add adaptive maxpool3d grad gpu vmap rule

mindspore/core/ops/cumprod.h

@@ -32,10 +32,8 @@ class MIND_API CumProd : public BaseOperator {
   void Init(const bool exclusive, const bool reverse);
   void SetExclusive(const bool exclusive);
   void SetReverse(const bool reverse);
-  void SetAxis(const int64_t axis);
   bool GetExclusive() const;
   bool GetReverse() const;
-  int64_t GetAxis() const;
 };
 abstract::AbstractBasePtr CumProdInfer(const abstract::AnalysisEnginePtr &, const PrimitivePtr &primitive,
                                        const std::vector<abstract::AbstractBasePtr> &input_args);

mindspore/core/ops/cumsum.h

@@ -38,8 +38,6 @@ class MIND_API CumSum : public BaseOperator {
   void set_exclusive(const bool exclusive);
   /// \brief Set reverse.
   void set_reverse(const bool reverse);
-  /// \brief Set axis.
-  void set_axis(const int64_t axis);
   /// \brief Get exclusive.
   ///
   /// \return exclusive.
@@ -49,8 +47,6 @@ class MIND_API CumSum : public BaseOperator {
   /// \return reverse.
   bool get_reverse() const;
   ///
-  /// \return axis.
-  int64_t get_axis() const;
 };
 abstract::AbstractBasePtr CumSumInfer(const abstract::AnalysisEnginePtr &, const PrimitivePtr &primitive,
                                       const std::vector<abstract::AbstractBasePtr> &input_args);

mindspore/python/mindspore/ops/_vmap/vmap_grad_nn_ops.py

@@ -226,11 +226,15 @@ def get_cdist_grad_vmap_rule(prim, axis_size):
     return vmap_rule
 
 
+@vmap_rules_getters.register(G.AdaptiveMaxPool3DGrad)
 @vmap_rules_getters.register(G.AdaptiveMaxPool2DGrad)
 def get_adaptive_avgpool2d_vmap_rule(prim, axis_size):
-    """VmapRule for `AdaptiveMaxPool2DGrad` operation."""
+    """VmapRule for `AdaptiveMaxPool2DGrad` and `AdaptiveMaxPool3DGrad` operation."""
     chw_reverse_index = -3
-    hw_reverse_index = -2
+    if prim.name == "AdaptiveMaxPool2DGrad":
+        hw_reverse_index = -2
+    else:
+        hw_reverse_index = -3
 
     def vmap_rule(ygrad_bdim, x_bdim, max_index_bdim):
         is_all_none, result = vmap_general_preprocess(prim, ygrad_bdim, x_bdim, max_index_bdim)
@@ -353,7 +357,7 @@ def get_batchnorm_grad_vmap_rule(prim, axis_size):
         if is_all_none:
             return result
         if data_format == "NHWC":
-            #BatchNormGrad with NHWC format is a GPU backend operation and not supported for now.
+            # BatchNormGrad with NHWC format is a GPU backend operation and not supported for now.
             return batchnorm_grad_nhwc_vmap(grad_bdim, x_bdim, scale_bdim, rsv_1_bdim, rsv_2_bdim, rsv_3_bdim)
         grad, grad_dim = grad_bdim
         input_x, input_x_dim = x_bdim

requirements.txt

@@ -3,6 +3,7 @@ protobuf >= 3.13.0
 asttokens >= 2.0.4
 pillow >= 6.2.0
 scipy >= 1.5.4
+decorator >= 4.4.0
 matplotlib >= 3.1.3         # for ut test
 opencv-python >= 4.1.2.30   # for ut test
 sklearn >= 0.0              # for st test