[NewIR]Split python api and vjp (#56518)

* support ir api form prim

* convert vector of int to intarray

* add reference of lbfgs

* add reference of lbfgs

* support ir api for prim

* Add more gen api

* concat python api to concat_grad

* fix gen conflict

* support vjp prim mode in new ir

* remove useless code

* add vjp autogen v1.0

* add test for prim

* resolve type conflict

* modify utils

* remove useless code

* add split op and modify some bug of vectorType

* fix conflict

* add concat python test

* add split python api to vjp

* modify build bug

* modify run bug

* fix conflict bug

* build bug fix

* modify python api bug

* modify test

* fix conflict

* fluid backward recover

* recover conflict

* reply review comments

* modify opruntimeinfo num

---------
Co-authored-by: Ncyber-pioneer <chenzhuo@tju.edu.cn>
Co-authored-by: NCharles-hit <wanghao107@baidu.com>
Co-authored-by: N0x45f <wangzhen45@baidu.com>
Co-authored-by: Nchenzhiyang <1792266893@qq.com>
Co-authored-by: NChen Zhiyang <chenzhiyang99@126.com>

paddle/fluid/ir/dialect/op_generator/op_gen.py

@@ -172,7 +172,7 @@ scalar_type_maps = {
     'bool': 'ir::BoolAttribute',
 }
 
-_NO_NEED_GEN_OPS = {'add_n'}
+_NO_NEED_GEN_OPS = {'add_n', 'split_grad'}
 
 
 def to_phi_and_fluid_op_name(op_item):

paddle/fluid/ir/dialect/op_generator/vjp_interface_gen_op_list.py

@@ -29,6 +29,7 @@ vjp_interface_declare_gen_op_list = [
     "sum",
     "add",
     "concat",
+    "split",
 ]
 vjp_interface_implementation_gen_op_list = [
     "tanh",

paddle/fluid/ir/dialect/paddle_dialect/ir/pd_dialect.cc

@@ -48,7 +48,7 @@ void PaddleDialect::initialize() {
 #define GET_OP_LIST
 #include "paddle/fluid/ir/dialect/paddle_dialect/ir/pd_op.h"  // NOLINT
       >();
-  RegisterOp<paddle::dialect::AddNOp>();
+  RegisterOps<paddle::dialect::AddNOp, paddle::dialect::SplitGradOp>();
 
   RegisterInterfaces<ParameterConvertInterface>();
 }

paddle/fluid/ir/dialect/paddle_dialect/ir/pd_manual_api.cc

@@ -18,5 +18,16 @@
 #include "paddle/ir/core/builtin_op.h"
 
 namespace paddle {
-namespace dialect {}  // namespace dialect
+namespace dialect {
+ir::OpResult split_grad(std::vector<ir::OpResult> out_grads,
+                        ir::OpResult axis) {
+  auto combine_op =
+      APIBuilder::Instance().GetBuilder()->Build<ir::CombineOp>(out_grads);
+  paddle::dialect::SplitGradOp split_grad_op =
+      APIBuilder::Instance().GetBuilder()->Build<paddle::dialect::SplitGradOp>(
+          combine_op.out(), axis);
+
+  return split_grad_op.x_grad();
+}
+}  // namespace dialect
 }  // namespace paddle

paddle/fluid/ir/dialect/paddle_dialect/ir/pd_manual_api.h

@@ -21,5 +21,9 @@
 #include "paddle/phi/common/place.h"
 
 namespace paddle {
-namespace dialect {}  // namespace dialect
+namespace dialect {
+
+ir::OpResult split_grad(std::vector<ir::OpResult> out_grads, ir::OpResult axis);
+
+}  // namespace dialect
 }  // namespace paddle

paddle/fluid/ir/dialect/paddle_dialect/ir/pd_manual_op.cc

@@ -14,6 +14,7 @@
 
 #include "paddle/fluid/ir/dialect/paddle_dialect/ir/pd_manual_op.h"
 #include "paddle/fluid/ir/dialect/paddle_dialect/ir/pd_attribute.h"
+#include "paddle/fluid/ir/dialect/paddle_dialect/ir/pd_op.h"
 #include "paddle/fluid/ir/dialect/paddle_dialect/ir/pd_type.h"
 #include "paddle/ir/core/builtin_attribute.h"
 #include "paddle/ir/core/builtin_op.h"
@@ -145,7 +146,221 @@ void AddNOp::InferMeta(phi::InferMetaContext *infer_meta) {
   fn(infer_meta);
 }
 
