Fuse multi transformer layer pass (#47541) (#47830)

* add fuse_multi_transformer_layer_pass

paddle/fluid/framework/ir/CMakeLists.txt

@@ -154,6 +154,7 @@ pass_library(skip_layernorm_fuse_pass base)
 pass_library(multihead_matmul_fuse_pass inference)
 pass_library(fused_multi_transformer_encoder_pass inference)
 pass_library(fused_multi_transformer_decoder_pass inference)
+pass_library(fuse_multi_transformer_layer_pass inference)
 pass_library(adaptive_pool2d_convert_global_pass inference)
 pass_library(unsqueeze2_eltwise_fuse_pass inference)
 pass_library(yolo_box_fuse_pass inference)
@@ -368,6 +369,10 @@ cc_test(
   test_fused_multi_transformer_decoder_pass
   SRCS fused_multi_transformer_decoder_pass_tester.cc
   DEPS fused_multi_transformer_decoder_pass)
+cc_test(
+  test_fuse_multi_transformer_layer_pass
+  SRCS fuse_multi_transformer_layer_pass_tester.cc
+  DEPS fuse_multi_transformer_layer_pass)
 cc_test(
   test_conv_bn_fuse_pass_cc
   SRCS conv_bn_fuse_pass_tester.cc

paddle/fluid/framework/ir/fuse_multi_transformer_layer_pass.cc

+// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "paddle/fluid/framework/ir/fuse_multi_transformer_layer_pass.h"
+
+#include <string>
+
+#include "paddle/fluid/framework/convert_utils.h"
+#include "paddle/fluid/framework/lod_tensor.h"
+#include "paddle/fluid/framework/op_version_registry.h"
+
+namespace paddle {
+namespace framework {
+class Scope;
+}  // namespace framework
+}  // namespace paddle
+
+namespace paddle {
+namespace framework {
+namespace ir {
+namespace patterns {
+
+std::unordered_map<std::string, std::string>
+MultiTransformerLayerPattern::operator()(bool enable_int8,
+                                         int num_fused_op,
+                                         bool is_decoder) {
+  std::string fused_multi_transformer_name =
+      enable_int8 ? "fused_multi_transformer_int8" : "fused_multi_transformer";
+
+  std::unordered_map<std::string, std::string> node_reprs;
+
+  // x0 and src_mask is unqiue input of subgraph
+  auto* x0 = pattern->NewNode(x0_repr());
+  x0->assert_is_op_input(fused_multi_transformer_name, "X")->AsInput();
+  auto* src_mask = pattern->NewNode(src_mask_repr());
+  src_mask->assert_is_op_input(fused_multi_transformer_name, "SrcMask")
+      ->AsInput();
+
+  for (int i = 0; i < num_fused_op; ++i) {
+    auto fuse_op_repr =
+        PDNodeName(name_scope_, repr_, id_, "fuse_op_" + std::to_string(i));
+    node_reprs["fuse_op_" + std::to_string(i)] = fuse_op_repr;
+    auto* fused_multi_transformer =
+        pattern->NewNode(fuse_op_repr)
+            ->assert_is_op(fused_multi_transformer_name);
+
+    auto out_repr =
+        PDNodeName(name_scope_, repr_, id_, "out_" + std::to_string(i));
+    node_reprs["out_" + std::to_string(i)] = out_repr;
+    auto* out = pattern->NewNode(out_repr)->assert_is_op_output(
+        fused_multi_transformer_name, "Out");
+
+    if (is_decoder) {
+      auto shape_repr =
+          PDNodeName(name_scope_, repr_, id_, "shape_" + std::to_string(i));
+      node_reprs["shape_" + std::to_string(i)] = shape_repr;
+      auto* shape = pattern->NewNode(shape_repr)->assert_is_op("shape");
+
+      auto shape_out_repr =
+          PDNodeName(name_scope_, repr_, id_, "shape_out_" + std::to_string(i));
+      node_reprs["shape_out_" + std::to_string(i)] = shape_out_repr;
+      auto* shape_out =
+          pattern->NewNode(shape_out_repr)->assert_is_op_output("shape", "Out");
+
+      shape->LinksFrom({src_mask}).LinksTo({shape_out});
+
+      auto slice_repr =
+          PDNodeName(name_scope_, repr_, id_, "slice_" + std::to_string(i));
+      node_reprs["slice_" + std::to_string(i)] = slice_repr;
+      auto* slice = pattern->NewNode(slice_repr)->assert_is_op("slice");
+
+      auto slice_out_repr =
+          PDNodeName(name_scope_, repr_, id_, "slice_out_" + std::to_string(i));
+      node_reprs["slice_out_" + std::to_string(i)] = slice_out_repr;
+      auto* slice_out =
+          pattern->NewNode(slice_out_repr)->assert_is_op_output("slice", "Out");
+
+      slice->LinksFrom({shape_out}).LinksTo({slice_out});
+
+      fused_multi_transformer->LinksFrom({x0, src_mask, slice_out})
+          .LinksTo({out});
+    } else {
+      auto cache_kv_repr =
+          PDNodeName(name_scope_, repr_, id_, "cache_kv_" + std::to_string(i));
+      node_reprs["cache_kv_" + std::to_string(i)] = cache_kv_repr;
+      auto* cache_kv = pattern->NewNode(cache_kv_repr);
+      cache_kv->assert_is_op_input(fused_multi_transformer_name, "CacheKV");
