!441 Add AdamApplyOne ir fusion pass

Merge pull request !441 from YuJianfeng/master

mindspore/ccsrc/pre_activate/ascend/ir_fusion/adam_apply_one_fusion.cc

@@ -42,17 +42,69 @@ AnfNodePtr AdamApplyOneFusion::CreateAdamApplyOneNode(const FuncGraphPtr &func_g
 
 const BaseRef AdamApplyOneFusion::DefinePattern() const {
   const auto prim_sqrt = std::make_shared<Primitive>(kSqrtOpName);
-  const auto prim_deal_div = std::make_shared<Primitive>(kRealDivOpName);
+  const auto prim_real_div = std::make_shared<Primitive>(kRealDivOpName);
   VectorRef mul2 = VectorRef({prim::kPrimMul, mul_x_input_vars_[2], input_vars_[1]});
   VectorRef mul3 = VectorRef({prim::kPrimMul, mul_x_input_vars_[3], VectorRef({prim::kPrimSquare, input_vars_[0]})});
   VectorRef sqrt0 = VectorRef({prim_sqrt, VectorRef({add1_var_, mul2, mul3})});
   VectorRef mul1 = VectorRef({prim::kPrimMul, mul_x_input_vars_[1], input_vars_[0]});
   VectorRef mul0 = VectorRef({prim::kPrimMul, mul_x_input_vars_[0], input_vars_[2]});
   VectorRef add0 = VectorRef({add0_var_, mul0, mul1});
-  VectorRef true_div0 = VectorRef({prim_deal_div, add0, VectorRef({prim::kPrimTensorAdd, sqrt0, add2_y_})});
+  VectorRef true_div0 = VectorRef({prim_real_div, add0, VectorRef({prim::kPrimTensorAdd, sqrt0, add2_y_})});
   return VectorRef({prim::kPrimSub, input_vars_[3], VectorRef({prim::kPrimMul, input_vars_[4], true_div0})});
 }
 
