diff --git a/paddle/cinn/ir/ir.cc b/paddle/cinn/ir/ir.cc
index c7402d56a74cb9..7157fcf5c15ce3 100644
--- a/paddle/cinn/ir/ir.cc
+++ b/paddle/cinn/ir/ir.cc
@@ -407,6 +407,7 @@ const std::string &Store::name() const {
 }
 
 Type Store::type() const { return value.type(); }
+
 std::vector<Expr *> Store::expr_fields() {
   std::vector<Expr *> exprs({&tensor, &value});
   for (auto &idx : indices) exprs.push_back(&idx);
@@ -776,6 +777,9 @@ Expr Reduce::Make(Reduce::ReduceType reduce_type,
   n->set_type(body.type());
   return Expr(n);
 }
+
+Type Reduce::type() const { return body.type().ElementOf(); }
+
 std::vector<Expr *> Reduce::expr_fields() {
   std::vector<Expr *> res;
   if (init.defined()) {
@@ -802,6 +806,12 @@ void Reduce::Verify() const {
   CHECK_EQ(init.type(), body.type());
 }
 
+Type Select::type() const {
+  PADDLE_ENFORCE_EQ(
+      true_value.type(), false_value.type(), "Type of Select must be same");
+  return true_value.type();
+}
+
 void Select::Verify() const {
   CHECK(condition.defined());
   CHECK(true_value.defined());
@@ -862,12 +872,16 @@ void MultiOperandVerify(llvm::ArrayRef<Expr> operands) {
   }
 }
 
+Type Product::type() const { return operands().front().type(); }
+
 void Product::Verify() const {
   CHECK_GT(operands().size(), 1UL)
       << "Product node should have more than 1 operands";
   MultiOperandVerify(operands());
 }
 
+Type Sum::type() const { return operands().front().type(); }
+
 void Sum::Verify() const {
   CHECK_GT(operands().size(), 1UL)
       << "Sum node should have more than 1 operands";
diff --git a/paddle/cinn/ir/ir.h b/paddle/cinn/ir/ir.h
index d711e93ce61abb..72429f6f0556a1 100644
--- a/paddle/cinn/ir/ir.h
+++ b/paddle/cinn/ir/ir.h
@@ -476,7 +476,7 @@ struct Reduce : public ExprNode<Reduce> {
                    Expr body,
                    const std::vector<Var>& reduce_axis);
 
-  Type type() const override { return body.type().ElementOf(); }
+  Type type() const override;
 
   std::vector<Expr*> expr_fields() override;
   std::vector<const Expr*> expr_fields() const override;
@@ -509,10 +509,7 @@ struct Select : public ExprNode<Select> {
     return Expr(node);
   }
 
-  Type type() const override {
-    CHECK_EQ(true_value.type(), false_value.type());
-    return true_value.type();
-  }
+  Type type() const override;
 
   void Verify() const override;
 
@@ -573,6 +570,7 @@ struct Store : public ExprNode<Store>, public LoadStoreAddrMnger {
   const std::string& name() const;
 
   Type type() const override;
+
   Expr index() const;
 
   static const IrNodeTy _node_type_ = IrNodeTy::Store;
@@ -932,7 +930,7 @@ struct Product : public ExprNode<Product> {
 
   using ExprNode<Product>::operand;
 
-  Type type() const override { return operands().front().type(); }
+  Type type() const override;
 
   void Verify() const override;
 
@@ -944,7 +942,7 @@ struct Sum : public ExprNode<Sum> {
 
   using ExprNode<Sum>::operand;
 
-  Type type() const override { return operands().front().type(); }
+  Type type() const override;
 
   void Verify() const override;
 
diff --git a/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/multiary_infer_sym.cc b/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/multiary_infer_sym.cc
index 4b5f146b66684a..4a8b0c1000cc0b 100644
--- a/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/multiary_infer_sym.cc
+++ b/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/multiary_infer_sym.cc
@@ -504,6 +504,25 @@ bool MemoryEfficientAttentionOpInferSymbolicShape(
   return true;
 }
 