+      cache_kv->AsInput();
+
+      auto fill_const_op_repr =
+          PDNodeName(name_scope_, repr_, id_, "fill_op_" + std::to_string(i));
+      node_reprs["fill_op_" + std::to_string(i)] = fill_const_op_repr;
+      auto fill_const_op = pattern->NewNode(fill_const_op_repr)
+                               ->assert_is_op("fill_constant_batch_size_like");
+
+      fused_multi_transformer->LinksFrom({x0, src_mask, cache_kv})
+          .LinksTo({out});
+      fill_const_op->LinksFrom({x0}).LinksTo({cache_kv});
+    }
+    x0 = out;
+  }
+  x0->AsOutput();
+  return node_reprs;
+}
+}  // namespace patterns
+
+inline void MergeInput(OpDesc* op,
+                       const std::vector<VariableNameMap>& input_name_maps,
+                       const std::string& input_name) {
+  std::vector<std::string> tmp = input_name_maps[0].at(input_name);
+  for (size_t i = 1; i < input_name_maps.size(); ++i) {
+    tmp.insert(tmp.end(),
+               input_name_maps[i].at(input_name).begin(),
+               input_name_maps[i].at(input_name).end());
+  }
+  op->SetInput(input_name, tmp);
+}
+
+template <typename T>
+inline void MergeAttrs(const std::vector<OpDesc*>& ops,
+                       const std::string& attr_name) {
+  std::vector<T> res;
+  for (size_t i = 0; i < ops.size(); ++i) {
+    auto scale_vec =
+        PADDLE_GET_CONST(std::vector<T>, ops[i]->GetAttr(attr_name));
+    res.insert(res.end(), scale_vec.begin(), scale_vec.end());
+  }
+  ops[0]->SetAttr(attr_name, res);
+}
+
+int FuseMultiTransformerLayerPass::BuildFusion(Graph* graph,
+                                               const std::string& name_scope,
+                                               Scope* scope) const {
+  GraphPatternDetector gpd;
+  auto* pattern = gpd.mutable_pattern();
+
+  // TODO(wufeisheng): Get enable_int8 attr from graph after
+  // fused_multi_transformer pass with int8 merged
+  bool enable_int8 = false;
+
+  int num_fuse_op = 0;
+  bool is_decoder = false;
+
+  if (graph->Has(kFusedMultiTransformerEncoderFusionCount)) {
+    num_fuse_op = graph->Get<int>(kFusedMultiTransformerEncoderFusionCount);
+    is_decoder = false;
+  } else if (graph->Has(kFusedMultiTransformerDecoderFusionCount)) {
+    num_fuse_op = graph->Get<int>(kFusedMultiTransformerDecoderFusionCount);
+    is_decoder = true;
+  }
+  if (num_fuse_op == 0) {
+    VLOG(4) << "fuse_multi_transformer_layer_pass will be skipped "
+               "cause num_fuse_op is not been set or set to 0";
+    return 0;
+  }
+  if (!is_decoder) {
+    VLOG(4) << "fuse_multi_transformer_layer_pass will match encoder pattern";
+  } else {
+    VLOG(4) << "fuse_multi_transformer_layer_pass will match decoder pattern";
+  }
+
+  patterns::MultiTransformerLayerPattern multi_layer_pattern(pattern,
+                                                             name_scope);
+  auto node_reprs = multi_layer_pattern(enable_int8, num_fuse_op, is_decoder);
+
+  int fusion_count{0};
+  auto handler = [&](const GraphPatternDetector::subgraph_t& subgraph,
+                     Graph* graph) {
+    ///////////////////
+    //// Get nodes ////
+    ///////////////////
+
+    GET_IR_NODE_FROM_SUBGRAPH(src_mask, src_mask, multi_layer_pattern);
+    GET_IR_NODE_FROM_SUBGRAPH(x0, x0, multi_layer_pattern);
+
+    std::vector<Node*> fuse_op_nodes;
+    std::vector<Node*> out_nodes;
+
+    std::vector<std::string> unused_node_prefixes = {
+        "shape_", "shape_out_", "slice_", "slice_out_"};
+    std::vector<Node*> unused_nodes;
+
+    std::vector<OpDesc*> fuse_op_descs;
+    std::vector<VariableNameMap> fuse_op_input_var_name_maps;
+    std::vector<VariableNameMap> fuse_op_output_var_name_maps;
+
+    for (int i = 0; i < num_fuse_op; ++i) {
+      PDNode* fuse_op_pdnode =
+          multi_layer_pattern.PatternBase::pattern->RetrieveNode(
+              node_reprs["fuse_op_" + std::to_string(i)]);
+      Node* fuse_op_node = subgraph.at(fuse_op_pdnode);
+      fuse_op_nodes.push_back(fuse_op_node);
+      fuse_op_descs.push_back(fuse_op_node->Op());
+      fuse_op_input_var_name_maps.emplace_back(fuse_op_node->Op()->Inputs());
+      fuse_op_output_var_name_maps.emplace_back(fuse_op_node->Op()->Outputs());
+
+      PDNode* out_pdnode =
+          multi_layer_pattern.PatternBase::pattern->RetrieveNode(
+              node_reprs["out_" + std::to_string(i)]);
+      out_nodes.push_back(subgraph.at(out_pdnode));
+
+      // fill_const op use x0 as input
