graph_pattern_detector.cc 46.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14
// Copyright (c) 2018 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.

15
#include <algorithm>
Q
Qiao Longfei 已提交
16
#include <array>
17 18 19 20
#include <string>
#include <vector>

#include "paddle/fluid/framework/ir/graph_helper.h"
21
#include "paddle/fluid/framework/ir/graph_pattern_detector.h"
22
#include "paddle/fluid/framework/ir/graph_traits.h"
23
#include "paddle/fluid/framework/ir/graph_viz_pass.h"
C
chengduo 已提交
24
#include "paddle/fluid/framework/operator.h"
25
#include "paddle/fluid/platform/enforce.h"
Y
Yan Chunwei 已提交
26
#include "paddle/fluid/string/pretty_log.h"
Y
Yan Chunwei 已提交
27
#include "paddle/fluid/string/printf.h"
28

29 30 31 32
namespace paddle {
namespace framework {
namespace ir {

Y
Yan Chunwei 已提交
33 34 35 36
using string::PrettyLogEndl;
using string::PrettyLog;
using string::Style;

37 38
size_t PDPattern::id_ = 0UL;

C
chengduo 已提交
39
PDNode *PDPattern::NewNode(const std::string &name) {
Y
Yan Chunwei 已提交
40
  if (!name.empty()) {
T
tensor-tang 已提交
41
    PADDLE_ENFORCE_EQ(node_map_.count(name), 0UL,
Y
Yan Chunwei 已提交
42 43 44 45 46
                      "PDNode's name should be unique, get duplicate [%s]",
                      name);
  }

  nodes_.emplace_back(new PDNode(this, name));
C
chengduo 已提交
47
  auto *cur = nodes_.back().get();
Y
Yan Chunwei 已提交
48 49 50 51
  node_map_[name] = cur;
  return cur;
}

C
chengduo 已提交
52
PDNode *PDPattern::NewNode(PDNode::teller_t &&teller, const std::string &name) {
53
  if (!name.empty()) {
T
tensor-tang 已提交
54
    PADDLE_ENFORCE_EQ(node_map_.count(name), 0UL,
55 56 57 58
                      "PDNode's name should be unique, get duplicate [%s]",
                      name);
  }

59
  nodes_.emplace_back(new PDNode(std::move(teller), this, name));
C
chengduo 已提交
60
  auto *cur = nodes_.back().get();
61
  node_map_[name] = cur;
62 63 64
  return cur;
}

C
chengduo 已提交
65
PDNode *PDPattern::RetrieveNode(const std::string &id) const {
66 67 68 69 70 71 72 73
  auto it = node_map_.find(id);
  if (it == node_map_.end()) {
    return nullptr;
  }

  return it->second;
}

C
chengduo 已提交
74
void PDPattern::AddEdge(PDNode *a, PDNode *b) {
75 76 77 78 79 80
  PADDLE_ENFORCE(a);
  PADDLE_ENFORCE(b);
  PADDLE_ENFORCE(a != b, "can't connect to the same nodes.");
  edges_.emplace_back(a, b);
}

C
chengduo 已提交
81
void GraphPatternDetector::operator()(Graph *graph,
82
                                      GraphPatternDetector::handle_t handler) {
83 84 85 86
  if (!MarkPDNodesInGraph(*graph)) {
    return;
  }

87 88 89
  auto subgraphs = DetectPatterns();
  UniquePatterns(&subgraphs);
  RemoveOverlappedMatch(&subgraphs);
Y
Yan Chunwei 已提交
90
  ValidateByNodeRole(&subgraphs);
91

Y
Yan Chunwei 已提交
92
  if (subgraphs.empty()) return;
Y
Yan Chunwei 已提交
93
  PrettyLogEndl(Style::detail(), "---  detect %d subgraphs", subgraphs.size());
94
  int id = 0;
C
chengduo 已提交
95
  for (auto &g : subgraphs) {
M
minqiyang 已提交
96
    VLOG(3) << "optimizing #" << id++ << " subgraph";
97 98 99 100
    handler(g, graph);
  }
}

C
chengduo 已提交
101
bool GraphPatternDetector::MarkPDNodesInGraph(const ir::Graph &graph) {
M
minqiyang 已提交
102
  VLOG(3) << "mark pdnodes in graph";
103 104
  if (graph.Nodes().empty()) return false;

C
chengduo 已提交
105 106
  for (auto &node : GraphTraits::DFS(graph)) {
    for (const auto &pdnode : pattern_.nodes()) {
107
      if (pdnode->Tell(&node)) {
108
        VLOG(4) << "Node " << node.Name() << " marked as " << pdnode->name();
109 110 111 112
        pdnodes2nodes_[pdnode.get()].insert(&node);
      }
    }
  }
Y
Yan Chunwei 已提交
113
  // Check to early stop if some PDNode can't find matched Node.
C
chengduo 已提交
114
  for (auto &pdnode : pattern_.nodes()) {
Y
Yan Chunwei 已提交
115
    if (!pdnodes2nodes_.count(pdnode.get())) {
M
minqiyang 已提交
116
      VLOG(4) << pdnode->name() << " can't find matched Node, early stop";
Y
Yan Chunwei 已提交
117
      // return false;
Y
Yan Chunwei 已提交
118 119
    }
  }
M
minqiyang 已提交
120
  VLOG(3) << pdnodes2nodes_.size() << " nodes marked";
121

122 123 124
  return !pdnodes2nodes_.empty();
}

Y
Yan Chunwei 已提交
125 126 127
// The intermediate Nodes can only link to the nodes inside the pattern, or this
// subgraph will be droped.
void GraphPatternDetector::ValidateByNodeRole(
C
chengduo 已提交
128
    std::vector<GraphPatternDetector::subgraph_t> *subgraphs) {
Y
Yan Chunwei 已提交
129 130 131 132 133
  std::vector<GraphPatternDetector::subgraph_t> result;

  subgraphs->erase(
      std::remove_if(
          subgraphs->begin(), subgraphs->end(),
C
chengduo 已提交
134
          [](const GraphPatternDetector::subgraph_t &subgraph) -> bool {
Y
Yan Chunwei 已提交
135
            // Collect the inputs and outputs.
C
chengduo 已提交
136 137
            std::unordered_set<Node *> ios;
            for (auto &item : subgraph) {
Y
Yan Chunwei 已提交
138 139 140 141
              if (!item.first->IsIntermediate()) {
                ios.insert(item.second);
              }
            }
C
chengduo 已提交
142
            for (auto &item : subgraph) {
Y
Yan Chunwei 已提交
143
              if (item.first->IsIntermediate()) {
C
chengduo 已提交
144
                for (auto *x : item.second->inputs) {
Y
Yan Chunwei 已提交
145 146 147 148
                  if (!ios.count(x)) {
                    return true;
                  }
                }
C
chengduo 已提交
149
                for (auto *x : item.second->outputs) {
Y
Yan Chunwei 已提交
150 151 152 153 154 155 156 157 158 159 160
                  if (!ios.count(x)) {
                    return true;
                  }
                }
              }
            }
            return false;
          }),
      subgraphs->end());
}

161
struct HitGroup {
C
chengduo 已提交
162
  std::unordered_map<PDNode *, Node *> roles;
163

C
chengduo 已提交
164
  bool Match(Node *node, PDNode *pat) {
165
    if (nodes_.count(node)) {
T
Tao Luo 已提交
166 167 168 169 170
      if (roles.count(pat) && roles[pat] == node) return true;
      return false;
    } else {
      if (roles.count(pat) && roles[pat] != node) return false;
      return true;
171
    }
172 173
  }

C
chengduo 已提交
174
  void Register(Node *node, PDNode *pat) {
175 176 177 178 179
    roles[pat] = node;
    nodes_.insert(node);
  }

 private:
C
chengduo 已提交
180
  std::unordered_set<Node *> nodes_;
181 182 183
};