+const BaseRef AdamApplyOneCond1Fusion::DefinePattern() const {
+  const auto prim_sqrt = std::make_shared<Primitive>(kSqrtOpName);
+  const auto prim_real_div = std::make_shared<Primitive>(kRealDivOpName);
+  VectorRef mul2 = VectorRef({prim::kPrimMul, mul_x_input_vars_[2], input_vars_[1]});
+  VectorRef mul3 = VectorRef({prim::kPrimMul, mul_x_input_vars_[3], VectorRef({prim::kPrimSquare, input_vars_[0]})});
+  VectorRef sqrt0 = VectorRef({prim_sqrt, VectorRef({add1_var_, mul2, mul3})});
+  VectorRef mul1 = VectorRef({prim::kPrimMul, mul_x_input_vars_[1], input_vars_[0]});
+  VectorRef mul0 = VectorRef({prim::kPrimMul, mul_x_input_vars_[0], input_vars_[2]});
+  VectorRef add0 = VectorRef({add0_var_, mul0, mul1});
+  VectorRef true_div0 = VectorRef({prim_real_div, add0, VectorRef({prim::kPrimTensorAdd, add2_y_, sqrt0})});
+  return VectorRef({prim::kPrimSub, input_vars_[3], VectorRef({prim::kPrimMul, input_vars_[4], true_div0})});
+}
+
+const BaseRef AdamApplyOneCond2Fusion::DefinePattern() const {
+  const auto prim_sqrt = std::make_shared<Primitive>(kSqrtOpName);
+  const auto prim_real_div = std::make_shared<Primitive>(kRealDivOpName);
+  VectorRef mul2 = VectorRef({prim::kPrimMul, mul_x_input_vars_[2], input_vars_[1]});
+  VectorRef mul3 = VectorRef({prim::kPrimMul, VectorRef({prim::kPrimSquare, input_vars_[0]}), mul_x_input_vars_[3]});
+  VectorRef sqrt0 = VectorRef({prim_sqrt, VectorRef({add1_var_, mul2, mul3})});
+  VectorRef mul1 = VectorRef({prim::kPrimMul, mul_x_input_vars_[1], input_vars_[0]});
+  VectorRef mul0 = VectorRef({prim::kPrimMul, mul_x_input_vars_[0], input_vars_[2]});
+  VectorRef add0 = VectorRef({add0_var_, mul0, mul1});
+  VectorRef true_div0 = VectorRef({prim_real_div, add0, VectorRef({prim::kPrimTensorAdd, sqrt0, add2_y_})});
+  return VectorRef({prim::kPrimSub, input_vars_[3], VectorRef({prim::kPrimMul, true_div0, input_vars_[4]})});
+}
+
+const BaseRef AdamApplyOneCond3Fusion::DefinePattern() const {
+  const auto prim_sqrt = std::make_shared<Primitive>(kSqrtOpName);
+  const auto prim_real_div = std::make_shared<Primitive>(kRealDivOpName);
+  VectorRef mul2 = VectorRef({prim::kPrimMul, mul_x_input_vars_[2], input_vars_[1]});
+  VectorRef mul3 = VectorRef({prim::kPrimMul, mul_x_input_vars_[3], VectorRef({prim::kPrimSquare, input_vars_[0]})});
+  VectorRef sqrt0 = VectorRef({prim_sqrt, VectorRef({add1_var_, mul2, mul3})});
+  VectorRef mul1 = VectorRef({prim::kPrimMul, mul_x_input_vars_[1], input_vars_[0]});
+  VectorRef mul0 = VectorRef({prim::kPrimMul, mul_x_input_vars_[0], input_vars_[2]});
+  VectorRef add0 = VectorRef({add0_var_, mul0, mul1});
+  VectorRef true_div0 = VectorRef({prim_real_div, add0, VectorRef({prim::kPrimTensorAdd, sqrt0, add2_y_})});
+  return VectorRef({prim::kPrimSub, input_vars_[3], VectorRef({prim::kPrimMul, true_div0, input_vars_[4]})});
+}
+
+const BaseRef AdamApplyOneCond4Fusion::DefinePattern() const {
+  const auto prim_sqrt = std::make_shared<Primitive>(kSqrtOpName);
+  const auto prim_real_div = std::make_shared<Primitive>(kRealDivOpName);
+  VectorRef mul2 = VectorRef({prim::kPrimMul, mul_x_input_vars_[2], input_vars_[1]});
+  VectorRef mul3 = VectorRef({prim::kPrimMul, mul_x_input_vars_[3], VectorRef({prim::kPrimSquare, input_vars_[0]})});
+  VectorRef sqrt0 = VectorRef({prim_sqrt, VectorRef({add1_var_, mul2, mul3})});
+  VectorRef mul1 = VectorRef({prim::kPrimMul, mul_x_input_vars_[1], input_vars_[0]});
+  VectorRef mul0 = VectorRef({prim::kPrimMul, mul_x_input_vars_[0], input_vars_[2]});
+  VectorRef add0 = VectorRef({add0_var_, mul0, mul1});
+  VectorRef true_div0 = VectorRef({prim_real_div, add0, VectorRef({prim::kPrimTensorAdd, add2_y_, sqrt0})});
+  return VectorRef({prim::kPrimSub, input_vars_[3], VectorRef({prim::kPrimMul, true_div0, input_vars_[4]})});
+}
+
 const AnfNodePtr AdamApplyOneFusion::Process(const FuncGraphPtr &func_graph, const AnfNodePtr &node,
                                              const EquivPtr &equiv) const {
   MS_EXCEPTION_IF_NULL(func_graph);

mindspore/ccsrc/pre_activate/ascend/ir_fusion/adam_apply_one_fusion.h

@@ -18,21 +18,23 @@
 
 #include <vector>
 #include <memory>
+#include <string>
 #include "pre_activate/common/optimizer.h"
 #include "utils/utils.h"
 
 namespace mindspore {
 namespace opt {
-constexpr size_t kAdamApplyOneInputNum = 5;
-constexpr size_t kAdamApplyOneMulInputNum = 4;
+constexpr size_t kAdamApplyOneInputVarNum = 5;
+constexpr size_t kAdamApplyOneMulInputVarNum = 4;
 