+const char *SplitGradOp::attributes_name[1] = {"axis"};
+
+OpInfoTuple SplitGradOp::GetOpInfo() {
+  std::vector<paddle::dialect::OpInputInfo> inputs = {
+      OpInputInfo("out_grad",
+                  "ir::VectorType<paddle::dialect::DenseTensorType>",
+                  false,
+                  false,
+                  false),
+      OpInputInfo(
+          "axis", "paddle::dialect::ScalarAttribute", false, false, true)};
+  std::vector<paddle::dialect::OpAttributeInfo> attributes = {};
+  std::vector<paddle::dialect::OpOutputInfo> outputs = {
+      OpOutputInfo("x_grad", "paddle::dialect::DenseTensorType", false, false)};
+  paddle::dialect::OpRunTimeInfo run_time_info =
+      OpRunTimeInfo("ConcatInferMeta",
+                    {"out_grad", "axis"},
+                    {"concat"},
+                    {"out_grad", "axis"},
+                    {"out_grad"},
+                    {},
+                    {},
+                    {});
+
+  return std::make_tuple(
+      inputs, attributes, outputs, run_time_info, "split_grad");
+}
+
+void SplitGradOp::Build(ir::Builder &builder,
+                        ir::OperationArgument &argument,
+                        ir::OpResult out_grad_,
+                        float axis) {
+  // Generate scalar mutable attribute: axis
+  paddle::dialect::FullOp full_axis_op = builder.Build<paddle::dialect::FullOp>(
+      std::vector<int64_t>{1}, axis, phi::DataType::FLOAT32, phi::CPUPlace());
+  ir::OpResult axis_ = full_axis_op->result(0);
+
+  VLOG(4) << "Builder construction inputs";
+  std::vector<ir::OpResult> argument_inputs = {out_grad_, axis_};
+  argument.AddOperands(argument_inputs.begin(), argument_inputs.end());
+
+  VLOG(4) << "Builder construction attributes";
+
+  VLOG(4) << "Builder construction outputs";
+  ir::VectorType out_grad = out_grad_.type().dyn_cast<ir::VectorType>();
+  std::vector<phi::DenseTensor> vec_dense_out_grad;
+  for (size_t i = 0; i < static_cast<size_t>(out_grad.size()); i++) {
+    vec_dense_out_grad.push_back(phi::DenseTensor(
+        std::make_unique<paddle::experimental::DefaultAllocator>(
+            paddle::platform::CPUPlace())
+            .get(),
+        phi::DenseTensorMeta(
+            paddle::dialect::TransToPhiDataType(
+                out_grad[i]
+                    .dyn_cast<paddle::dialect::DenseTensorType>()
+                    .dtype()),
+            out_grad[i].dyn_cast<paddle::dialect::DenseTensorType>().dims(),
+            out_grad[i]
+                .dyn_cast<paddle::dialect::DenseTensorType>()
+                .data_layout(),
+            out_grad[i].dyn_cast<paddle::dialect::DenseTensorType>().lod(),
+            out_grad[i]
+                .dyn_cast<paddle::dialect::DenseTensorType>()
+                .offset())));
+  }
+  std::vector<phi::MetaTensor> vec_meta_out_grad;
+  for (size_t i = 0; i < vec_dense_out_grad.size(); i++) {
+    vec_meta_out_grad.push_back(phi::MetaTensor(&vec_dense_out_grad[i]));
+  }
+
+  std::vector<const phi::MetaTensor *> meta_out_grad;
+  for (size_t i = 0; i < static_cast<size_t>(vec_meta_out_grad.size()); i++) {
+    meta_out_grad.push_back(&vec_meta_out_grad[i]);
+  }
+  phi::DenseTensor dense_x_grad;
+  phi::MetaTensor meta_x_grad(&dense_x_grad);
+
+  phi::ConcatInferMeta(meta_out_grad, axis, &meta_x_grad);
+
+  std::vector<ir::Type> argument_outputs;
+  ir::Type x_grad_dense_tensor_type = paddle::dialect::DenseTensorType::get(
+      ir::IrContext::Instance(),
+      paddle::dialect::TransToIrDataType(dense_x_grad.dtype()),
+      dense_x_grad.dims(),
+      dense_x_grad.layout(),
+      dense_x_grad.lod(),
+      dense_x_grad.offset());
+  argument_outputs.push_back(x_grad_dense_tensor_type);
+  argument.AddOutputs(argument_outputs.begin(), argument_outputs.end());
+}
+
+void SplitGradOp::Build(ir::Builder &builder,
+                        ir::OperationArgument &argument,
+                        ir::OpResult out_grad_,
+                        ir::OpResult axis_) {
+  VLOG(4) << "Builder construction inputs";
+  std::vector<ir::OpResult> argument_inputs = {out_grad_, axis_};
+  argument.AddOperands(argument_inputs.begin(), argument_inputs.end());
+
+  VLOG(4) << "Builder construction attributes";
+
+  VLOG(4) << "Builder construction outputs";
+  ir::VectorType out_grad = out_grad_.type().dyn_cast<ir::VectorType>();