// Tell whether Node a links to b.
C
chengduo 已提交
184 185
bool IsNodesLink(Node *a, Node *b) {
  for (auto *node : a->outputs) {
186 187 188 189 190 191 192
    if (b == node) {
      return true;
    }
  }
  return false;
}

193 194
std::vector<GraphPatternDetector::subgraph_t>
GraphPatternDetector::DetectPatterns() {
195
  // Init empty subgraphs.
196
  std::vector<GraphPatternDetector::subgraph_t> result;
197
  std::vector<HitGroup> init_groups;
198
  std::array<std::vector<HitGroup>, 2> bi_records;
C
chengduo 已提交
199
  auto *first_pnode = pattern_.edges().empty() ? pattern().nodes().front().get()
200
                                               : pattern_.edges().front().first;
201
  if (!pdnodes2nodes_.count(first_pnode)) return result;
C
chengduo 已提交
202
  for (auto *node : pdnodes2nodes_[first_pnode]) {
203 204 205 206 207 208 209 210 211 212
    HitGroup group;
    group.roles[first_pnode] = node;
    init_groups.emplace_back(group);
  }

  int step = 0;
  bi_records[0] = std::move(init_groups);

  // Extend a PDNode to subgraphs by deducing the connection relations defined
  // in edges of PDNodes.
C
chengduo 已提交
213
  for (const auto &edge : pattern_.edges()) {
M
minqiyang 已提交
214
    VLOG(4) << "check " << edge.first->name() << " -> " << edge.second->name();
Y
Yan Chunwei 已提交
215
    // TODO(Superjomn) Fix bug here, the groups might be duplicate here.
216 217
    // Each role has two PDNodes, which indicates two roles.
    // Detect two Nodes that can match these two roles and they are connected.
C
chengduo 已提交
218 219
    auto &pre_groups = bi_records[step % 2];
    auto &cur_groups = bi_records[1 - (step++ % 2)];
220
    cur_groups.clear();
221
    if (pre_groups.empty()) break;
222
    // source -> target
C
chengduo 已提交
223 224
    for (Node *source : pdnodes2nodes_[edge.first]) {
      for (Node *target : pdnodes2nodes_[edge.second]) {
M
minqiyang 已提交
225
        VLOG(8) << "check " << source->id() << " -- " << target->id();
226
        // TODO(Superjomn) add some prune strategies.
C
chengduo 已提交
227
        for (const auto &group : pre_groups) {
T
Tao Luo 已提交
228 229 230 231 232 233
          if (IsNodesLink(source, target)) {
            HitGroup new_group = group;
            bool flag = new_group.Match(source, edge.first) &&
                        new_group.Match(target, edge.second);
            if (flag) {
              new_group.Register(source, edge.first);
234 235 236 237 238 239 240 241
              new_group.Register(target, edge.second);
              cur_groups.push_back(new_group);
              // TODO(Superjomn) need to unique
            }
          }
        }
      }
    }
M
minqiyang 已提交
242
    VLOG(3) << "step " << step << " get records: " << cur_groups.size();
C
chengduo 已提交
243 244
    for (auto &group : cur_groups) {
      for (auto &item : group.roles) {
M
minqiyang 已提交
245
        VLOG(4) << "node " << item.second->id() << " as " << item.first->name();
Y
Yan Chunwei 已提交
246
      }
M
minqiyang 已提交
247
      VLOG(4) << "=========================================================";
Y
Yan Chunwei 已提交
248
    }
249 250
  }

C
chengduo 已提交
251
  for (auto &group : bi_records[step % 2]) {
252
    GraphPatternDetector::subgraph_t subgraph;
C
chengduo 已提交
253
    for (auto &role : group.roles) {
254 255 256 257 258 259 260
      subgraph.emplace(role.first, role.second);
    }
    result.emplace_back(subgraph);
  }
  return result;
}

Y
Yan Chunwei 已提交
261 262
struct GraphItemLessThan {
  bool operator()(const std::pair<PDNode *, Node *> &a,
Y
Yan Chunwei 已提交
263
                  const std::pair<PDNode *, Node *> &b) {
Y
Yan Chunwei 已提交
264 265 266 267 268
    if (a.first != b.first) {
      return a.first < b.first;
    } else {
      return a.second < b.second;
    }
Y
Yan Chunwei 已提交
269
  }
Y
Yan Chunwei 已提交
270
};
Y
Yan Chunwei 已提交
271

272 273
// TODO(Superjomn) enhance the function as it marks unique unique as duplicates
// see https://github.com/PaddlePaddle/Paddle/issues/13550
274
void GraphPatternDetector::UniquePatterns(
C
chengduo 已提交
275
    std::vector<GraphPatternDetector::subgraph_t> *subgraphs) {
276
  if (subgraphs->empty()) return;
277
  std::vector<GraphPatternDetector::subgraph_t> result;
278 279

  std::unordered_set<size_t> set;
Y
Yan Chunwei 已提交
280
  std::hash<std::string> hasher;
C
chengduo 已提交
281
  for (auto &g : *subgraphs) {
Y
Yan Chunwei 已提交
282 283
    // Sort the items in the sub-graph, and transform to a string key.
    std::vector<std::pair<PDNode *, Node *>> sorted_keys(g.begin(), g.end());
Y
Yan Chunwei 已提交
284
    std::sort(sorted_keys.begin(), sorted_keys.end(), GraphItemLessThan());
Y
Yan Chunwei 已提交
285 286 287
    std::stringstream ss;
    for (auto &item : sorted_keys) {
      ss << item.first << ":" << item.second;
288
    }
Y
Yan Chunwei 已提交
289
    auto key = hasher(ss.str());
290 291 292 293 294 295 296 297
    if (!set.count(key)) {
      result.emplace_back(g);
      set.insert(key);
    }
  }
  *subgraphs = result;
}

298
void GraphPatternDetector::RemoveOverlappedMatch(
C
chengduo 已提交
299
    std::vector<subgraph_t> *subgraphs) {
300
  std::vector<subgraph_t> result;
C
chengduo 已提交
301
  std::unordered_set<Node *> node_set;
302

C
chengduo 已提交
303
  for (const auto &subgraph : *subgraphs) {
304
    bool valid = true;
C
chengduo 已提交
305
    for (auto &item : subgraph) {
Y
Yan Chunwei 已提交
306
      if (item.first->IsIntermediate() && node_set.count(item.second)) {
307 308 309 310 311
        valid = false;
        break;
      }
    }
    if (valid) {
C
chengduo 已提交
312
      for (auto &item : subgraph) {
313 314 315 316 317 318 319 320
        node_set.insert(item.second);
      }
      result.push_back(subgraph);
    }
  }
  *subgraphs = result;
}

321 322 323 324 325
std::string PDPattern::DotString() const {
  using inference::analysis::Dot;
  Dot dot;
  int id = 0;
  // Create Nodes
C
chengduo 已提交
326 327
  std::unordered_map<PDNode *, std::string> node2dot;
  for (const auto &node : nodes()) {
328 329 330 331 332
    std::string node_id = "Node" + std::to_string(id++);
    dot.AddNode(node_id, {}, node->name());
    node2dot[node.get()] = node_id;
  }
  // Create Edges
C
chengduo 已提交
333
  for (const auto &edge : edges()) {
334 335 336 337
    if (!node2dot.count(edge.first) || !node2dot.count(edge.second)) {
      LOG(ERROR) << "no node " << edge.first << " " << edge.second;
      continue;
    }
C
chengduo 已提交
338 339
    auto &src = node2dot.at(edge.first);
    auto &trg = node2dot.at(edge.second);
340 341 342 343 344
    dot.AddEdge(src, trg, {});
  }
  return dot.Build();
}