 class AdamApplyOneFusion : public PatternProcessPass {
  public:
-  explicit AdamApplyOneFusion(bool multigraph = true) : PatternProcessPass("adam_apply_one_fusion", multigraph) {
-    for (size_t i = 0; i < kAdamApplyOneInputNum; ++i) {
+  explicit AdamApplyOneFusion(const std::string &name = "adam_apply_one_fusion", bool multigraph = true)
+      : PatternProcessPass(name, multigraph) {
+    for (size_t i = 0; i < kAdamApplyOneInputVarNum; ++i) {
       input_vars_.push_back(std::make_shared<Var>());
     }
-    for (size_t i = 0; i < kAdamApplyOneMulInputNum; ++i) {
+    for (size_t i = 0; i < kAdamApplyOneMulInputVarNum; ++i) {
       mul_x_input_vars_.push_back(std::make_shared<Var>());
     }
     add2_y_ = std::make_shared<Var>();
@@ -44,7 +46,7 @@ class AdamApplyOneFusion : public PatternProcessPass {
   const BaseRef DefinePattern() const override;
   const AnfNodePtr Process(const FuncGraphPtr &, const AnfNodePtr &, const EquivPtr &) const override;
 
- private:
+ protected:
   AnfNodePtr CreateAdamApplyOneNode(const FuncGraphPtr &func_graph, const EquivPtr &equiv) const;
   std::vector<VarPtr> input_vars_;
   std::vector<VarPtr> mul_x_input_vars_;
@@ -52,6 +54,42 @@ class AdamApplyOneFusion : public PatternProcessPass {
   VarPtr add0_var_;
   VarPtr add1_var_;
 };
+
+class AdamApplyOneCond1Fusion : public AdamApplyOneFusion {
+ public:
+  explicit AdamApplyOneCond1Fusion(bool multigraph = true)
+      : AdamApplyOneFusion("adam_apply_one_cond1_fusion", multigraph) {}
+
+  ~AdamApplyOneCond1Fusion() override = default;
+  const BaseRef DefinePattern() const override;
+};
+
+class AdamApplyOneCond2Fusion : public AdamApplyOneFusion {
+ public:
+  explicit AdamApplyOneCond2Fusion(bool multigraph = true)
+      : AdamApplyOneFusion("adam_apply_one_cond2_fusion", multigraph) {}
+
+  ~AdamApplyOneCond2Fusion() override = default;
+  const BaseRef DefinePattern() const override;
+};
+
+class AdamApplyOneCond3Fusion : public AdamApplyOneFusion {
+ public:
+  explicit AdamApplyOneCond3Fusion(bool multigraph = true)
+      : AdamApplyOneFusion("adam_apply_one_cond3_fusion", multigraph) {}
+
+  ~AdamApplyOneCond3Fusion() override = default;
+  const BaseRef DefinePattern() const override;
+};
+
+class AdamApplyOneCond4Fusion : public AdamApplyOneFusion {
+ public:
+  explicit AdamApplyOneCond4Fusion(bool multigraph = true)
+      : AdamApplyOneFusion("adam_apply_one_cond4_fusion", multigraph) {}
+
+  ~AdamApplyOneCond4Fusion() override = default;
+  const BaseRef DefinePattern() const override;
+};
 }  // namespace opt
 }  // namespace mindspore
 #endif  // MINDSPORE_CCSRC_PRE_ACTIVATE_ASCEND_IR_FUSION_ADAM_APPLY_ONE_FUSION_H_

tests/ut/cpp/pre_activate/ascend/ir_fusion/adam_apply_one_fusion_test.cc

@@ -66,5 +66,156 @@ TEST_F(TestHWAdamApplyOneFusion, test_adam_apply_one_fusion) {
   EXPECT_TRUE(CheckEqualGraph(g_after, new_graph));
 }
 