+  int axis = axis_.owner()
+                 ->dyn_cast<paddle::dialect::FullOp>()
+                 .attributes()
+                 .at("value")
+                 .dyn_cast<paddle::dialect::ScalarAttribute>()
+                 .data()
+                 .to<int>();
+
+  std::vector<phi::DenseTensor> vec_dense_out_grad;
+  for (size_t i = 0; i < static_cast<size_t>(out_grad.size()); i++) {
+    vec_dense_out_grad.push_back(phi::DenseTensor(
+        std::make_unique<paddle::experimental::DefaultAllocator>(
+            paddle::platform::CPUPlace())
+            .get(),
+        phi::DenseTensorMeta(
+            TransToPhiDataType(out_grad[i]
+                                   .dyn_cast<paddle::dialect::DenseTensorType>()
+                                   .dtype()),
+            out_grad[i].dyn_cast<paddle::dialect::DenseTensorType>().dims(),
+            out_grad[i]
+                .dyn_cast<paddle::dialect::DenseTensorType>()
+                .data_layout(),
+            out_grad[i].dyn_cast<paddle::dialect::DenseTensorType>().lod(),
+            out_grad[i]
+                .dyn_cast<paddle::dialect::DenseTensorType>()
+                .offset())));
+  }
+  std::vector<phi::MetaTensor> vec_meta_out_grad;
+  for (size_t i = 0; i < vec_dense_out_grad.size(); i++) {
+    vec_meta_out_grad.push_back(phi::MetaTensor(&vec_dense_out_grad[i]));
+  }
+
+  std::vector<const phi::MetaTensor *> meta_out_grad;
+  for (size_t i = 0; i < static_cast<size_t>(vec_meta_out_grad.size()); i++) {
+    meta_out_grad.push_back(&vec_meta_out_grad[i]);
+  }
+  phi::DenseTensor dense_x_grad;
+  phi::MetaTensor meta_x_grad(&dense_x_grad);
+
+  phi::ConcatInferMeta(meta_out_grad, axis, &meta_x_grad);
+
+  std::vector<ir::Type> argument_outputs;
+  ir::Type x_grad_dense_tensor_type = paddle::dialect::DenseTensorType::get(
+      ir::IrContext::Instance(),
+      TransToIrDataType(dense_x_grad.dtype()),
+      dense_x_grad.dims(),
+      dense_x_grad.layout(),
+      dense_x_grad.lod(),
+      dense_x_grad.offset());
+  argument_outputs.push_back(x_grad_dense_tensor_type);
+  argument.AddOutputs(argument_outputs.begin(), argument_outputs.end());
+}
+
+void SplitGradOp::Verify() {
+  VLOG(4) << "Start Verifying inputs, outputs and attributes for: SplitGradOp.";
+  VLOG(4) << "Verifying inputs:";
+  {
+    auto input_size = num_operands();
+    PADDLE_ENFORCE_EQ(
+        input_size,
+        2u,
+        phi::errors::PreconditionNotMet(
+            "The size %d of inputs must be equal to 2.", input_size));
+    if (auto vec_type =
+            (*this)->operand_source(0).type().dyn_cast<ir::VectorType>()) {
+      for (size_t i = 0; i < vec_type.size(); ++i) {
+        PADDLE_ENFORCE(vec_type[i].isa<paddle::dialect::DenseTensorType>(),
+                       phi::errors::PreconditionNotMet(
+                           "Type validation failed for the 0th input."));
+      }
+    } else {
+      PADDLE_ENFORCE((*this)
+                         ->operand_source(0)
+                         .type()
+                         .isa<paddle::dialect::DenseTensorType>(),
+                     phi::errors::PreconditionNotMet(
+                         "Type validation failed for the 0th input."));
+    }
+    PADDLE_ENFORCE((*this)
+                       ->operand_source(1)
+                       .type()
+                       .isa<paddle::dialect::DenseTensorType>(),
+                   phi::errors::PreconditionNotMet(
+                       "Type validation failed for the 1th input."));
+  }
+  VLOG(4) << "Verifying attributes:";
+  {
+    // Attributes num is 0, not need to check attributes type.
+  }
+  VLOG(4) << "Verifying outputs:";
+  {
+    auto output_size = num_results();
+    PADDLE_ENFORCE_EQ(
+        output_size,
+        1u,
+        phi::errors::PreconditionNotMet(
+            "The size %d of outputs must be equal to 1.", output_size));
+    PADDLE_ENFORCE(
+        (*this)->result(0).type().isa<paddle::dialect::DenseTensorType>(),
+        phi::errors::PreconditionNotMet(
+            "Type validation failed for the 0th output."));
+  }
+  VLOG(4) << "End Verifying for: SplitGradOp.";
+}
+
+void SplitGradOp::InferMeta(phi::InferMetaContext *infer_meta) {
+  auto fn = PD_INFER_META(phi::ConcatInferMeta);
+  fn(infer_meta);
+}
+
 }  // namespace dialect
 }  // namespace paddle
 