C
chengduo 已提交
345
PDNode &PDNode::LinksTo(const std::vector<PDNode *> &others) {
346
  // extend outlinks.
C
chengduo 已提交
347
  for (PDNode *x : others) {
348 349 350 351 352
    pattern_->AddEdge(this, x);
  }
  return *this;
}

C
chengduo 已提交
353
PDNode &PDNode::LinksFrom(const std::vector<PDNode *> &others) {
354
  // extend outlinks.
C
chengduo 已提交
355
  for (PDNode *x : others) {
356 357 358 359 360
    pattern_->AddEdge(x, this);
  }
  return *this;
}

C
chengduo 已提交
361 362
PDNode *PDNode::assert_is_op() {
  asserts_.emplace_back([](Node *x) { return x && x->IsOp(); });
Y
Yan Chunwei 已提交
363 364
  return this;
}
C
chengduo 已提交
365 366 367

PDNode *PDNode::assert_is_op(const std::string &op_type) {
  asserts_.emplace_back([op_type](Node *x) {
Y
Yan Chunwei 已提交
368 369 370 371
    return x && x->IsOp() && x->Op()->Type() == op_type;
  });
  return this;
}
C
chengduo 已提交
372 373 374 375 376 377 378 379

PDNode *PDNode::assert_is_var() {
  asserts_.emplace_back([](Node *x) { return x && x->IsVar(); });
  return this;
}

PDNode *PDNode::assert_is_not_ctrl_var() {
  asserts_.emplace_back([](Node *x) { return x && !x->IsCtrlVar(); });
Y
Yan Chunwei 已提交
380 381
  return this;
}
C
chengduo 已提交
382 383

PDNode *PDNode::assert_var_not_persistable() {
Y
Yan Chunwei 已提交
384
  assert_is_var();
C
chengduo 已提交
385
  asserts_.emplace_back([](Node *x) { return !x->Var()->Persistable(); });
Y
Yan Chunwei 已提交
386 387
  return this;
}
C
chengduo 已提交
388 389

PDNode *PDNode::assert_is_persistable_var() {
Y
Yan Chunwei 已提交
390
  assert_is_var();
C
chengduo 已提交
391
  asserts_.emplace_back([=](Node *x) { return x->Var()->Persistable(); });
Y
Yan Chunwei 已提交
392 393
  return this;
}
C
chengduo 已提交
394 395 396

PDNode *PDNode::assert_is_op_nth_input(const std::string &op_type,
                                       const std::string &argument, int nth) {
Y
Yan Chunwei 已提交
397 398
  assert_is_var();
  assert_is_op_input(op_type);
C
chengduo 已提交
399 400
  asserts_.emplace_back([=](Node *x) {
    for (auto *op : x->outputs) {
401 402 403
      if (op->IsOp() && op->Op()->Type() == op_type &&
          IsNthInput(x, op, argument, nth))
        return true;
Y
Yan Chunwei 已提交
404 405 406 407 408
    }
    return false;
  });
  return this;
}
C
chengduo 已提交
409 410 411

PDNode *PDNode::assert_is_op_nth_output(const std::string &op_type,
                                        const std::string &argument, int nth) {
Y
Yan Chunwei 已提交
412
  assert_is_var();
C
chengduo 已提交
413 414
  asserts_.emplace_back([=](Node *x) {
    for (auto *op : x->inputs) {
415 416 417
      if (op->IsOp() && op->Op()->Type() == op_type &&
          IsNthOutput(x, op, argument, nth))
        return true;
Y
Yan Chunwei 已提交
418 419 420 421 422
    }
    return false;
  });
  return this;
}
C
chengduo 已提交
423 424

PDNode *PDNode::assert_is_only_input_of_op(const std::string &op_type) {
Y
Yan Chunwei 已提交
425
  assert_is_var();
C
chengduo 已提交
426 427
  asserts_.emplace_back([=](Node *x) {
    for (auto *op : x->outputs) {
Y
Yan Chunwei 已提交
428 429 430 431 432 433 434 435 436
      if (op && op->IsOp() && op->Op() && op->Op()->Type() == op_type &&
          op->inputs.size() == 1) {
        return true;
      }
    }
    return false;
  });
  return this;
}
C
chengduo 已提交
437 438

PDNode *PDNode::assert_is_only_output_of_op(const std::string &op_type) {
Y
Yan Chunwei 已提交
439
  assert_is_var();
C
chengduo 已提交
440 441
  asserts_.emplace_back([=](Node *x) {
    for (auto *op : x->inputs) {
Y
Yan Chunwei 已提交
442 443 444 445 446 447 448 449 450
      if (op && op->IsOp() && op->Op() && op->Op()->Type() == op_type &&
          op->outputs.size() == 1) {
        return true;
      }
    }
    return false;
  });
  return this;
}
C
chengduo 已提交
451 452

PDNode *PDNode::assert_is_op_output(const std::string &op_type) {
Y
Yan Chunwei 已提交
453
  assert_is_var();
C
chengduo 已提交
454 455
  asserts_.emplace_back([=](Node *x) {
    for (auto *op : x->inputs) {
Y
Yan Chunwei 已提交
456 457 458 459 460 461 462 463
      if (op && op->IsOp() && op->Op() && op->Op()->Type() == op_type) {
        return true;
      }
    }
    return false;
  });
  return this;
}
C
chengduo 已提交
464 465 466

PDNode *PDNode::assert_is_op_output(const std::string &op_type,
                                    const std::string &argument) {
467 468 469 470
  assert_is_var();
  assert_is_op_nth_output(op_type, argument, 0);
  return this;
}
C
chengduo 已提交
471
PDNode *PDNode::assert_is_op_input(const std::string &op_type) {
Y
Yan Chunwei 已提交
472
  assert_is_var();
C
chengduo 已提交
473 474
  asserts_.emplace_back([=](Node *x) {
    for (auto *op : x->outputs) {
Y
Yan Chunwei 已提交
475 476 477 478 479 480 481 482
      if (op && op->IsOp() && op->Op() && op->Op()->Type() == op_type) {
        return true;
      }
    }
    return false;
  });
  return this;
}
C
chengduo 已提交
483 484 485

PDNode *PDNode::assert_is_op_input(const std::string &op_type,
                                   const std::string &argument) {
486 487 488 489
  assert_is_var();
  assert_is_op_nth_input(op_type, argument, 0);
  return this;
}
C
chengduo 已提交
490 491

PDNode *PDNode::assert_op_has_n_inputs(const std::string &op_type, size_t n) {
Y
Yan Chunwei 已提交
492
  assert_is_op(op_type);
C
chengduo 已提交
493
  asserts_.emplace_back([=](Node *x) { return x->inputs.size() == n; });
Y
Yan Chunwei 已提交
494 495
  return this;
}
C
chengduo 已提交
496 497

PDNode *PDNode::assert_op_has_n_outputs(const std::string &op_type, size_t n) {
Y
Yan Chunwei 已提交
498
  assert_is_op(op_type);
C
chengduo 已提交
499
  asserts_.emplace_back([=](Node *x) { return x->outputs.size() == n; });
Y
Yan Chunwei 已提交
500 501
  return this;
}
C
chengduo 已提交
502 503

PDNode *PDNode::assert_more(PDNode::teller_t &&teller) {
Y
Yan Chunwei 已提交
504 505 506 507
  asserts_.emplace_back(std::move(teller));
  return this;
}

C
chengduo 已提交
508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590
PDNode *PDNode::assert_is_ops(const std::unordered_set<std::string> &op_types) {
  asserts_.emplace_back([op_types](Node *x) {
    return x && x->IsOp() && op_types.count(x->Op()->Type());
  });
  return this;
}

