Fix

Author: co63oc

paddle/phi/ops/yaml/inconsistent/dygraph_backward.yaml

@@ -94,6 +94,16 @@
     func : c_embedding_grad
   no_need_buffer : weight
 
+- backward_op : c_softmax_with_cross_entropy_grad
+  forward: c_softmax_with_cross_entropy (Tensor logits, Tensor label,  int64_t ignore_index=-100, int ring_id=0, int rank=0, int nranks=0) -> Tensor(softmax), Tensor(loss)
+  args: (Tensor softmax, Tensor label, Tensor loss_grad,int64_t ignore_index=-100, int ring_id=0, int rank=0, int nranks=0)
+  output: Tensor(logits_grad)
+  infer_meta :
+    func: CSoftmaxWithCrossEntropyGradInferMeta
+  kernel:
+    func: c_softmax_with_cross_entropy_grad
+    data_type: loss_grad
+
 - backward_op : divide_double_grad
   forward : divide_grad (Tensor x, Tensor y, Tensor out, Tensor grad_out, int axis = -1) -> Tensor(grad_x), Tensor(grad_y)
   args : (Tensor y, Tensor out, Tensor grad_out, Tensor grad_x, Tensor grad_x_grad, Tensor grad_y_grad, int axis = -1)
@@ -277,16 +287,6 @@
     func: set_value_with_scalar_grad
     param: [out_grad, starts, ends, steps, axes, decrease_axes, none_axes]
 
-- backward_op : c_softmax_with_cross_entropy_grad
-  forward: c_softmax_with_cross_entropy (Tensor logits, Tensor label,  int64_t ignore_index=-100, int ring_id=0, int rank=0, int nranks=0) -> Tensor(softmax), Tensor(loss)
-  args: (Tensor softmax, Tensor label, Tensor loss_grad,int64_t ignore_index=-100, int ring_id=0, int rank=0, int nranks=0)
-  output: Tensor(logits_grad)
-  infer_meta :
-    func: CSoftmaxWithCrossEntropyGradInferMeta
-  kernel:
-    func: c_softmax_with_cross_entropy_grad
-    data_type: loss_grad
-
 - backward_op : softmax_grad
   forward : softmax (Tensor x, int axis) -> Tensor(out)
   args : (Tensor out, Tensor out_grad, int axis)