 IR_DEFINE_EXPLICIT_TYPE_ID(paddle::dialect::AddNOp)
+IR_DEFINE_EXPLICIT_TYPE_ID(paddle::dialect::SplitGradOp)

paddle/fluid/ir/dialect/paddle_dialect/ir/pd_manual_op.h

@@ -14,7 +14,7 @@
 
 #ifdef GET_MANUAL_OP_LIST
 #undef GET_MANUAL_OP_LIST
-paddle::dialect::AddNOp
+paddle::dialect::AddNOp, paddle::dialect::SplitGradOp
 
 #else
 
@@ -51,9 +51,33 @@ class AddNOp : public ir::Op<AddNOp, OpYamlInfoInterface> {
   static void InferMeta(phi::InferMetaContext *infer_meta);
 };
 
+class SplitGradOp : public ir::Op<SplitGradOp, OpYamlInfoInterface> {
+ public:
+  using Op::Op;
+  static const char *name() { return "pd.split_grad"; }
+  static const char *attributes_name[1];
+  static constexpr uint32_t attributes_num = 1;
+  static OpInfoTuple GetOpInfo();
+  static void Build(ir::Builder &builder,             // NOLINT
+                    ir::OperationArgument &argument,  // NOLINT
+                    ir::OpResult x_,
+                    float axis = 0);
+  static void Build(ir::Builder &builder,             // NOLINT
+                    ir::OperationArgument &argument,  // NOLINT
+                    ir::OpResult out_grad_,
+                    ir::OpResult axis_);
+
+  void Verify();
+  ir::Value out_grad() { return operand_source(0); }
+  ir::Value axis() { return operand_source(1); }
+  ir::OpResult x_grad() { return result(0); }
+  static void InferMeta(phi::InferMetaContext *infer_meta);
+};
+
 }  // namespace dialect
 }  // namespace paddle
 
 IR_DECLARE_EXPLICIT_TYPE_ID(paddle::dialect::AddNOp)
+IR_DECLARE_EXPLICIT_TYPE_ID(paddle::dialect::SplitGradOp)
 