+      if (!is_decoder && i != 0) {
+        PDNode* fill_op_pdnode =
+            multi_layer_pattern.PatternBase::pattern->RetrieveNode(
+                node_reprs["fill_op_" + std::to_string(i)]);
+        Node* fill_op_node = subgraph.at(fill_op_pdnode);
+        fill_op_node->Op()->SetInput("Input", {x0->Name()});
+        IR_NODE_UNLINK(out_nodes[i - 1], fill_op_node);
+        IR_NODE_LINK_TO(x0, fill_op_node);
+      } else if (is_decoder && i != 0) {
+        for (const auto& unused_node_prefix : unused_node_prefixes) {
+          PDNode* unused_pdnode =
+              multi_layer_pattern.PatternBase::pattern->RetrieveNode(
+                  node_reprs[unused_node_prefix + std::to_string(i)]);
+          Node* unused_node = subgraph.at(unused_pdnode);
+          unused_nodes.push_back(unused_node);
+        }
+      }
+    }
+
+    ///////////////
+    //// Merge ////
+    ///////////////
+
+    // Merge inputs
+    std::vector<std::string> inputs_names = {"CacheKV",
+                                             "FFN1Bias",
+                                             "FFN1Weight",
+                                             "FFN2Bias",
+                                             "FFN2Weight",
+                                             "FFNLnBias",
+                                             "FFNLnScale",
+                                             "LnBias",
+                                             "LnScale",
+                                             "OutLinearBias",
+                                             "OutLinearW",
+                                             "QKVBias",
+                                             "QKVW"};
+
+    for (const auto& input_name : inputs_names) {
+      MergeInput(fuse_op_descs[0], fuse_op_input_var_name_maps, input_name);
+    }
+
+    // Merge outputs
+    fuse_op_descs[0]->SetOutput(
+        "Out", fuse_op_output_var_name_maps[num_fuse_op - 1]["Out"]);
+    auto& merged_cache_kv_out_names =
+        fuse_op_output_var_name_maps[0]["CacheKVOut"];
+    for (int i = 1; i < num_fuse_op; ++i) {
+      const auto& out_var_names = fuse_op_output_var_name_maps[i]["CacheKVOut"];
+      merged_cache_kv_out_names.insert(merged_cache_kv_out_names.end(),
+                                       out_var_names.begin(),
+                                       out_var_names.end());
+    }
+    fuse_op_descs[0]->SetOutput("CacheKVOut", merged_cache_kv_out_names);
+
+    ////////////////
+    //// ReLink ////
+    ////////////////
+    // Before relink, out nodes (0 -> num_layer-1) should be removed
+    std::unordered_set<const Node*> marked_out_nodes(out_nodes.begin(),
+                                                     out_nodes.end() - 1);
+    GraphSafeRemoveNodes(graph, marked_out_nodes);
+
+    // Relink all input nodes of fused_multi_transformer ops to the first op
+    auto& merged_inputs = fuse_op_nodes[0]->inputs;
+    for (int i = 1; i < num_fuse_op; ++i) {
+      merged_inputs.insert(merged_inputs.end(),
+                           fuse_op_nodes[i]->inputs.begin(),
+                           fuse_op_nodes[i]->inputs.end());
+    }
+
+    // Relink fuse op -> out
+    IR_NODE_UNLINK(fuse_op_nodes[num_fuse_op - 1], out_nodes[num_fuse_op - 1]);
+    IR_NODE_LINK_TO(fuse_op_nodes[0], out_nodes[num_fuse_op - 1]);
+
+    /////////////////////////////
+    //// Delete unused nodes ////
+    /////////////////////////////
+    // Delete fused_multi_transformer op expect for the first one
+    std::unordered_set<const Node*> marked_fuse_op_nodes(
+        fuse_op_nodes.begin() + 1, fuse_op_nodes.end());
+
+    if (is_decoder) {
+      marked_fuse_op_nodes.insert(unused_nodes.begin(), unused_nodes.end());
+    }
+
+    GraphSafeRemoveNodes(graph, marked_fuse_op_nodes);
+    ++fusion_count;
+  };
+
+  gpd(graph, handler);
+  return fusion_count;
+}
+
+void FuseMultiTransformerLayerPass::ApplyImpl(Graph* graph) const {
+  FusePassBase::Init(name_scope_, graph);
+  auto* scope = param_scope();
+  PADDLE_ENFORCE_NOT_NULL(
+      scope,
+      platform::errors::Fatal("During the fuse_multi_transformer_layer pass, "
+                              "The scope should not be null."));
+  int fusion_count = BuildFusion(graph, name_scope_, scope);
+
+  AddStatis(fusion_count);
+}
+
+}  // namespace ir
+}  // namespace framework
+}  // namespace paddle
+
+REGISTER_PASS(fuse_multi_transformer_layer_pass,
+              paddle::framework::ir::FuseMultiTransformerLayerPass);