+TEST_F(TestHWAdamApplyOneFusion, test_adam_apply_one_cond1_fusion) {
+  /*
+   * def before_cond1(input0, input1, input2, input3, input4, mul0_x, mul1_x, mul2_x, mul3_x, add2_y):
+   *    square0 = Square(input0)
+   *    mul1 = Mul(mul1_x, input0)
+   *    mul0 = Mul(mul0_x, input2)
+   *    mul2 = Mul(mul2_x, input1)
+   *    mul3 = Mul(mul3_x, square0)
+   *    add0 = Add(mul0, mul1)
+   *    add1 = Add(mul2, mul3)
+   *    sqrt0 = Sqrt(add1)
+   *    add2 = Add(add2_y, sqrt0)
+   *    true_div0 = RealDiv(add0, add2)
+   *    mul4 = Mul(input4, true_div0)
+   *    sub0 = Sub(input3, mul4)
+   *    outputs = make_tuple(add1, add0, sub0)
+   *    output = tuple_getitem(outputs, 0)
+   *    return output
+   */
+  FuncGraphPtr g = get_py_fun_.CallAndParseRet("test_adam_apply_one_fusion", "before_cond1");
+  std::vector<int> shp{2, 32, 224, 224};
+  auto x_abstract = std::make_shared<abstract::AbstractTensor>(kFloat32, shp);
+  AbstractBasePtrList args_spec_list;
+  for (size_t i = 0; i < 10; ++i) {
+    args_spec_list.push_back(x_abstract);
+  }
+  auto fg = GetKernelGraph(g, args_spec_list);
+
+  auto optimizer = std::make_shared<opt::GraphOptimizer>();
+  auto pm = std::make_shared<opt::PassManager>();
+  pm->AddPass(std::make_shared<opt::AdamApplyOneCond1Fusion>());
+  optimizer->AddPassManager(pm);
+  FuncGraphPtr new_graph = optimizer->Optimize(fg);
+
+  FuncGraphPtr g_after = get_py_fun_.CallAndParseRet("test_adam_apply_one_fusion", "after");
+  EXPECT_TRUE(CheckEqualGraph(g_after, new_graph));
+}
+
+TEST_F(TestHWAdamApplyOneFusion, test_adam_apply_one_cond2_fusion) {
+  /*
+   * def before_cond2(input0, input1, input2, input3, input4, mul0_x, mul1_x, mul2_x, mul3_x, add2_y):
+   *    square0 = Square(input0)
+   *    mul1 = Mul(mul1_x, input0)
+   *    mul0 = Mul(mul0_x, input2)
+   *    mul2 = Mul(mul2_x, input1)
+   *    mul3 = Mul(square0, mul3_x)
+   *    add0 = Add(mul0, mul1)
+   *    add1 = Add(mul2, mul3)
+   *    sqrt0 = Sqrt(add1)
+   *    add2 = Add(sqrt0, add2_y)
+   *    true_div0 = RealDiv(add0, add2)
+   *    mul4 = Mul(true_div0, input4)
+   *    sub0 = Sub(input3, mul4)
+   *    outputs = make_tuple(add1, add0, sub0)
+   *    output = tuple_getitem(outputs, 0)
+   *    return output
+   */
+  FuncGraphPtr g = get_py_fun_.CallAndParseRet("test_adam_apply_one_fusion", "before_cond2");
+  std::vector<int> shp{2, 32, 224, 224};
+  auto x_abstract = std::make_shared<abstract::AbstractTensor>(kFloat32, shp);
+  AbstractBasePtrList args_spec_list;
+  for (size_t i = 0; i < 10; ++i) {
+    args_spec_list.push_back(x_abstract);
+  }
+  auto fg = GetKernelGraph(g, args_spec_list);
+
+  auto optimizer = std::make_shared<opt::GraphOptimizer>();
+  auto pm = std::make_shared<opt::PassManager>();
+  pm->AddPass(std::make_shared<opt::AdamApplyOneCond2Fusion>());
+  optimizer->AddPassManager(pm);
+  FuncGraphPtr new_graph = optimizer->Optimize(fg);
+
+  FuncGraphPtr g_after = get_py_fun_.CallAndParseRet("test_adam_apply_one_fusion", "after");
+  EXPECT_TRUE(CheckEqualGraph(g_after, new_graph));
+}
+