 #endif

paddle/fluid/ir/dialect/paddle_dialect/ir/pd_op_vjp_manual.cc

@@ -53,5 +53,39 @@ std::vector<std::vector<ir::OpResult>> SumOp::Vjp(
   }
   return res;
 }
+
+std::vector<std::vector<ir::OpResult>> SplitOp::Vjp(
+    ir::Operation* op,
+    const std::vector<std::vector<ir::OpResult>>& out_grads,
+    const std::vector<std::vector<bool>>& stop_gradients) {
+  SplitOp op_obj = op->dyn_cast<SplitOp>();
+
+  Tensor axis(std::make_shared<primitive::LazyTensor>(op_obj.axis()));
+  std::vector<Tensor> out_grads_;
+  for (size_t idx = 0; idx < out_grads[0].size(); idx++) {
+    out_grads_.emplace_back(
+        std::make_shared<primitive::LazyTensor>(out_grads[0][idx]));
+  }
+
+  std::vector<std::vector<Tensor>> tensor_res =
+      primitive::split_vjp(out_grads_, axis, stop_gradients);
+
+  std::vector<std::vector<ir::OpResult>> res(tensor_res.size(),
+                                             std::vector<ir::OpResult>());
+
+  for (uint64_t i = 0; i < tensor_res.size(); i++) {
+    res[i].resize(tensor_res[i].size());
+    for (uint64_t j = 0; j < tensor_res[i].size(); j++) {
+      if (tensor_res[i][j].defined()) {
+        res[i][j] = std::static_pointer_cast<primitive::LazyTensor>(
+                        tensor_res[i][j].impl())
+                        ->getValue()
+                        .dyn_cast<ir::OpResult>();
+      }
+    }
+  }
+  return res;
+}
+
 }  // namespace dialect
 }  // namespace paddle

paddle/fluid/primitive/backend/manual/manual_backend.h

@@ -33,6 +33,9 @@ std::vector<Tensor> concat_grad(const std::vector<Tensor>& x,
                                 const Tensor& out_grad,
                                 const Tensor& axis);
 
+template <typename T>
+Tensor split_grad(const std::vector<Tensor>& out_grads, const Tensor& axis);
+
 }  // namespace backend
 }  // namespace primitive
 }  // namespace paddle

paddle/fluid/primitive/backend/manual/manual_static_backend.cc

@@ -54,6 +54,23 @@ std::vector<Tensor> concat_grad<LazyTensor>(const std::vector<Tensor>& x,
   return op_result;
 }
 
+template <>
+Tensor split_grad<LazyTensor>(const std::vector<Tensor>& out_grads,
+                              const Tensor& axis) {
+  std::vector<ir::OpResult> out_grads_res;
+  for (uint64_t idx = 0; idx < out_grads.size(); idx++) {
+    out_grads_res.emplace_back(
+        std::static_pointer_cast<LazyTensor>(out_grads[idx].impl())
+            ->getValue()
+            .dyn_cast<ir::OpResult>());
+  }
+  ir::OpResult axis_res = std::static_pointer_cast<LazyTensor>(axis.impl())
+                              ->getValue()
+                              .dyn_cast<ir::OpResult>();
+  ir::OpResult op_res = paddle::dialect::split_grad(out_grads_res, axis_res);
+  return Tensor(std::make_shared<primitive::LazyTensor>(op_res));
+}
+
 }  // namespace backend
 }  // namespace primitive
 }  // namespace paddle

paddle/fluid/primitive/rule/vjp/manual/manual_vjp.cc

@@ -48,5 +48,25 @@ std::vector<std::vector<paddle::Tensor>> concat_vjp(
   return vjp_res;
 }
 
