未验证 提交 3f372814 编写于 作者: W wanghuancoder 提交者: GitHub

add some no need buff (#42556)

上级 efa21b12
......@@ -567,6 +567,7 @@
param : [x]
kernel :
func : expand_grad
no_need_buffer : x
- backward_api : expm1_grad
forward : expm1 (Tensor x) -> Tensor(out)
......@@ -820,6 +821,7 @@
kernel :
func : layer_norm_grad
data_type : out_grad
no_need_buffer : bias
optional : scale, bias
- backward_api : leaky_relu_double_grad
......@@ -1260,6 +1262,7 @@
param: [x]
kernel :
func : pad3d_grad
no_need_buffer : x
- backward_api : pixel_shuffle_grad
forward : pixel_shuffle (Tensor x, int upscale_factor, str data_format) -> Tensor(out)
......@@ -1409,6 +1412,7 @@
param : [grad_out]
kernel :
func : reshape_double_grad
no_need_buffer : grad_out
- backward_api : reshape_grad
forward : reshape_with_xshape (Tensor x, IntArray shape) -> Tensor(out), Tensor(xshape)
......@@ -1435,6 +1439,7 @@
kernel :
func : roi_align_grad
data_type : boxes
no_need_buffer : x
optional : boxes_num
- backward_api : roi_pool_grad
......@@ -1749,6 +1754,7 @@
param : [x]
kernel :
func : sum_grad
no_need_buffer : x
backward : sum_double_grad
- backward_api : sum_triple_grad
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册