paddle/fluid/framework/ir/fuse_multi_transformer_layer_pass.h

+// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include <memory>
+#include <string>
+
+#include "paddle/fluid/framework/ir/fuse_pass_base.h"
+#include "paddle/fluid/framework/ir/graph.h"
+
+namespace paddle {
+namespace framework {
+namespace ir {
+namespace patterns {
+
+struct MultiTransformerLayerPattern : public PatternBase {
+  MultiTransformerLayerPattern(PDPattern* pattern,
+                               const std::string& name_scope)
+      : PatternBase(pattern, name_scope, "fuse_multi_transformer_layer") {}
+
+  std::unordered_map<std::string, std::string> operator()(
+      bool enable_int8, int num_fused_op = 1, bool is_decoder = false);
+
+  PATTERN_DECL_NODE(src_mask);
+  PATTERN_DECL_NODE(x0);
+};
+
+}  // namespace patterns
+
+class FuseMultiTransformerLayerPass : public FusePassBase {
+ public:
+  FuseMultiTransformerLayerPass() {}
+  virtual ~FuseMultiTransformerLayerPass() {}
+
+ protected:
+  void ApplyImpl(Graph* graph) const;
+
+  const std::string name_scope_{"fuse_multi_transformer_layer"};
+
+ private:
+  int BuildFusion(Graph* graph,
+                  const std::string& name_scope,
+                  Scope* scope) const;
+};
+
+}  // namespace ir
+}  // namespace framework
+}  // namespace paddle