+bool RoiAlignOpInferSymbolicShape(
+    pir::Operation *op, pir::InferSymbolicShapeContext *infer_context) {
+  const auto &x = op->operand_source(0);
+  const auto &boxes = op->operand_source(1);
+
+  const auto &num_boxes =
+      infer_context->GetShapeOrDataForValue(boxes).shape()[0];
+  symbol::DimExpr channel_num =
+      infer_context->GetShapeOrDataForValue(x).shape()[1];
+
+  int32_t out_h = op->attribute<pir::Int32Attribute>("pooled_height").data();
+  int32_t out_w = op->attribute<pir::Int32Attribute>("pooled_width").data();
+
+  std::vector<symbol::DimExpr> out_dim = {num_boxes, channel_num, out_h, out_w};
+  infer_context->SetShapeOrDataForValue(
+      op->result(0), symbol::TensorShapeOrDataDimExprs(out_dim));
+  return true;
+}
+
 bool MeshgridOpInferSymbolicShape(
     pir::Operation *op, pir::InferSymbolicShapeContext *infer_context) {
   const symbol::TensorListShapeOrDataDimExprs &shape_data_list =
diff --git a/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/multiary_infer_sym.h b/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/multiary_infer_sym.h
index ccb95c344cf28d..ba28e796a51138 100644
--- a/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/multiary_infer_sym.h
+++ b/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/multiary_infer_sym.h
@@ -30,6 +30,7 @@ OP_DECLARE_INFER_SYMBOLIC_SHAPE(Logspace)
 OP_DECLARE_INFER_SYMBOLIC_SHAPE(MemoryEfficientAttention)
 OP_DECLARE_INFER_SYMBOLIC_SHAPE(Meshgrid)
 OP_DECLARE_INFER_SYMBOLIC_SHAPE(NearestInterp)
+OP_DECLARE_INFER_SYMBOLIC_SHAPE(RoiAlign)
 OP_DECLARE_INFER_SYMBOLIC_SHAPE(Stack)
 OP_DECLARE_INFER_SYMBOLIC_SHAPE(TrilinearInterp)
 OP_DECLARE_INFER_SYMBOLIC_SHAPE(Where)
diff --git a/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/unary_infer_sym.cc b/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/unary_infer_sym.cc
index 5efddf897a6d03..f84ca545188bfa 100644
--- a/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/unary_infer_sym.cc
+++ b/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/unary_infer_sym.cc
@@ -246,6 +246,84 @@ bool DiagonalOpInferSymbolicShape(
   return true;
 }
 