PDNode *PDNode::assert_is_ops_nth_input(
    const std::unordered_set<std::string> &op_types,
    const std::string &argument, int nth) {
  assert_is_var();
  assert_is_ops_input(op_types);
  asserts_.emplace_back([=](Node *x) {
    for (auto *op : x->outputs) {
      if (op->IsOp() && op_types.count(op->Op()->Type()) &&
          IsNthInput(x, op, argument, nth))
        return true;
    }
    return false;
  });
  return this;
}

PDNode *PDNode::assert_is_ops_nth_output(
    const std::unordered_set<std::string> &op_types,
    const std::string &argument, int nth) {
  assert_is_var();
  asserts_.emplace_back([=](Node *x) {
    for (auto *op : x->inputs) {
      if (op->IsOp() && op_types.count(op->Op()->Type()) &&
          IsNthOutput(x, op, argument, nth))
        return true;
    }
    return false;
  });
  return this;
}
PDNode *PDNode::assert_is_ops_output(
    const std::unordered_set<std::string> &op_types) {
  assert_is_var();
  asserts_.emplace_back([=](Node *x) {
    for (auto *op : x->inputs) {
      if (op && op->IsOp() && op->Op() && op_types.count(op->Op()->Type())) {
        return true;
      }
    }
    return false;
  });
  return this;
}

PDNode *PDNode::assert_is_ops_output(
    const std::unordered_set<std::string> &op_types,
    const std::string &argument) {
  assert_is_var();
  assert_is_ops_nth_output(op_types, argument, 0);
  return this;
}

PDNode *PDNode::assert_is_ops_input(
    const std::unordered_set<std::string> &op_types) {
  assert_is_var();
  asserts_.emplace_back([=](Node *x) {
    for (auto *op : x->outputs) {
      if (op && op->IsOp() && op->Op() && op_types.count(op->Op()->Type())) {
        return true;
      }
    }
    return false;
  });
  return this;
}

PDNode *PDNode::assert_is_ops_input(
    const std::unordered_set<std::string> &op_types,
    const std::string &argument) {
  assert_is_var();
  assert_is_ops_nth_input(op_types, argument, 0);
  return this;
}

bool VarLinksToOp(Node *node, const std::string &op_type) {
  for (auto *out : node->outputs) {
591 592 593 594 595 596
    if (out->IsOp() && out->Op()->Type() == op_type) {
      return true;
    }
  }
  return false;
}
C
chengduo 已提交
597 598

bool IsNthInput(Node *var, Node *op, const std::string &argument, size_t nth) {
599 600 601 602 603
  PADDLE_ENFORCE(var->IsVar());
  PADDLE_ENFORCE(op->IsOp());
  if (op->Op()->Input(argument).size() <= nth) return false;
  return var->Name() == op->Op()->Input(argument)[nth];
}
C
chengduo 已提交
604 605

bool IsNthOutput(Node *var, Node *op, const std::string &argument, size_t nth) {
606 607 608 609 610
  PADDLE_ENFORCE(var->IsVar());
  PADDLE_ENFORCE(op->IsOp());
  if (op->Op()->Output(argument).size() <= nth) return false;
  return var->Name() == op->Op()->Output(argument)[nth];
}
C
chengduo 已提交
611 612 613 614 615

void GraphSafeRemoveNodes(Graph *graph,
                          const std::unordered_set<const Node *> &nodes) {
  for (auto *node : nodes) {
    graph->RemoveNode(const_cast<Node *>(node));
616 617
  }

C
chengduo 已提交
618
  for (auto *node : graph->Nodes()) {
619 620
    for (auto it = node->inputs.begin(); it != node->inputs.end();) {
      if (nodes.count(*it)) {
C
chengduo 已提交
621
        it = const_cast<Node *>(node)->inputs.erase(it);
622
      } else {
623
        it++;
624
      }
625 626 627
    }
    for (auto it = node->outputs.begin(); it != node->outputs.end();) {
      if (nodes.count(*it)) {
C
chengduo 已提交
628
        it = const_cast<Node *>(node)->outputs.erase(it);
629
      } else {
630
        it++;
631
      }
632 633 634
    }
  }
}
C
chengduo 已提交
635 636 637

bool VarLinksFromOp(Node *node, const std::string &op_type) {
  for (auto *out : node->inputs) {
638 639 640 641 642 643 644
    if (out->IsOp() && out->Op()->Type() == op_type) {
      return true;
    }
  }
  return false;
}

S
Sylwester Fraczek 已提交
645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750
PDNode *patterns::ConvBN::operator()(paddle::framework::ir::PDNode *conv_input,
                                     bool with_eltwise_add) {
  // Create Operators
  conv_input->assert_is_op_input("conv2d", "Input");
  auto *conv_op = pattern->NewNode(conv_repr())->assert_is_op("conv2d");

  PDNode *eltwise_op = nullptr;
  if (with_eltwise_add) {
    eltwise_op =
        pattern->NewNode(eltwise_repr())->assert_is_op("elementwise_add");
  }
  auto *batch_norm_op =
      pattern->NewNode(batch_norm_repr())->assert_is_op("batch_norm");
  // Create variables
  // Conv Filter
  auto *conv_weight_var = pattern->NewNode(conv_weight_repr())
                              ->AsInput()
                              ->assert_is_persistable_var()
                              ->assert_is_op_input("conv2d", "Filter");

  auto *conv_out_var = pattern->NewNode(conv_out_repr())
                           ->AsIntermediate()
                           ->assert_is_only_output_of_op("conv2d");

  PDNode *eltwise_y_in_var = nullptr;
  PDNode *eltwise_out_var = nullptr;
  if (with_eltwise_add) {
    // Conv output as Bias input
    conv_out_var->assert_is_op_input("elementwise_add", "X");
    // Bias
    eltwise_y_in_var = pattern->NewNode(eltwise_y_in_repr())
                           ->assert_is_op_input("elementwise_add", "Y")
                           ->AsInput();
    eltwise_out_var = pattern->NewNode(eltwise_out_repr())
                          ->AsIntermediate()
                          ->assert_is_only_output_of_op("elementwise_add");
  } else {
    // Conv output as BN input
    conv_out_var->assert_is_op_input("batch_norm", "X");
  }

  // BN Scale
  auto *bn_scale_var = pattern->NewNode(bn_scale_repr())
                           ->AsInput()
                           ->assert_is_persistable_var()
                           ->assert_is_op_input("batch_norm", "Scale");
  // BN Bias
  auto *bn_bias_var = pattern->NewNode(bn_bias_repr())
                          ->AsInput()
                          ->assert_is_persistable_var()
                          ->assert_is_op_input("batch_norm", "Bias");
  // BN Mean
  auto *bn_mean_var = pattern->NewNode(bn_mean_repr())
                          ->AsInput()
                          ->assert_is_persistable_var()
                          ->assert_is_op_input("batch_norm", "Mean");
  // BN Variance
  auto *bn_variance_var = pattern->NewNode(bn_variance_repr())
                              ->AsInput()
                              ->assert_is_persistable_var()
                              ->assert_is_op_input("batch_norm", "Variance");