paddle/fluid/framework/ir/fuse_multi_transformer_layer_pass_tester.cc

+/* Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+    http://www.apache.org/licenses/LICENSE-2.0
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License. */
+
+#include <gtest/gtest.h>
+
+#include "paddle/fluid/framework/ir/fuse_multi_transformer_layer_pass.h"
+#include "paddle/fluid/framework/ir/pass_tester_helper.h"
+#include "paddle/fluid/framework/op_version_registry.h"
+
+#define DEF_INPUT_DATA                                                  \
+  Layers layers;                                                        \
+  int num_layers = 3;                                                   \
+  auto* x = layers.data("x", {1, 128, 1024});                           \
+  auto* src_mask = layers.data("src_mask", {1, 16, 128, 128});          \
+  auto* ln_scale = layers.data("ln_scale", {1024}, true);               \
+  auto* ln_bias = layers.data("ln_bias", {1024}, true);                 \
+  auto* ffn_ln_scale = layers.data("ffn_ln_scale", {1024}, true);       \
+  auto* ffn_ln_bias = layers.data("ffn_ln_bias", {1024}, true);         \
+  auto* qkv_w = layers.data("qkv_w", {3, 16, 64, 1024}, true);          \
+  auto* out_linear_w = layers.data("out_linear_w", {1024, 1024}, true); \
+  auto* ffn1_w = layers.data("ffn1_w", {1024, 4096}, true);             \
+  auto* ffn2_w = layers.data("ffn2_w", {4096, 1024}, true);             \
+  auto* qkv_bias = layers.data("qkv_bias", {3072}, true);               \
+  auto* out_linear_bias = layers.data("out_linear_bias", {1024}, true); \
+  auto* ffn1_bias = layers.data("ffn1_bias", {4096}, true);             \
+  auto* ffn2_bias = layers.data("ffn2_bias", {1024}, true);
+
+namespace paddle {
+namespace framework {
+namespace ir {
+
+void AddVarToScope(Scope* param_scope,
+                   const std::string& name,
+                   const DDim& dims) {
+  auto* tensor = param_scope->Var(name)->GetMutable<phi::DenseTensor>();
+  tensor->Resize(dims);
+  tensor->mutable_data<float>(platform::CPUPlace());
+}
+
+Scope* CreateParamScope() {
+  auto param_scope = new Scope();
+  AddVarToScope(param_scope, "ln_scale", {1024});
+  AddVarToScope(param_scope, "ln_bias", {1024});
+  AddVarToScope(param_scope, "ffn_ln_scale", {1024});
+  AddVarToScope(param_scope, "ffn_ln_bias", {1024});
+
+  AddVarToScope(param_scope, "qkv_w", {3, 16, 64, 1024});
+  AddVarToScope(param_scope, "out_linear_w", {1024, 1024});
+  AddVarToScope(param_scope, "ffn1_w", {1024, 4096});
+  AddVarToScope(param_scope, "ffn2_w", {4096, 1024});
+  AddVarToScope(param_scope, "qkv_bias", {3072});
+  AddVarToScope(param_scope, "out_linear_bias", {1024});
+  AddVarToScope(param_scope, "ffn1_bias", {4096});
+  AddVarToScope(param_scope, "ffn2_bias", {1024});
+
+  return param_scope;
+}
+TEST(FuseMultiTransformerLayerPass, encoder_fp) {
+  DEF_INPUT_DATA
+
+  // Layers
+  for (int i = 0; i < num_layers; ++i) {
+    auto* cache_kv = layers.fill_constant_batch_size_like(
+        x,
+        static_cast<int>(proto::VarType::FP32),
+        0,
+        1,
+        {2, -1, 16, 1024, 64},
+        0);
+    auto* out = layers.fused_multi_transformer(x,
+                                               cache_kv,
+                                               src_mask,
+                                               qkv_w,
+                                               qkv_bias,
+                                               out_linear_w,
+                                               out_linear_bias,
+                                               ffn1_w,
+                                               ffn1_bias,
+                                               ffn2_w,
+                                               ffn2_bias,
+                                               ln_scale,
+                                               ln_bias,
+                                               ffn_ln_scale,
+                                               ffn_ln_bias,
+                                               0.1,
+                                               1e-12);
+
+    x = out;
+  }
+  std::unique_ptr<ir::Graph> graph(new ir::Graph(layers.main_program()));
+  graph->Set("__param_scope__", CreateParamScope());
+  graph->Set(kFusedMultiTransformerEncoderFusionCount, new int(num_layers));
+
+  auto pass = PassRegistry::Instance().Get("fuse_multi_transformer_layer_pass");
+  if (pass.get() == nullptr)
+    LOG(INFO) << "get fuse_multi_transformer_layer_pass failed";
+
+  graph.reset(pass->Apply(graph.release()));
+  int num_nodes_after = GetNumOpNodes(graph, "fused_multi_transformer");
+
+  PADDLE_ENFORCE_EQ(
+      num_nodes_after,
+      1,
+      platform::errors::InvalidArgument(
+          "After the fuse_multi_transformer_layer_pass, "
+          "The node num in graph should be 1, but the result is %d",
+          num_nodes_after));
+}
+TEST(FuseMultiTransformerLayerPass, decoder_fp) {
+  DEF_INPUT_DATA
+
+  x = layers.data("x", {1, 1, 1024});
+  auto* cache_kv = layers.data("cache_kv", {2, 1, 16, 1024, 64}, true);
+  src_mask = layers.data("src_mask", {1, 16, 1, 129});
+
+  // Layers
+  for (int i = 0; i < num_layers; ++i) {
+    auto* shape_out = layers.shape(src_mask);
+    auto* time_stamp = layers.slice(shape_out, {0}, {3}, {4});
+    auto* out = layers.fused_multi_transformer(x,
+                                               cache_kv,
+                                               src_mask,
+                                               qkv_w,
+                                               qkv_bias,
+                                               out_linear_w,
+                                               out_linear_bias,
+                                               ffn1_w,
+                                               ffn1_bias,
+                                               ffn2_w,
+                                               ffn2_bias,
+                                               ln_scale,
+                                               ln_bias,
+                                               ffn_ln_scale,
+                                               ffn_ln_bias,
+                                               0.1,
+                                               1e-12,
+                                               time_stamp);
+
+    x = out;
+  }
+  std::unique_ptr<ir::Graph> graph(new ir::Graph(layers.main_program()));
+  auto param_scope = CreateParamScope();
+  AddVarToScope(param_scope, "cache_kv", {2, 1, 16, 1024, 64});
+  graph->Set("__param_scope__", param_scope);
+
+  graph->Set(kFusedMultiTransformerDecoderFusionCount, new int(num_layers));
+
+  auto pass = PassRegistry::Instance().Get("fuse_multi_transformer_layer_pass");
+  if (pass.get() == nullptr)
+    LOG(INFO) << "get fuse_multi_transformer_layer_pass failed";
+
+  graph.reset(pass->Apply(graph.release()));
+  int num_nodes_after = GetNumOpNodes(graph, "fused_multi_transformer");
+
+  PADDLE_ENFORCE_EQ(
+      num_nodes_after,
+      1,
+      platform::errors::InvalidArgument(
+          "After the fuse_multi_transformer_layer_pass, "
+          "The node num in graph should be 1, but the result is %d",
+          num_nodes_after));
+}
+}  // namespace ir
+}  // namespace framework
+}  // namespace paddle
+
+USE_PASS(fuse_multi_transformer_layer_pass);

