From 14d66fb2775ce5aa4e54aab567ff183456e37a2a Mon Sep 17 00:00:00 2001 From: Tongzhou Wang Date: Mon, 30 Nov 2020 13:06:39 -0500 Subject: [PATCH] Update loss module doc --- torch/nn/modules/loss.py | 44 ++++++++++++++++++++-------------------- 1 file changed, 22 insertions(+), 22 deletions(-) diff --git a/torch/nn/modules/loss.py b/torch/nn/modules/loss.py index a61a39fd28b6..7ec3f36d849e 100644 --- a/torch/nn/modules/loss.py +++ b/torch/nn/modules/loss.py @@ -58,7 +58,7 @@ class L1Loss(_Loss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -148,7 +148,7 @@ class NLLLoss(_WeightedLoss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` ignore_index (int, optional): Specifies a target value that is ignored and does not contribute to the input gradient. When :attr:`size_average` is ``True``, the loss is averaged over @@ -251,7 +251,7 @@ class PoissonNLLLoss(_Loss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` eps (float, optional): Small value to avoid evaluation of :math:`\log(0)` when :attr:`log_input = False`. Default: 1e-8 reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the @@ -339,7 +339,7 @@ class KLDivLoss(_Loss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -413,7 +413,7 @@ class MSELoss(_Loss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -494,7 +494,7 @@ class BCELoss(_WeightedLoss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -593,7 +593,7 @@ class BCEWithLogitsLoss(_Loss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -667,7 +667,7 @@ class HingeEmbeddingLoss(_Loss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -722,7 +722,7 @@ class MultiLabelMarginLoss(_Loss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -763,9 +763,9 @@ def forward(self, input: Tensor, target: Tensor) -> Tensor: class SmoothL1Loss(_Loss): r"""Creates a criterion that uses a squared term if the absolute element-wise error falls below beta and an L1 term otherwise. - It is less sensitive to outliers than the `MSELoss` and in some cases + It is less sensitive to outliers than the :class:`torch.nn.MSELoss` and in some cases prevents exploding gradients (e.g. see `Fast R-CNN` paper by Ross Girshick). - Also known as the Huber loss: + Omitting a scaling factor of :attr:`beta`, this loss is also known as the Huber loss: .. math:: \text{loss}(x, y) = \frac{1}{n} \sum_{i} z_{i} @@ -782,10 +782,10 @@ class SmoothL1Loss(_Loss): :math:`x` and :math:`y` arbitrary shapes with a total of :math:`n` elements each the sum operation still operates over all the elements, and divides by :math:`n`. - beta is an optional parameter that defaults to 1. + :attr:`beta` is an optional parameter that defaults to 1. - Note: When beta is set to 0, this is equivalent to :class:`L1Loss`. - Passing a negative value in for beta will result in an exception. + Note: When :attr:`beta` is set to 0, this is equivalent to :class:`L1Loss`. + Passing a negative value in for :attr:`beta` will result in an exception. The division by :math:`n` can be avoided if sets ``reduction = 'sum'``. @@ -794,7 +794,7 @@ class SmoothL1Loss(_Loss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -839,7 +839,7 @@ class SoftMarginLoss(_Loss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -916,7 +916,7 @@ class CrossEntropyLoss(_WeightedLoss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` ignore_index (int, optional): Specifies a target value that is ignored and does not contribute to the input gradient. When :attr:`size_average` is ``True``, the loss is averaged over non-ignored targets. @@ -988,7 +988,7 @@ class MultiLabelSoftMarginLoss(_WeightedLoss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -1039,7 +1039,7 @@ class CosineEmbeddingLoss(_Loss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -1081,7 +1081,7 @@ class MarginRankingLoss(_Loss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -1153,7 +1153,7 @@ class MultiMarginLoss(_WeightedLoss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per @@ -1220,7 +1220,7 @@ class TripletMarginLoss(_Loss): the losses are averaged over each loss element in the batch. Note that for some losses, there are multiple elements per sample. If the field :attr:`size_average` is set to ``False``, the losses are instead summed for each minibatch. Ignored - when reduce is ``False``. Default: ``True`` + when :attr:`reduce` is ``False``. Default: ``True`` reduce (bool, optional): Deprecated (see :attr:`reduction`). By default, the losses are averaged or summed over observations for each minibatch depending on :attr:`size_average`. When :attr:`reduce` is ``False``, returns a loss per