  // BN output
  auto *bn_out_var = pattern->NewNode(bn_out_repr())
                         ->AsOutput()
                         ->assert_is_op_output("batch_norm");

  auto *bn_mean_out_var = pattern->NewNode(bn_mean_out_repr())
                              ->AsOutput()
                              ->assert_is_op_output("batch_norm", "MeanOut");

  auto *bn_variance_out_var =
      pattern->NewNode(bn_variance_out_repr())
          ->AsOutput()
          ->assert_is_op_output("batch_norm", "VarianceOut");

  auto *bn_saved_mean_var =
      pattern->NewNode(bn_saved_mean_repr())
          ->AsOutput()
          ->assert_is_op_output("batch_norm", "SavedMean");

  auto *bn_saved_variance_var =
      pattern->NewNode(bn_saved_variance_repr())
          ->AsOutput()
          ->assert_is_op_output("batch_norm", "SavedVariance");

  conv_op->LinksFrom({conv_input, conv_weight_var}).LinksTo({conv_out_var});

  if (with_eltwise_add) {
    eltwise_op->LinksFrom({conv_out_var, eltwise_y_in_var})
        .LinksTo({eltwise_out_var});
    batch_norm_op
        ->LinksFrom({eltwise_out_var, bn_scale_var, bn_bias_var, bn_mean_var,
                     bn_variance_var})
        .LinksTo({bn_out_var, bn_mean_out_var, bn_variance_out_var,
                  bn_saved_mean_var, bn_saved_variance_var});
  } else {
    batch_norm_op
        ->LinksFrom({conv_out_var, bn_scale_var, bn_bias_var, bn_mean_var,
                     bn_variance_var})
        .LinksTo({bn_out_var, bn_mean_out_var, bn_variance_out_var,
                  bn_saved_mean_var, bn_saved_variance_var});
  }
  return bn_out_var;
}

C
chengduo 已提交
751 752
PDNode *patterns::ConvReLU::operator()(
    paddle::framework::ir::PDNode *conv_input) {
753 754
  // Create Operators
  conv_input->assert_is_op_input("conv2d", "Input");
C
chengduo 已提交
755 756
  auto *conv_op = pattern->NewNode(conv_repr())->assert_is_op("conv2d");
  auto *relu_op = pattern->NewNode(relu_repr())->assert_is_op("relu");
757 758
  // Create variables
  // Filter
C
chengduo 已提交
759
  auto *conv_weight_var = pattern->NewNode(conv_weight_repr())
760 761 762 763
                              ->AsInput()
                              ->assert_is_persistable_var()
                              ->assert_is_op_input("conv2d", "Filter");
  // intermediate variable, will be removed in the IR after fuse.
C
chengduo 已提交
764
  auto *conv_out_var = pattern->NewNode(conv_out_repr())
765 766 767 768
                           ->AsIntermediate()
                           ->assert_is_only_output_of_op("conv2d")
                           ->assert_is_op_input("relu");
  // output
C
chengduo 已提交
769
  auto *relu_out_var = pattern->NewNode(relu_out_repr())
770 771 772
                           ->AsOutput()
                           ->assert_is_op_output("relu");

773
  conv_op->LinksFrom({conv_input, conv_weight_var}).LinksTo({conv_out_var});
774 775 776 777
  relu_op->LinksFrom({conv_out_var}).LinksTo({relu_out_var});
  return relu_out_var;
}

T
tensor-tang 已提交
778 779 780 781
PDNode *patterns::SeqConvEltAddRelu::operator()(
    paddle::framework::ir::PDNode *seqconv_input) {
  // Create Operators
  seqconv_input->assert_is_op_input("sequence_conv", "X");
T
tensor-tang 已提交
782 783 784 785
  auto *seqconv_op = pattern->NewNode(seqconv_repr())
                         ->assert_is_op("sequence_conv")
                         ->assert_op_attr<bool>("paddingTrainable", false)
                         ->assert_op_attr<int>("contextStride", 1);
T
tensor-tang 已提交
786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822

  auto *eltadd_op =
      pattern->NewNode(eltadd_repr())->assert_is_op("elementwise_add");
  auto *relu_op = pattern->NewNode(relu_repr())->assert_is_op("relu");
  // Create variables
  // Filter
  auto *seqconv_weight_var =
      pattern->NewNode(seqconv_weight_repr())
          ->AsInput()
          ->assert_is_persistable_var()
          ->assert_is_op_input("sequence_conv", "Filter");
  // Bias
  auto *eltadd_bias_var = pattern->NewNode(eltadd_bias_repr())
                              ->AsInput()
                              ->assert_is_op_input("elementwise_add");
  // intermediate variable, will be removed in the IR after fuse.
  auto *seqconv_out_var = pattern->NewNode(seqconv_out_repr())
                              ->AsIntermediate()
                              ->assert_is_only_output_of_op("sequence_conv")
                              ->assert_is_op_input("elementwise_add");
  auto *eltadd_out_var = pattern->NewNode(eltadd_out_repr())
                             ->AsIntermediate()
                             ->assert_is_only_output_of_op("elementwise_add")
                             ->assert_is_only_input_of_op("relu");
  // output
  auto *relu_out_var = pattern->NewNode(relu_out_repr())
                           ->AsOutput()
                           ->assert_is_op_output("relu");

  seqconv_op->LinksFrom({seqconv_input, seqconv_weight_var})
      .LinksTo({seqconv_out_var});
  eltadd_op->LinksFrom({seqconv_out_var, eltadd_bias_var})
      .LinksTo({eltadd_out_var});
  relu_op->LinksFrom({eltadd_out_var}).LinksTo({relu_out_var});
  return relu_out_var;
}

C
chengduo 已提交
823
PDNode *patterns::FC::operator()(paddle::framework::ir::PDNode *x,
Y
Yan Chunwei 已提交
824 825 826
                                 bool with_bias) {
  // Create shared nodes.
  x->assert_is_op_input("mul", "X");
C
chengduo 已提交
827
  auto *mul = pattern->NewNode(mul_repr())->assert_is_op("mul");
Y
Yan Chunwei 已提交
828

C
chengduo 已提交
829
  auto *mul_w_var = pattern->NewNode(w_repr())
Y
Yan Chunwei 已提交
830 831 832 833
                        ->AsInput()
                        ->assert_is_persistable_var()
                        ->assert_is_op_input("mul", "Y");

C
chengduo 已提交
834
  auto *mul_out_var =
Y
Yan Chunwei 已提交
835 836 837 838 839 840 841 842 843 844
      pattern->NewNode(mul_out_repr())->assert_is_op_output("mul");

  if (!with_bias) {  // not with bias
    // Add links.
    mul->LinksFrom({x, mul_w_var}).LinksTo({mul_out_var});
    return mul_out_var;

  } else {  // with bias
    mul_out_var->AsIntermediate()->assert_is_op_input("elementwise_add");
    // Create operators.
C
chengduo 已提交
845
    auto *elementwise_add = pattern->NewNode(elementwise_add_repr())
Y
Yan Chunwei 已提交
846 847
                                ->assert_is_op("elementwise_add");
    // Create variables.
C
chengduo 已提交
848
    auto *bias = pattern->NewNode(bias_repr())
Y
Yan Chunwei 已提交
849 850 851
                     ->assert_is_op_input("elementwise_add")
                     ->AsInput();

C
chengduo 已提交
852
    auto *fc_out = pattern->NewNode(Out_repr())
Y
Yan Chunwei 已提交
853 854 855 856 857 858
                       ->AsOutput()
                       ->assert_is_op_output("elementwise_add");

    mul->LinksFrom({mul_w_var, x}).LinksTo({mul_out_var});
    elementwise_add->LinksFrom({mul_out_var, bias}).LinksTo({fc_out});
    return fc_out;
859 860
  }
}
T
tensor-tang 已提交
861

862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879
PDNode *patterns::Embedding::operator()(PDNode *x) {
  x->assert_is_op_input("lookup_table", "Ids");
  auto *lookup_table_op =
      pattern->NewNode(lookup_table_repr())->assert_is_op("lookup_table");
#define NEW_NODE(arg__, io__)                    \
  auto *arg__ = pattern->NewNode(arg__##_repr()) \
                    ->assert_is_op_##io__("lookup_table", #arg__);

  NEW_NODE(W, input);

  NEW_NODE(Out, output);
#undef NEW_NODE

  lookup_table_op->LinksFrom({x, W});
  lookup_table_op->LinksTo({Out});
  return Out;
}