paddle/fluid/framework/ir/fused_multi_transformer_decoder_pass.cc

@@ -1565,6 +1565,7 @@ void FusedMultiTransformerDecoderPass::ApplyImpl(Graph* graph) const {
   int fusion_count = BuildFusion(graph, name_scope_, scope);
   if (fusion_count > 0) {
     graph->Set(kFusedMultiTransformerDecoderPass, new bool(true));
+    graph->Set(kFusedMultiTransformerDecoderFusionCount, new int(fusion_count));
   }
   AddStatis(fusion_count);
 }
@@ -2178,6 +2179,7 @@ void FusedMultiTransformerDecoderFuseQKVPass::ApplyImpl(Graph* graph) const {
   int fusion_count = BuildFusion(graph, name_scope_, scope);
   if (fusion_count > 0) {
     graph->Set(kFusedMultiTransformerDecoderFuseQKVPass, new bool(true));
+    graph->Set(kFusedMultiTransformerDecoderFusionCount, new int(fusion_count));
   }
   AddStatis(fusion_count);
 }
@@ -2833,6 +2835,7 @@ void MultiDevicesFusedMultiTransformerDecoderFuseQKVPass::ApplyImpl(
   int fusion_count = BuildFusion(graph, name_scope_, scope);
   if (fusion_count > 0) {
     graph->Set(kFusedMultiTransformerDecoderFuseQKVPass, new bool(true));
+    graph->Set(kFusedMultiTransformerDecoderFusionCount, new int(fusion_count));
   }
   AddStatis(fusion_count);
 }