+std::vector<std::vector<paddle::Tensor>> split_vjp(
+    const std::vector<Tensor>& out_grads,
+    const Tensor& axis,
+    const std::vector<std::vector<bool>>& stop_gradients) {
+  std::vector<std::vector<paddle::Tensor>> vjp_res(3, std::vector<Tensor>(1));
+  // get concat_grad res.
+  Tensor op_res = backend::split_grad<primitive::LazyTensor>(out_grads, axis);
+
+  // construct vjp result by op result and stop_gradients info
+  if (!stop_gradients[0][0]) {
+    vjp_res[0][0] = op_res;
+  }
+
+  // vjp_res[1] is sections's grad which is attribute (no grad).
+  // vjp_res[2] is axis's grad which is attribute (no grad).
+  vjp_res[1].resize(stop_gradients[1].size());
+  vjp_res[2].resize(stop_gradients[2].size());
+  return vjp_res;
+}
+
 }  // namespace primitive
 }  // namespace paddle

paddle/fluid/primitive/rule/vjp/manual/manual_vjp.h

@@ -30,5 +30,10 @@ std::vector<std::vector<paddle::Tensor>> concat_vjp(
     const Tensor& axis,
     const std::vector<std::vector<bool>>& stop_gradients);
 
+std::vector<std::vector<paddle::Tensor>> split_vjp(
+    const std::vector<Tensor>& out_grads,
+    const Tensor& axis,
+    const std::vector<std::vector<bool>>& stop_gradients);
+
 }  // namespace primitive
 }  // namespace paddle

paddle/fluid/pybind/ops_api.cc

@@ -52,6 +52,10 @@ static PyObject *concat(PyObject *self, PyObject *args, PyObject *kwargs) {
   return static_api_concat(self, args, kwargs);
 }
 
+static PyObject *split(PyObject *self, PyObject *args, PyObject *kwargs) {
+  return static_api_split(self, args, kwargs);
+}
+
 static PyMethodDef OpsAPI[] = {{"add_n",
                                 (PyCFunction)(void (*)(void))add_n,
                                 METH_VARARGS | METH_KEYWORDS,
@@ -76,6 +80,10 @@ static PyMethodDef OpsAPI[] = {{"add_n",
                                 (PyCFunction)(void (*)(void))full,
                                 METH_VARARGS | METH_KEYWORDS,
                                 "C++ interface function for full."},
+                               {"split",
+                                (PyCFunction)(void (*)(void))split,
+                                METH_VARARGS | METH_KEYWORDS,
+                                "C++ interface function for split."},
                                {"data",
                                 (PyCFunction)(void (*)(void))data,
                                 METH_VARARGS | METH_KEYWORDS,

python/paddle/autograd/backward.py

@@ -201,7 +201,6 @@ def prune_ops(total_ops, inputs_set, outputs_set, no_grad_set):
                     outputs_set.add(operand)
         else:
             relevant_op_flags[i] = False
-
     # recover full op or full_Intarray op created by mutable attribute.
     total_ops_list = list(total_ops)
     for i, op in enumerate(total_ops_list):
@@ -354,12 +353,16 @@ def append_backward_ops(
 
     def make_output_grad(op):
         zero_flag = [False] * op.num_results()
+        output_grads = []
         for i, value in enumerate(op.results()):
             if (
                 value not in state.value_to_valuegrad
                 or state.value_to_valuegrad[value] is None
             ):
-                if value.first_use().owner().name() == "builtin.split":
+                if (
+                    not value.use_empty()
+                    and value.first_use().owner().name() == "builtin.split"
+                ):
                     # pattern case:
                     # this fwd_op's output is vectorType, it will split to
                     # Type by builtin.split op, so need get from split op's ouput
@@ -367,7 +370,7 @@ def append_backward_ops(
                         value.first_use().owner()
                     )
                     zero_flag[i] = all(split_zero_flag)
-                    grad_value = [op_list[0] for op_list in split_output_grad]
+                    state.value_to_valuegrad[value] = [split_output_grad]
                 else:
                     # first case:
                     # this fwd_op's output didn't used by other fwd_op,
@@ -388,7 +391,7 @@ def append_backward_ops(
                     )
                     zero_flag[i] = True
 
-                state.value_to_valuegrad[value] = [[grad_value]]
+                    state.value_to_valuegrad[value] = [[grad_value]]
 
             if len(state.value_to_valuegrad[value]) > 1:
                 # one value is input of more than one fwd_op,
@@ -411,8 +414,8 @@ def append_backward_ops(
                     value
                 ]
 