C
chengduo 已提交
880
PDNode *patterns::LSTM::operator()(PDNode *x) {
881
  x->assert_is_op_input("lstm", "Input");
C
chengduo 已提交
882
  auto *lstm_op = pattern->NewNode(lstm_repr())->assert_is_op("lstm");
Y
Yan Chunwei 已提交
883
#define NEW_NODE(arg__, io__) \
C
chengduo 已提交
884
  auto *arg__ =               \
Y
Yan Chunwei 已提交
885
      pattern->NewNode(arg__##_repr())->assert_is_op_##io__("lstm", #arg__);
886 887 888 889 890

  // Currently, the H0 and C0 are optional
  // TODO(Superjomn) upgrade the fuse framework to support optional.
  // NEW_NODE(H0, input);
  // NEW_NODE(C0, input);
Y
Yan Chunwei 已提交
891 892
  NEW_NODE(Weight, input);
  NEW_NODE(Bias, input);
893

Y
Yan Chunwei 已提交
894 895 896 897 898
  NEW_NODE(Hidden, output);
  NEW_NODE(Cell, output);
  NEW_NODE(BatchGate, output);
  NEW_NODE(BatchCellPreAct, output);
#undef NEW_NODE
899 900 901 902 903

  lstm_op->LinksFrom({x, Weight, Bias});
  lstm_op->LinksTo({Hidden, Cell, BatchGate, BatchCellPreAct});
  return Hidden;
}
T
tensor-tang 已提交
904

C
chengduo 已提交
905
PDNode *patterns::GRU::operator()(PDNode *x) {
T
tensor-tang 已提交
906
  x->assert_is_op_input("gru", "Input");
C
chengduo 已提交
907
  auto *gru_op = pattern->NewNode(gru_repr())->assert_is_op("gru");
Y
Yan Chunwei 已提交
908
#define NEW_NODE(arg__, io__) \
C
chengduo 已提交
909
  auto *arg__ =               \
Y
Yan Chunwei 已提交
910
      pattern->NewNode(arg__##_repr())->assert_is_op_##io__("gru", #arg__);
T
tensor-tang 已提交
911

Y
Yan Chunwei 已提交
912
  NEW_NODE(Weight, input);
T
tensor-tang 已提交
913 914
  // TODO(Superjomn): upgrade the fuse framework to support optional.
  // H0 and bias are optional
Y
Yan Chunwei 已提交
915
  NEW_NODE(Bias, input);  // also optional
T
tensor-tang 已提交
916 917
  // NEW_NODE(H0, input);

Y
Yan Chunwei 已提交
918
  NEW_NODE(Hidden, output);
T
tensor-tang 已提交
919
  // below are intermediate
Y
Yan Chunwei 已提交
920 921 922 923
  NEW_NODE(BatchGate, output);
  NEW_NODE(BatchResetHiddenPrev, output);
  NEW_NODE(BatchHidden, output);
#undef NEW_NODE
T
tensor-tang 已提交
924

T
tensor-tang 已提交
925 926 927 928
  BatchGate->AsIntermediate();
  BatchResetHiddenPrev->AsIntermediate();
  BatchHidden->AsIntermediate();

T
tensor-tang 已提交
929 930 931 932 933
  gru_op->LinksFrom({x, Weight, Bias});
  gru_op->LinksTo({Hidden, BatchGate, BatchResetHiddenPrev, BatchHidden});
  return Hidden;
}

C
chengduo 已提交
934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
PDNode *patterns::ActElewiseAdd::operator()(
    paddle::framework::ir::PDNode *in_var,
    std::unordered_set<std::string> act_types) {
  in_var->assert_is_ops_input(act_types, "X");

  auto *act = pattern->NewNode(act_repr())->assert_is_ops(act_types);
  auto *act_out_var = pattern->NewNode(act_out_repr())
                          ->assert_is_not_ctrl_var()
                          ->assert_is_ops_output(act_types);
  act_out_var->AsIntermediate()->assert_is_op_input("elementwise_add");

  auto *ele_x_var = pattern->NewNode(ele_x_repr())
                        ->assert_is_not_ctrl_var()
                        ->assert_is_op_input("elementwise_add")
                        ->AsInput();
  auto *elementwise_add =
      pattern->NewNode(ele_add_repr())->assert_is_op("elementwise_add");

  auto *elewise_add_out = pattern->NewNode(elewise_add_out_repr())
                              ->AsOutput()
                              ->assert_is_op_output("elementwise_add", "Out");

  act->LinksFrom({in_var}).LinksTo({act_out_var});
  elementwise_add->LinksFrom({act_out_var, ele_x_var})
      .LinksTo({elewise_add_out});

  return elewise_add_out;
}

PDNode *patterns::ElewiseAddAct::operator()(
    paddle::framework::ir::PDNode *ele_x_var,
    std::unordered_set<std::string> act_types) {
  auto *ele_y_var = pattern->NewNode(ele_y_repr())
                        ->assert_is_op_input("elementwise_add", "Y");

  auto *ele_add =
      pattern->NewNode(ele_add_repr())->assert_is_op("elementwise_add");

  auto *ele_out_var = pattern->NewNode(elewise_add_out_repr())
                          ->assert_is_op_output("elementwise_add", "Out");

  ele_out_var->AsIntermediate()->assert_is_ops_input(act_types);

  auto *act = pattern->NewNode(act_repr())->assert_is_ops(act_types);

  auto *act_out_var =
      pattern->NewNode(act_out_repr())->assert_is_ops_output(act_types, "Out");

  ele_add->LinksFrom({ele_x_var, ele_y_var}).LinksTo({ele_out_var});
  act->LinksFrom({ele_out_var}).LinksTo({act_out_var});

  return act_out_var;
}

PDNode *patterns::ElewiseAddActInplaceGrad::operator()(
    paddle::framework::ir::PDNode *d_act_out_var,
    std::unordered_set<std::string> act_types) {
  // act_grad: in["Out", "Out@GRAD"], out["X@GRAD"]
  // ele_add_grad: in["Y", "Out@GRAD"], out["X@GRAD", "Y@GRAD"]
  auto *act_grad = pattern->NewNode(act_grad_repr())->assert_is_ops(act_types);

  auto *act_out_var =
      pattern->NewNode(act_out_repr())->assert_is_ops_input(act_types, "Out");

  auto *d_intermediate_var =
      pattern->NewNode(d_itermediate_out_repr())
          ->assert_is_ops_output(act_types, GradVarName("X"));

  act_grad->LinksFrom({d_act_out_var, act_out_var})
      .LinksTo({d_intermediate_var});

  auto *ele_y_var = pattern->NewNode(ele_y_repr())
                        ->assert_is_not_ctrl_var()
                        ->assert_is_op_input("elementwise_add_grad", "Y");

  auto *ele_add_grad = pattern->NewNode(ele_add_grad_repr())
                           ->assert_is_op("elementwise_add_grad");

  auto *d_ele_x_var =
      pattern->NewNode(d_ele_x_repr())
          ->assert_is_not_ctrl_var()
          ->assert_is_op_output("elementwise_add_grad", GradVarName("X"));

  auto *d_ele_y_var =
      pattern->NewNode(d_ele_y_repr())
          ->assert_is_not_ctrl_var()
          ->assert_is_op_output("elementwise_add_grad", GradVarName("Y"));

  ele_add_grad->LinksFrom({d_intermediate_var, ele_y_var})
      .LinksTo({d_ele_x_var, d_ele_y_var});