+TEST_F(TestHWAdamApplyOneFusion, test_adam_apply_one_cond3_fusion) {
+  /*
+   * def before_cond3(input0, input1, input2, input3, input4, mul0_x, mul1_x, mul2_x, mul3_x, add2_y):
+   *    square0 = Square(input0)
+   *    mul1 = Mul(mul1_x, input0)
+   *    mul0 = Mul(mul0_x, input2)
+   *    mul2 = Mul(mul2_x, input1)
+   *    mul3 = Mul(mul3_x, square0)
+   *    add0 = Add(mul0, mul1)
+   *    add1 = Add(mul2, mul3)
+   *    sqrt0 = Sqrt(add1)
+   *    add2 = Add(sqrt0, add2_y)
+   *    true_div0 = RealDiv(add0, add2)
+   *    mul4 = Mul(true_div0, input4)
+   *    sub0 = Sub(input3, mul4)
+   *    outputs = make_tuple(add1, add0, sub0)
+   *    output = tuple_getitem(outputs, 0)
+   *    return output
+   */
+  FuncGraphPtr g = get_py_fun_.CallAndParseRet("test_adam_apply_one_fusion", "before_cond3");
+  std::vector<int> shp{2, 32, 224, 224};
+  auto x_abstract = std::make_shared<abstract::AbstractTensor>(kFloat32, shp);
+  AbstractBasePtrList args_spec_list;
+  for (size_t i = 0; i < 10; ++i) {
+    args_spec_list.push_back(x_abstract);
+  }
+  auto fg = GetKernelGraph(g, args_spec_list);
+
+  auto optimizer = std::make_shared<opt::GraphOptimizer>();
+  auto pm = std::make_shared<opt::PassManager>();
+  pm->AddPass(std::make_shared<opt::AdamApplyOneCond3Fusion>());
+  optimizer->AddPassManager(pm);
+  FuncGraphPtr new_graph = optimizer->Optimize(fg);
+
+  FuncGraphPtr g_after = get_py_fun_.CallAndParseRet("test_adam_apply_one_fusion", "after");
+  EXPECT_TRUE(CheckEqualGraph(g_after, new_graph));
+}
+
+TEST_F(TestHWAdamApplyOneFusion, test_adam_apply_one_cond4_fusion) {
+  /*
+   * def before_cond4(input0, input1, input2, input3, input4, mul0_x, mul1_x, mul2_x, mul3_x, add2_y):
+   *    square0 = Square(input0)
+   *    mul1 = Mul(mul1_x, input0)
+   *    mul0 = Mul(mul0_x, input2)
+   *    mul2 = Mul(mul2_x, input1)
+   *    mul3 = Mul(mul3_x, square0)
+   *    add0 = Add(mul0, mul1)
+   *    add1 = Add(mul2, mul3)
+   *    sqrt0 = Sqrt(add1)
+   *    add2 = Add(add2_y, sqrt0)
+   *    true_div0 = RealDiv(add0, add2)
+   *    mul4 = Mul(true_div0, input4)
+   *    sub0 = Sub(input3, mul4)
+   *    outputs = make_tuple(add1, add0, sub0)
+   *    output = tuple_getitem(outputs, 0)
+   *    return output
+   */
+  FuncGraphPtr g = get_py_fun_.CallAndParseRet("test_adam_apply_one_fusion", "before_cond4");
+  std::vector<int> shp{2, 32, 224, 224};
+  auto x_abstract = std::make_shared<abstract::AbstractTensor>(kFloat32, shp);
+  AbstractBasePtrList args_spec_list;
+  for (size_t i = 0; i < 10; ++i) {
+    args_spec_list.push_back(x_abstract);
+  }
+  auto fg = GetKernelGraph(g, args_spec_list);
+
+  auto optimizer = std::make_shared<opt::GraphOptimizer>();
+  auto pm = std::make_shared<opt::PassManager>();
+  pm->AddPass(std::make_shared<opt::AdamApplyOneCond4Fusion>());
+  optimizer->AddPassManager(pm);
+  FuncGraphPtr new_graph = optimizer->Optimize(fg);
+
+  FuncGraphPtr g_after = get_py_fun_.CallAndParseRet("test_adam_apply_one_fusion", "after");
+  EXPECT_TRUE(CheckEqualGraph(g_after, new_graph));
+}
 }  // namespace opt
 }  // namespace mindspore