paddle/fluid/framework/ir/fused_multi_transformer_encoder_pass.cc

@@ -1728,6 +1728,7 @@ void FusedMultiTransformerEncoderPass::ApplyImpl(Graph* graph) const {
   int fusion_count = BuildFusion(graph, name_scope_, scope);
   if (fusion_count > 0) {
     graph->Set(kFusedMultiTransformerEncoderPass, new bool(true));
+    graph->Set(kFusedMultiTransformerEncoderFusionCount, new int(fusion_count));
   }
   AddStatis(fusion_count);
 }
@@ -2380,6 +2381,7 @@ void FusedMultiTransformerEncoderFuseQKVPass::ApplyImpl(Graph* graph) const {
   int fusion_count = BuildFusion(graph, name_scope_, scope);
   if (fusion_count > 0) {
     graph->Set(kFusedMultiTransformerEncoderFuseQKVPass, new bool(true));
+    graph->Set(kFusedMultiTransformerEncoderFusionCount, new int(fusion_count));
   }
   AddStatis(fusion_count);
 }
@@ -3076,6 +3078,7 @@ void MultiDevicesFusedMultiTransformerEncoderFuseQKVPass::ApplyImpl(
   if (fusion_count > 0) {
     graph->Set(kMultiDevicesFusedMultiTransformerEncoderFuseQKVPass,
                new bool(true));
+    graph->Set(kFusedMultiTransformerEncoderFusionCount, new int(fusion_count));
   }
   AddStatis(fusion_count);
 }

paddle/fluid/framework/ir/pass.cc

@@ -46,7 +46,7 @@ static const std::vector<std::string> support_subgraph_passes = {
     "fused_multi_transformer_decoder_fuse_qkv_pass",
     "multi_devices_fused_multi_transformer_encoder_fuse_qkv_pass",
     "multi_devices_fused_multi_transformer_decoder_fuse_qkv_pass",
-};
+    "fuse_multi_transformer_layer_pass"};
 
 Graph *Pass::Apply(Graph *graph) const {
   VLOG(10) << "start to apply pass " << Type() << " to graph";

paddle/fluid/framework/ir/pass.h

@@ -59,6 +59,10 @@ constexpr char kMultiDevicesFusedMultiTransformerEncoderFuseQKVPass[] =
     "multi_devices_fused_multi_transformer_encoder_fuse_qkv_pass_flag";
 constexpr char kMultiDevicesFusedMultiTransformerDecoderFuseQKVPass[] =
     "multi_devices_fused_multi_transformer_decoder_fuse_qkv_pass_flag";
+constexpr char kFusedMultiTransformerEncoderFusionCount[] =
+    "fused_multi_transformer_encoder_fusion_count";
+constexpr char kFusedMultiTransformerDecoderFusionCount[] =
+    "fused_multi_transformer_decoder_fusion_count";
 constexpr char kPrelnEmbEltwiseLayernormPass[] =
     "preln_embedding_eltwise_layernorm_fuse_pass_flag";
 