  return ele_add_grad;
}

M
Michal Gallus 已提交
1028
PDNode *patterns::ConvBias::operator()(
1029
    paddle::framework::ir::PDNode *conv_input, bool is_conv3d) {
Y
Yihua Xu 已提交
1030
  std::string type = is_conv3d ? "conv3d" : "conv2d";
M
Michal Gallus 已提交
1031
  // Create Operators
Y
Yihua Xu 已提交
1032 1033
  conv_input->assert_is_op_input(type, "Input");
  auto *conv_op = pattern->NewNode(conv_repr())->assert_is_op(type);
M
Michal Gallus 已提交
1034 1035 1036 1037
  auto *eltiwse_op =
      pattern->NewNode(eltwise_repr())->assert_is_op("elementwise_add");
  // Create variables
  // Filter
Y
Yihua Xu 已提交
1038 1039 1040 1041
  auto *conv_weight_var = pattern->NewNode(conv_weight_repr())
                              ->AsInput()
                              ->assert_is_persistable_var()
                              ->assert_is_op_input(type, "Filter");
M
Michal Gallus 已提交
1042
  // intermediate variable, will be removed in the IR after fuse.
Y
Yihua Xu 已提交
1043 1044 1045 1046
  auto *conv_out_var = pattern->NewNode(conv_out_repr())
                           ->AsIntermediate()
                           ->assert_is_only_output_of_op(type)
                           ->assert_is_op_input("elementwise_add");
M
Michal Gallus 已提交
1047 1048 1049
  // Bias stored in elementwise_add
  auto *eltwise_bias_var = pattern->NewNode(eltwise_bias_repr())
                               ->AsInput()
M
Michal Gallus 已提交
1050
                               ->assert_is_persistable_var()
M
Michal Gallus 已提交
1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061
                               ->assert_is_op_input("elementwise_add", "Y");
  // output
  auto *eltwise_out_var = pattern->NewNode(eltwise_out_repr())
                              ->AsOutput()
                              ->assert_is_op_output("elementwise_add");
  conv_op->LinksFrom({conv_input, conv_weight_var}).LinksTo({conv_out_var});
  eltiwse_op->LinksFrom({conv_out_var, eltwise_bias_var})
      .LinksTo({eltwise_out_var});
  return eltwise_out_var;
}

1062 1063 1064 1065
PDNode *patterns::Conv::operator()() {
  auto conv_op = pattern->NewNode(conv_op_repr())->assert_is_op("conv2d");

  auto input_var = pattern->NewNode(conv_input_repr())
1066
                       ->AsInput()
1067 1068 1069
                       ->assert_is_op_input("conv2d", "Input");

  auto filter_var = pattern->NewNode(conv_filter_repr())
1070
                        ->AsInput()
1071 1072 1073
                        ->assert_is_op_input("conv2d", "Filter");

  auto output_var = pattern->NewNode(conv_output_repr())
1074
                        ->AsOutput()
1075 1076
                        ->assert_is_op_output("conv2d", "Output");

1077
  conv_op->LinksFrom({input_var, filter_var});
1078 1079 1080 1081 1082
  conv_op->LinksTo({output_var});

  return output_var;
}

1083
PDNode *patterns::ElementwiseAdd::operator()(PDNode *x_var, PDNode *y_var) {
1084 1085 1086
  auto elementwise_add_op = pattern->NewNode(elementwise_add_op_repr())
                                ->assert_is_op("elementwise_add");

1087 1088
  x_var->AsInput()->assert_is_op_input("elementwise_add", "X");
  y_var->AsInput()->assert_is_op_input("elementwise_add", "Y");
1089 1090 1091 1092
  auto out_var = pattern->NewNode(elementwise_add_out_repr())
                     ->AsOutput()
                     ->assert_is_op_output("elementwise_add", "Out");

1093
  elementwise_add_op->LinksFrom({x_var, y_var});
1094 1095 1096 1097
  elementwise_add_op->LinksTo({out_var});

  return out_var;
}
1098

H
hjchen2 已提交
1099
std::unordered_set<std::string> conv_act_set({"identity", "relu"});
1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164

PDNode *patterns::ConvElementwiseaddAct::operator()(PDNode *conv_in) {
  conv_in->AsInput();
  auto conv_op = pattern->NewNode(conv_op_repr())->assert_is_op("conv2d");
  auto conv_out = pattern->NewNode(conv_out_repr())
                      ->assert_is_op_output("conv2d")
                      ->assert_is_op_input("elementwise_add", "X")
                      ->AsIntermediate();
  auto conv_filter = pattern->NewNode(conv_filter_repr())
                         ->assert_is_op_input("conv2d", "Filter")
                         ->AsInput();
  auto elementwise_add_op = pattern->NewNode(elementwise_add_op_repr())
                                ->assert_is_op("elementwise_add");
  auto elementwise_add_in_y = pattern->NewNode(elementwise_add_in_y_repr())
                                  ->assert_is_op_input("elementwise_add", "Y")
                                  ->AsInput();
  auto elementwise_add_out = pattern->NewNode(elementwise_add_out_repr())
                                 ->assert_is_op_output("elementwise_add")
                                 ->AsIntermediate();

  auto act_op = pattern->NewNode(act_op_repr())
                    ->assert_is_op()
                    ->assert_more([&](Node *node) {
                      auto op_type = node->Name();
                      return conv_act_set.count(op_type);
                    });

  auto act_out = pattern->NewNode(act_out_repr())
                     ->assert_is_var()
                     // is activation op's output.
                     ->assert_more([&](Node *node) {
                       for (auto *in_op : node->inputs) {
                         if (conv_act_set.count(in_op->Name())) {
                           return true;
                         }
                       }
                       return false;
                     })
                     ->AsOutput();

  conv_op->LinksFrom({conv_in, conv_filter});
  conv_out->LinksFrom({conv_op});
  elementwise_add_op->LinksFrom({conv_out, elementwise_add_in_y})
      .LinksTo({elementwise_add_out});
  act_op->LinksFrom({elementwise_add_out}).LinksTo({act_out});

  return act_out;
}

PDNode *patterns::ConvElementwiseadd2Act::operator()(PDNode *conv_in) {
  auto conv_op = pattern->NewNode(conv_op_repr())->assert_is_op("conv2d");
  auto conv_filter = pattern->NewNode(conv_filter_repr())
                         ->assert_is_op_input("conv2d", "Filter")
                         ->AsInput();
  auto conv_out = pattern->NewNode(conv_out_repr())
                      ->assert_is_op_output("conv2d")
                      ->assert_is_op_input("elementwise_add", "X")
                      ->AsIntermediate();
  auto elementwise_add_op = pattern->NewNode(elementwise_add_op_repr())
                                ->assert_is_op("elementwise_add");
  auto elementwise_add_in_y = pattern->NewNode(elementwise_add_in_y_repr())
                                  ->assert_is_op_input("elementwise_add", "Y")
                                  ->AsInput();
  auto elementwise_add_out = pattern->NewNode(elementwise_add_out_repr())
                                 ->assert_is_op_output("elementwise_add")
H
hjchen2 已提交
1165
                                 ->assert_is_op_input("elementwise_add", "Y")
1166 1167 1168 1169 1170
                                 ->AsIntermediate();

  auto elementwise_add_op_1 = pattern->NewNode(elementwise_add_op_1_repr())
                                  ->assert_is_op("elementwise_add");
  auto elementwise_add_in_y_1 = pattern->NewNode(elementwise_add_in_y_1_repr())
H
hjchen2 已提交
1171
                                    ->assert_is_op_input("elementwise_add", "X")
1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
                                    ->AsInput();
  auto elementwise_add_out_1 = pattern->NewNode(elementwise_add_out_1_repr())
                                   ->assert_is_op_output("elementwise_add")
                                   ->AsIntermediate();

  auto act_op = pattern->NewNode(act_op_repr())
                    ->assert_is_op()
                    ->assert_more([&](Node *node) {
                      auto op_type = node->Name();
                      return conv_act_set.count(op_type);
                    });
  auto act_out = pattern->NewNode(act_out_repr())
                     ->assert_is_var()
                     // is activation op's output.
                     ->assert_more([&](Node *node) {
                       for (auto *in_op : node->inputs) {
                         if (conv_act_set.count(in_op->Name())) {
                           return true;
                         }
                       }
                       return false;
                     })
                     ->AsOutput();

  conv_op->LinksFrom({conv_in, conv_filter}).LinksTo({conv_out});
  elementwise_add_op->LinksFrom({conv_out, elementwise_add_in_y})
      .LinksTo({elementwise_add_out});
H
hjchen2 已提交
1199 1200
  elementwise_add_op_1->LinksFrom({elementwise_add_out, elementwise_add_in_y_1})
      .LinksTo({elementwise_add_out_1});
1201 1202 1203 1204
  act_op->LinksFrom({elementwise_add_out_1}).LinksTo({act_out});
  return act_out;
}

