OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 00:20:18 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity

fix more derivatives

Browse Source 
ghstack-source-id: 86068308f3
Pull Request resolved: https://github.com/pytorch/pytorch/pull/29677
This commit is contained in:
Vitaly Fedyunin 2019-11-18 08:28:00 -08:00
parent 9af850845f
commit 5c9582d681
1 changed files with 6 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
tools/autograd/derivatives.yaml
View File

@ -507,7 +507,7 @@
- name: masked_fill_.Tensor(Tensor(a!) self, Tensor mask, Tensor value) -> Tensor(a!)
self: grad.clone().masked_fill_(mask, 0)
value: at::where(mask, grad, zeros_like(grad)).sum()
value: at::where(mask, grad, zeros_like(grad, at::MemoryFormat::Preserve)).sum()
mask: non_differentiable
- name: masked_scatter_(Tensor(a!) self, Tensor mask, Tensor source) -> Tensor(a!)
@ -519,7 +519,7 @@
# normally broadcasting is handled implicitly, but here, because we call an inplace
# function as an optimization and the LHS doesn't broadcast for inplace functions,
# we need to explicitly broadcast.
self: zeros_like(self.expand(at::infer_size(self.sizes(), mask.sizes()))).masked_scatter_(mask, grad)
self: zeros_like(self.expand(at::infer_size(self.sizes(), mask.sizes())), at::MemoryFormat::Preserve).masked_scatter_(mask, grad)
mask: non_differentiable
- name: max.dim(Tensor self, int dim, bool keepdim=False) -> (Tensor values, Tensor indices)
@ -685,7 +685,7 @@
self: prod_backward(grad, self.to(grad.scalar_type()), result, dim, keepdim)
- name: put_(Tensor(a!) self, Tensor index, Tensor source, bool accumulate=False) -> Tensor(a!)
self: grad.clone().put_(index, zeros_like(source), accumulate)
self: grad.clone().put_(index, zeros_like(source, at::MemoryFormat::Preserve), accumulate)
index: non_differentiable
source: grad.take(index)
@ -917,8 +917,8 @@
- name: _s_where(Tensor condition, Tensor self, Tensor other) -> Tensor
condition: non_differentiable
self: where(condition, grad, zeros_like(grad))
other: where(condition, zeros_like(grad), grad)
self: where(condition, grad, zeros_like(grad, at::MemoryFormat::Preserve))
other: where(condition, zeros_like(grad, at::MemoryFormat::Preserve), grad)
# weight_norm_cuda_interface_backward does not have an explicitly defined derivative, so if we do happen
# to be running backward with create_graph=True, fall back to a backward function that uses
@ -1276,7 +1276,7 @@
- name: kl_div_backward(Tensor grad_output, Tensor self, Tensor target, int reduction=Mean) -> Tensor
grad_output: kl_div_double_backward_grad_output(grad, self, target, reduction)
self: zeros_like(grad, at::MemoryFormat::Preserve)
target: zeros_like(grad)
target: zeros_like(grad, at::MemoryFormat::Preserve)
- name: l1_loss_backward(Tensor grad_output, Tensor self, Tensor target, int reduction) -> Tensor
grad_output: l1_loss_double_backward_grad_output(grad, self, target, reduction)