+bool DistributeFpnProposalsOpInferSymbolicShape(
+    pir::Operation *op, pir::InferSymbolicShapeContext *infer_context) {
+  const auto &attributes = op->attributes();
+  int32_t min_level =
+      attributes.at("min_level").dyn_cast<pir::Int32Attribute>().data();
+  int32_t max_level =
+      attributes.at("max_level").dyn_cast<pir::Int32Attribute>().data();
+  int32_t num_levels = max_level - min_level + 1;
+  int64_t batch_size = 1;
+
+  symbol::DimExpr num_rois = [&]() {
+    pir::Value rois_num = op->operand_source(1);
+    const auto &rois_num_shape_or_data =
+        infer_context->GetShapeOrDataForValue(rois_num);
+
+    batch_size = rois_num_shape_or_data.shape()[0].Get<int64_t>();
+    PADDLE_ENFORCE_EQ(rois_num_shape_or_data.data().has_value(),
+                      true,
+                      ::common::errors::InvalidArgument(
+                          "InferSymbolicShape of DistributeFpnProposalsdOp "
+                          "only support input with rois_num."));
+
+    symbol::DimExpr rois_total_num = 0;
+    for (int i = 0; i < batch_size; i++) {
+      const auto &rois_num_value = rois_num_shape_or_data.data().value()[i];
+
+      CHECK(rois_num_value.isa<int64_t>() || rois_num_value.isa<std::string>())
+          << "rois_num must be int64 or SymName.";
+      if (rois_num_value.isa<int64_t>()) {
+        return symbol::DimExpr(rois_num_value.Get<int64_t>());
+      } else {
+        return symbol::DimExpr(rois_num_value.Get<std::string>());
+      }
+      rois_total_num = rois_total_num + rois_num_value;
+    }
+
+    return rois_total_num;
+  }();
+
+  const auto &multi_rois_out_shape = [&]() {
+    symbol::TensorListShapeOrDataDimExprs multi_rois_out_shape;
+    if (num_levels == 1) {
+      multi_rois_out_shape.emplace_back(
+          symbol::TensorShapeOrDataDimExprs({num_rois, 4}));
+    } else {
+      symbol::DimExpr last_dim = num_rois;
+      for (int i = 0; i < num_levels - 1; i++) {
+        const auto &next_sym_name = infer_context->GetNextSymName();
+        std::vector<symbol::DimExpr> level_dim = {next_sym_name, 4};
+        multi_rois_out_shape.emplace_back(
+            symbol::TensorShapeOrDataDimExprs(level_dim));
+        last_dim = last_dim - level_dim[0];
+      }
+      multi_rois_out_shape.emplace_back(
+          symbol::TensorShapeOrDataDimExprs({last_dim, 4}));
+    }
+
+    return multi_rois_out_shape;
+  }();
+
+  const auto &rois_num_per_level_out_shape = [&]() {
+    symbol::TensorListShapeOrDataDimExprs rois_num_per_level_out_shape;
+    rois_num_per_level_out_shape.resize(
+        num_levels, symbol::TensorShapeOrDataDimExprs({batch_size}));
+    return rois_num_per_level_out_shape;
+  }();
+
+  const auto &restore_ind = [&]() {
+    return symbol::TensorShapeOrDataDimExprs({num_rois, 1});
+  }();
+
+  infer_context->SetShapeOrDataForValue(op->result(0), multi_rois_out_shape);
+  infer_context->SetShapeOrDataForValue(op->result(1),
+                                        rois_num_per_level_out_shape);
+  infer_context->SetShapeOrDataForValue(op->result(2), restore_ind);
+  return true;
+}
+
 bool EinsumOpInferSymbolicShape(pir::Operation *op,
                                 pir::InferSymbolicShapeContext *infer_context) {
   PADDLE_THROW(phi::errors::Unimplemented(
diff --git a/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/unary_infer_sym.h b/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/unary_infer_sym.h
index 8155b4f0ac5052..e260add8ddda0c 100644
--- a/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/unary_infer_sym.h
+++ b/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/unary_infer_sym.h
@@ -29,6 +29,7 @@ OP_DECLARE_INFER_SYMBOLIC_SHAPE(Cumsum)
 OP_DECLARE_INFER_SYMBOLIC_SHAPE(Cumsum_)
 OP_DECLARE_INFER_SYMBOLIC_SHAPE(DiagEmbed)
 OP_DECLARE_INFER_SYMBOLIC_SHAPE(Diagonal)
+OP_DECLARE_INFER_SYMBOLIC_SHAPE(DistributeFpnProposals)
 OP_DECLARE_INFER_SYMBOLIC_SHAPE(Einsum)
 OP_DECLARE_INFER_SYMBOLIC_SHAPE(Kthvalue)
 OP_DECLARE_INFER_SYMBOLIC_SHAPE(Logcumsumexp)
diff --git a/paddle/fluid/pir/dialect/operator/ir/ops.yaml b/paddle/fluid/pir/dialect/operator/ir/ops.yaml
index 926363d002af12..3fd2c36fda46be 100644
--- a/paddle/fluid/pir/dialect/operator/ir/ops.yaml
+++ b/paddle/fluid/pir/dialect/operator/ir/ops.yaml
@@ -525,6 +525,7 @@
     func : distribute_fpn_proposals
     data_type : fpn_rois
   optional : rois_num, multi_level_rois_num
+  interfaces : paddle::dialect::InferSymbolicShapeInterface
 
 - op : distributed_fused_lamb
   args : (Tensor[] param, Tensor[] grad, Tensor fp32_fused_param, Tensor fp32_fused_grad, Tensor fp16_fused_param, Tensor fp16_fused_grad, Tensor moment1, Tensor moment2, Tensor beta1pow, Tensor beta2pow, Tensor fused_param_offsets, Tensor fp32_shard_fused_param_offsets, Tensor fp16_shard_fused_param_offsets, Tensor param_info, Tensor param_order, Tensor learning_rate, Tensor global_scale, float beta1, float beta2, float epsilon, float max_global_grad_norm, float weight_decay, bool clip_after_allreduce, int[] ring_ids= {}, int acc_steps = 1, bool use_master_param_norm = true, bool use_master_acc_grad = true, bool is_grad_scaled_by_nranks = true, int64_t nranks = 1, bool use_hierarchical_allreduce = false)
diff --git a/paddle/phi/api/yaml/ops.yaml b/paddle/phi/api/yaml/ops.yaml
index a59f50b7a8ac64..0c77952bd9743f 100755
--- a/paddle/phi/api/yaml/ops.yaml
+++ b/paddle/phi/api/yaml/ops.yaml
@@ -2500,6 +2500,7 @@
     data_type : x
   optional : boxes_num
   backward : roi_align_grad
+  interfaces : paddle::dialect::InferSymbolicShapeInterface
 
 - op : roi_pool
   args : (Tensor x, Tensor boxes, Tensor boxes_num, int pooled_height=1, int pooled_width=1, float spatial_scale=1.0)