tests/ut/cpp/python_input/gtest_input/pre_activate/adam_apply_one_fusion_test.py

@@ -58,6 +58,78 @@ def test_adam_apply_one_fusion(tag):
         output = tuple_getitem(outputs, 0)
         return output
 
+    @fns
+    def before_cond1(input0, input1, input2, input3, input4, mul0_x, mul1_x, mul2_x, mul3_x, add2_y):
+        square0 = Square(input0)
+        mul1 = Mul(mul1_x, input0)
+        mul0 = Mul(mul0_x, input2)
+        mul2 = Mul(mul2_x, input1)
+        mul3 = Mul(mul3_x, square0)
+        add0 = Add(mul0, mul1)
+        add1 = Add(mul2, mul3)
+        sqrt0 = Sqrt(add1)
+        add2 = Add(add2_y, sqrt0)
+        true_div0 = RealDiv(add0, add2)
+        mul4 = Mul(input4, true_div0)
+        sub0 = Sub(input3, mul4)
+        outputs = make_tuple(add1, add0, sub0)
+        output = tuple_getitem(outputs, 0)
+        return output
+
+    @fns
+    def before_cond2(input0, input1, input2, input3, input4, mul0_x, mul1_x, mul2_x, mul3_x, add2_y):
+        square0 = Square(input0)
+        mul1 = Mul(mul1_x, input0)
+        mul0 = Mul(mul0_x, input2)
+        mul2 = Mul(mul2_x, input1)
+        mul3 = Mul(square0, mul3_x)
+        add0 = Add(mul0, mul1)
+        add1 = Add(mul2, mul3)
+        sqrt0 = Sqrt(add1)
+        add2 = Add(sqrt0, add2_y)
+        true_div0 = RealDiv(add0, add2)
+        mul4 = Mul(true_div0, input4)
+        sub0 = Sub(input3, mul4)
+        outputs = make_tuple(add1, add0, sub0)
+        output = tuple_getitem(outputs, 0)
+        return output
+
+    @fns
+    def before_cond3(input0, input1, input2, input3, input4, mul0_x, mul1_x, mul2_x, mul3_x, add2_y):
+        square0 = Square(input0)
+        mul1 = Mul(mul1_x, input0)
+        mul0 = Mul(mul0_x, input2)
+        mul2 = Mul(mul2_x, input1)
+        mul3 = Mul(mul3_x, square0)
+        add0 = Add(mul0, mul1)
+        add1 = Add(mul2, mul3)
+        sqrt0 = Sqrt(add1)
+        add2 = Add(sqrt0, add2_y)
+        true_div0 = RealDiv(add0, add2)
+        mul4 = Mul(true_div0, input4)
+        sub0 = Sub(input3, mul4)
+        outputs = make_tuple(add1, add0, sub0)
+        output = tuple_getitem(outputs, 0)
+        return output
+
+    @fns
+    def before_cond4(input0, input1, input2, input3, input4, mul0_x, mul1_x, mul2_x, mul3_x, add2_y):
+        square0 = Square(input0)
+        mul1 = Mul(mul1_x, input0)
+        mul0 = Mul(mul0_x, input2)
+        mul2 = Mul(mul2_x, input1)
+        mul3 = Mul(mul3_x, square0)
+        add0 = Add(mul0, mul1)
+        add1 = Add(mul2, mul3)
+        sqrt0 = Sqrt(add1)
+        add2 = Add(add2_y, sqrt0)
+        true_div0 = RealDiv(add0, add2)
+        mul4 = Mul(true_div0, input4)
+        sub0 = Sub(input3, mul4)
+        outputs = make_tuple(add1, add0, sub0)
+        output = tuple_getitem(outputs, 0)
+        return output
+
     @fns
     def after(input0, input1, input2, input3, input4, mul0_x, mul1_x, mul2_x, mul3_x, add2_y):
         adam_apply_one = AdamApplyOne(input0, input1, input2, input3, input4, mul0_x, mul1_x, mul2_x, mul3_x, add2_y)