-            output_grad = state.value_to_valuegrad[value][0]
-        return zero_flag, output_grad
+            output_grads.append(state.value_to_valuegrad[value][0][0])
+        return zero_flag, output_grads
 
     def make_input_stopgradient(op):
         input_grad_stopgradients = []
@@ -529,8 +532,9 @@ def create_backward_prune_set(inputs, outputs, no_grad_set, state):
             inputs_set.add(state.value_to_valuegrad[output][0][0])
     inputs_set_tmp = set()
     for out_grad in inputs_set:
-        for item in out_grad.first_use().owner().operands_source():
-            inputs_set_tmp.add(item)
+        if not out_grad.use_empty():
+            for item in out_grad.first_use().owner().operands_source():
+                inputs_set_tmp.add(item)
     inputs_set.update(inputs_set_tmp)
 
     no_gradvar_set = set()  # grad_value of value in no_grad_set

python/paddle/tensor/manipulation.py

@@ -1976,6 +1976,14 @@ def split(x, num_or_sections, axis=0, name=None):
         else:
             return _C_ops.split(input, num_or_sections, dim)
     else:
+        if paddle.ir.core._use_new_ir_api():
+            if not isinstance(num_or_sections, int):
+                return paddle._ir_ops.split(input, num_or_sections, dim)
+            else:
+                raise NotImplementedError(
+                    "_ir_ops.split_with_num is not implemented, please change sections as list"
+                )
+
         check_variable_and_dtype(
             input,
             'input',

test/cpp/prim/test_vjp.cc

@@ -402,5 +402,74 @@ TEST(VJP, Add_BackwardTest) {
   ASSERT_EQ(dx.data<float>()[0], 1.0);
   ASSERT_EQ(dy.data<float>()[0], 1.0);
 }
+
+TEST(VJP, SplitBackwardTest) {
+  ir::IrContext* ctx = ir::IrContext::Instance();
+  ir::Program program((ctx));
+  paddle::dialect::APIBuilder::Instance().SetProgram(&program);
+
+  std::shared_ptr<ir::Builder> builder =
+      paddle::dialect::APIBuilder::Instance().GetBuilder();
+  paddle::dialect::FullOp op1 = builder->Build<paddle::dialect::FullOp>(
+      std::vector<int64_t>{2, 2}, 2.0, phi::DataType::FLOAT32, phi::CPUPlace());
+
+  paddle::dialect::SplitOp op2 = builder->Build<paddle::dialect::SplitOp>(
+      op1.out(), std::vector<int64_t>{1, 1}, 0);
+
+  ir::SplitOp op3 = builder->Build<ir::SplitOp>(op2.out());
+
+  paddle::dialect::FullOp op4 = builder->Build<paddle::dialect::FullOp>(
+      std::vector<int64_t>{1, 2}, 1.0, phi::DataType::FLOAT32, phi::CPUPlace());
+
+  std::vector<std::vector<bool>> stop_gradients{{false}, {true}, {true}};
+  std::vector<std::vector<ir::OpResult>> out_grads{{op3.result(0), op4.out()}};
+  ir::OpInfo op2_info = ctx->GetRegisteredOpInfo("pd.split");
+
+  auto concat_vjp_interface_impl =
+      op2_info.GetInterfaceImpl<paddle::dialect::VjpInterface>();
+
+  concat_vjp_interface_impl->vjp_(op2.operation(), out_grads, stop_gradients);
+  auto kernel_program = paddle::dialect::PdOpLowerToKernelPass(&program);
+
+  auto place = platform::CPUPlace();
+  Scope scope;
+  ProgramDesc prog_desc;
+  InterpreterCore test_core(place, {}, std::move(kernel_program), &scope);
+  std::stringstream os;
+  os << reinterpret_cast<NewIRInterpreter*>(
+      const_cast<InterpreterBaseImpl*>(test_core.Impl()));
+  std::string prefix_str = os.str();
+  test_core.SetSkipGcVars({prefix_str + "_inner_var_4",
+                           prefix_str + "_inner_var_5",
+                           prefix_str + "_inner_var_8"});
+  test_core.Run({});
+  auto out_tensor_0 =
+      test_core.local_scope() == nullptr
+          ? scope.FindVar(prefix_str + "_inner_var_4")->Get<phi::DenseTensor>()
+          : test_core.local_scope()
+                ->FindVar(prefix_str + "_inner_var_4")
+                ->Get<phi::DenseTensor>();
+  auto out_tensor_1 =
+      test_core.local_scope() == nullptr
+          ? scope.FindVar(prefix_str + "_inner_var_5")->Get<phi::DenseTensor>()
+          : test_core.local_scope()
+                ->FindVar(prefix_str + "_inner_var_5")
+                ->Get<phi::DenseTensor>();
+  auto grad_out_tensor_0 =
+      test_core.local_scope() == nullptr
+          ? scope.FindVar(prefix_str + "_inner_var_8")->Get<phi::DenseTensor>()
+          : test_core.local_scope()
+                ->FindVar(prefix_str + "_inner_var_8")
+                ->Get<phi::DenseTensor>();
+  ASSERT_EQ(out_tensor_0.data<float>()[0], 2.0);
+  ASSERT_EQ(out_tensor_0.data<float>()[1], 2.0);
+  ASSERT_EQ(out_tensor_1.data<float>()[0], 2.0);
+  ASSERT_EQ(out_tensor_1.data<float>()[1], 2.0);
+  ASSERT_EQ(grad_out_tensor_0.data<float>()[0], 2.0);
+  ASSERT_EQ(grad_out_tensor_0.data<float>()[1], 2.0);
+  ASSERT_EQ(grad_out_tensor_0.data<float>()[2], 1.0);
+  ASSERT_EQ(grad_out_tensor_0.data<float>()[3], 1.0);
+}
+
 }  // namespace framework
 }  // namespace paddle