paddle/fluid/framework/ir/pass_tester_helper.h

@@ -528,6 +528,119 @@ struct Layers {
     return out;
   }
 
+  VarDesc* shape(VarDesc* input) {
+    VarDesc* out = lod_tensor(unique_name());
+    OpDesc* op = program_.MutableBlock(0)->AppendOp();
+    op->SetType("shape");
+    op->SetInput("Input", {input->Name()});
+    op->SetOutput("Out", {out->Name()});
+    return out;
+  }
+
+  VarDesc* slice(VarDesc* input,
+                 std::vector<int> axes,
+                 std::vector<int> starts,
+                 std::vector<int> ends) {
+    VarDesc* out = lod_tensor(unique_name());
+    OpDesc* op = program_.MutableBlock(0)->AppendOp();
+    op->SetType("slice");
+    op->SetInput("Input", {input->Name()});
+    op->SetOutput("Out", {out->Name()});
+    op->SetAttr("axes", axes);
+    op->SetAttr("starts", starts);
+    op->SetAttr("ends", ends);
+    return out;
+  }
+
+  VarDesc* fill_constant_batch_size_like(VarDesc* x,
+                                         int dtype,
+                                         int input_dim_idx,
+                                         int output_dim_idx,
+                                         std::vector<int> shape,
+                                         float value) {
+    VarDesc* out = lod_tensor(unique_name());
+    OpDesc* op = program_.MutableBlock(0)->AppendOp();
+    op->SetType("fill_constant_batch_size_like");
+    op->SetInput("Input", {x->Name()});
+    op->SetAttr("dtype", dtype);
+    op->SetAttr("input_dim_idx", input_dim_idx);
+    op->SetAttr("output_dim_idx", output_dim_idx);
+    op->SetAttr("shape", shape);
+    op->SetAttr("value", value);
+    op->SetOutput("Out", {out->Name()});
+    return out;
+  }
+
+  VarDesc* fused_multi_transformer(VarDesc* x,
+                                   VarDesc* cache_kv,
+                                   VarDesc* src_mask,
+                                   VarDesc* qkv_w,
+                                   VarDesc* qkv_bias,
+                                   VarDesc* out_linear_w,
+                                   VarDesc* out_linear_bias,
+                                   VarDesc* ffn1_w,
+                                   VarDesc* ffn1_bias,
+                                   VarDesc* ffn2_w,
+                                   VarDesc* ffn2_bias,
+                                   VarDesc* ln_scale,
+                                   VarDesc* ln_bias,
+                                   VarDesc* ffn_ln_scale,
+                                   VarDesc* ffn_ln_bias,
+                                   float epsilon,
+                                   float dropout_rate,
+                                   VarDesc* time_stamp = nullptr,
+                                   VarDesc* qkv_out_scale = nullptr,
+                                   VarDesc* out_linear_out_scale = nullptr,
+                                   VarDesc* ffn1_out_scale = nullptr,
+                                   VarDesc* ffn2_out_scale = nullptr,
+                                   std::vector<float> qkv_in_scale = {},
+                                   std::vector<float> out_linear_in_scale = {},
+                                   std::vector<float> ffn1_in_scale = {},
+                                   std::vector<float> ffn2_in_scale = {}) {
+    VarDesc* out = lod_tensor(unique_name());
+    OpDesc* op = program_.MutableBlock(0)->AppendOp();
+    std::string op_type = qkv_out_scale ? "fused_multi_transformer_int8"
+                                        : "fused_multi_transformer";
+    op->SetType(op_type);
+    op->SetInput("X", {x->Name()});
+    op->SetInput("CacheKV", {cache_kv->Name()});
+    op->SetInput("SrcMask", {src_mask->Name()});
+    op->SetInput("QKVW", {qkv_w->Name()});
+    op->SetInput("QKVBias", {qkv_bias->Name()});
+    op->SetInput("OutLinearW", {out_linear_w->Name()});
+    op->SetInput("OutLinearBias", {out_linear_bias->Name()});
+    op->SetInput("FFN1Weight", {ffn1_w->Name()});
+    op->SetInput("FFN1Bias", {ffn1_bias->Name()});
+    op->SetInput("FFN2Weight", {ffn2_w->Name()});
+    op->SetInput("FFN2Bias", {ffn2_bias->Name()});
+    op->SetInput("LnScale", {ln_scale->Name()});
+    op->SetInput("LnBias", {ln_bias->Name()});
+    op->SetInput("FFNLnScale", {ffn_ln_scale->Name()});
+    op->SetInput("FFNLnBias", {ffn_ln_bias->Name()});
+    op->SetAttr("pre_layer_norm", true);
+    op->SetAttr("is_test", true);
+    op->SetAttr("dropout_implementation", "upscale_in_train");
+    op->SetAttr("dropout_rate", dropout_rate);
+    op->SetAttr("epsilon", epsilon);
+    op->SetOutput("Out", {out->Name()});
+
+    if (time_stamp) {
+      op->SetInput("TimeStep", {time_stamp->Name()});
+    }
+
+    if (qkv_out_scale) {
+      op->SetInput("QKVOutScale", {qkv_out_scale->Name()});
+      op->SetInput("OutLinearOutScale", {out_linear_out_scale->Name()});
+      op->SetInput("FFN1OutScale", {ffn1_out_scale->Name()});
+      op->SetInput("FFN2OutScale", {ffn2_out_scale->Name()});
+      op->SetAttr("qkv_in_scale", qkv_in_scale);
+      op->SetAttr("out_linear_in_scale", out_linear_in_scale);
+      op->SetAttr("ffn1_in_scale", ffn1_in_scale);
+      op->SetAttr("ffn2_in_scale", ffn2_in_scale);
+    }
+    return out;
+  }
+
   void backward(std::vector<VarDesc*> targets) {
     // This function is designed to simulate the structure of training program,
     //  but is constructed differently as the actual program.

paddle/fluid/inference/api/paddle_pass_builder.cc

@@ -205,6 +205,7 @@ GpuPassStrategy::GpuPassStrategy() : PassStrategy({}) {
         "fused_multi_transformer_decoder_fuse_qkv_pass",                //
         "multi_devices_fused_multi_transformer_encoder_fuse_qkv_pass",  //
         "multi_devices_fused_multi_transformer_decoder_fuse_qkv_pass",  //
+        "fuse_multi_transformer_layer_pass",                            //
         "gpu_cpu_squeeze2_matmul_fuse_pass",                            //
         "gpu_cpu_reshape2_matmul_fuse_pass",                            //
         "gpu_cpu_flatten2_matmul_fuse_pass",                            //