.pows(-1 * probability + 1.0, self.gamma)
if target.shape[1] == 1:
loss = (-weight * log_probability).mean(axis=1)
loss = (-1 * weight * log_probability).mean(axis=1)
else:
loss = (-weight * targets * log_probability).mean(axis=-1)
loss = (-1 * weight * targets * log_probability).mean(axis=-1)
return self.get_loss(loss)