test/ir/new_ir/test_build_op.py

@@ -121,5 +121,25 @@ class TestBuildOp4(unittest.TestCase):
         paddle.framework.set_flags({"FLAGS_enable_new_ir_api": False})
 
 
+class TestBuildOp5(unittest.TestCase):
+    def test_build_split_op(self):
+        newir_program = get_ir_program()
+        tanh_out = newir_program.block().ops[-1].result(0)
+        paddle.framework.set_flags({"FLAGS_enable_new_ir_api": True})
+        with paddle.ir.core.program_guard(newir_program):
+            out = paddle.split(tanh_out, [2, 2], 0)
+        self.assertEqual(out[0].get_defining_op().name(), "builtin.split")
+        self.assertEqual(
+            out[0]
+            .get_defining_op()
+            .operands()[0]
+            .source()
+            .get_defining_op()
+            .name(),
+            "pd.split",
+        )
+        paddle.framework.set_flags({"FLAGS_enable_new_ir_api": False})
+
+
 if __name__ == "__main__":
     unittest.main()

test/ir/new_ir/test_ir_backward.py

@@ -94,6 +94,32 @@ class TesBackward_1(unittest.TestCase):
         self.assertEqual(newir_program.block().ops[-1].name(), "pd.mean")
         paddle.framework.set_flags({"FLAGS_enable_new_ir_api": False})
 
+    def test_split(self):
+        # test create output_grad in backward use full op
+        newir_program = get_ir_program_0()
+        input = newir_program.block().ops[-1].operand(0).source()
+        tanh_out = newir_program.block().ops[-1].result(0)
+        paddle.framework.set_flags({"FLAGS_enable_new_ir_api": True})
+        with paddle.ir.core.program_guard(newir_program):
+            out = paddle.split(tanh_out, [2, 2], 0)
+            input_grad = grad(out, input)
+
+        ops_name = [
+            "pd.data",
+            "pd.tanh",
+            "pd.full_int_array",
+            "pd.full",
+            "pd.split",
+            "builtin.split",
+            "pd.full",
+            "builtin.combine",
+            "pd.split_grad",
+            "pd.tanh_grad",
+        ]
+        for i, op in enumerate(newir_program.block().ops):
+            self.assertEqual(op.name(), ops_name[i])
+        paddle.framework.set_flags({"FLAGS_enable_new_ir_api": False})
+
 
 def get_ir_program_1():
     x = paddle.randn([2, 2])