N
nhzlx 已提交
1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231
PDNode *patterns::ConvElementwiseadd::operator()(PDNode *conv_in) {
  conv_in->AsInput();
  auto conv_op = pattern->NewNode(conv_op_repr())->assert_is_op("conv2d");
  auto conv_out = pattern->NewNode(conv_out_repr())
                      ->assert_is_op_output("conv2d")
                      ->assert_is_op_input("elementwise_add", "X")
                      ->AsIntermediate();
  auto conv_filter = pattern->NewNode(conv_filter_repr())
                         ->assert_is_op_input("conv2d", "Filter")
                         ->AsInput();
  auto elementwise_add_op = pattern->NewNode(elementwise_add_op_repr())
                                ->assert_is_op("elementwise_add");
  auto elementwise_add_in_y = pattern->NewNode(elementwise_add_in_y_repr())
                                  ->assert_is_op_input("elementwise_add", "Y")
                                  ->AsInput();
  auto elementwise_add_out = pattern->NewNode(elementwise_add_out_repr())
                                 ->assert_is_op_output("elementwise_add")
                                 ->AsOutput();

  conv_op->LinksFrom({conv_in, conv_filter});
  conv_out->LinksFrom({conv_op});
  elementwise_add_op->LinksFrom({conv_out, elementwise_add_in_y})
      .LinksTo({elementwise_add_out});

  return elementwise_add_out;
}

N
nhzlx 已提交
1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
PDNode *patterns::ConvAffineChannel::operator()(
    paddle::framework::ir::PDNode *conv_input, bool with_eltwise_add) {
  // Create Operators
  conv_input->assert_is_op_input("conv2d", "Input");
  auto *conv_op = pattern->NewNode(conv_repr())->assert_is_op("conv2d");

  PDNode *eltwise_op = nullptr;
  if (with_eltwise_add) {
    eltwise_op =
        pattern->NewNode(eltwise_repr())->assert_is_op("elementwise_add");
  }

  auto *affine_channel_op =
      pattern->NewNode(affine_channel_repr())->assert_is_op("affine_channel");
  // Create variables
  // Conv Filter
  auto *conv_weight_var = pattern->NewNode(conv_weight_repr())
                              ->AsInput()
                              ->assert_is_persistable_var()
                              ->assert_is_op_input("conv2d", "Filter");

  auto *conv_out_var = pattern->NewNode(conv_out_repr())
                           ->AsIntermediate()
                           ->assert_is_only_output_of_op("conv2d");

  PDNode *eltwise_y_in_var = nullptr;
  PDNode *eltwise_out_var = nullptr;
  if (with_eltwise_add) {
    // Conv output as Bias input
    conv_out_var->assert_is_op_input("elementwise_add", "X");
    // Bias
    eltwise_y_in_var = pattern->NewNode(eltwise_y_in_repr())
                           ->assert_is_op_input("elementwise_add", "Y")
                           ->AsInput();
    eltwise_out_var = pattern->NewNode(eltwise_out_repr())
                          ->AsIntermediate()
                          ->assert_is_only_output_of_op("elementwise_add");
  } else {
    // Conv output as AffineChannel input
    conv_out_var->assert_is_op_input("affine_channel", "X");
  }

  // AC Scale
  auto *ac_scale_var = pattern->NewNode(ac_scale_repr())
                           ->AsInput()
                           ->assert_is_persistable_var()
                           ->assert_is_op_input("affine_channel", "Scale");
  // AC Bias
  auto *ac_bias_var = pattern->NewNode(ac_bias_repr())
                          ->AsInput()
                          ->assert_is_persistable_var()
                          ->assert_is_op_input("affine_channel", "Bias");

  // AC output
  auto *ac_out_var = pattern->NewNode(ac_out_repr())
                         ->AsOutput()
                         ->assert_is_op_output("affine_channel");

  conv_op->LinksFrom({conv_input, conv_weight_var}).LinksTo({conv_out_var});

  if (with_eltwise_add) {
    eltwise_op->LinksFrom({conv_out_var, eltwise_y_in_var})
        .LinksTo({eltwise_out_var});
    affine_channel_op->LinksFrom({eltwise_out_var, ac_scale_var, ac_bias_var})
        .LinksTo({ac_out_var});
  } else {
    affine_channel_op->LinksFrom({conv_out_var, ac_scale_var, ac_bias_var})
        .LinksTo({ac_out_var});
  }
  return ac_out_var;
}

1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366
// a -> transpose_op(1) -> transpose_out_a -> flatten_op(1) -> flatten_out_a
// b -> transpose_op(2) -> transpose_out_b -> flatten_op(2) -> flatten_out_b
// ...
// z -> transpose_op(n) -> transpose_out_z -> flatten_op(n) -> flatten_out_z
// flatten_out_a -> concat_op  flatten_out_b -> concat_op ... flatten_out_z ->
// concat_op
PDNode *patterns::TransposeFlattenConcat::operator()(
    std::vector<PDNode *> conv_in, int times) {
  // The times represents the repeat times of the
  // {trans, trans_out, flatten, flatten_out}
  const int kNumFields = 4;
  const int kTransOutOffset = 1;
  const int kFlattenOffset = 2;
  const int kFlattenOutOffset = 3;

  std::vector<PDNode *> nodes;

  for (int i = 0; i < times; i++) {
    nodes.push_back(
        pattern->NewNode(GetNodeName("transpose" + std::to_string(i)))
            ->assert_is_op("transpose2"));
    nodes.push_back(
        pattern->NewNode(GetNodeName("transpose_out" + std::to_string(i)))
            ->assert_is_op_output("transpose2")
            ->assert_is_op_input("flatten2", "X")
            ->AsIntermediate());
    nodes.push_back(pattern->NewNode(GetNodeName("flatten" + std::to_string(i)))
                        ->assert_is_op("flatten2"));

    nodes.push_back(
        pattern->NewNode(GetNodeName("flatten_out" + std::to_string(i)))
            ->assert_is_op_output("flatten2")
            ->assert_is_op_nth_input("concat", "X", i)
            ->AsIntermediate());
  }

  auto concat_op = pattern->NewNode(GetNodeName("concat"))
                       ->assert_is_op("concat")
                       ->assert_op_has_n_inputs("concat", times);
  auto concat_out = pattern->NewNode(GetNodeName("concat_out"))
                        ->assert_is_op_output("concat")
                        ->AsOutput();

  std::vector<PDNode *> flatten_outs;
  for (int i = 0; i < times; i++) {
    conv_in[i]->AsInput();
    // trans
    nodes[i * kNumFields]->LinksFrom({conv_in[i]});
    // trans_out
    nodes[i * kNumFields + kTransOutOffset]->LinksFrom({nodes[i * kNumFields]});
    // flatten
    nodes[i * kNumFields + kFlattenOffset]->LinksFrom(
        {nodes[i * kNumFields + kTransOutOffset]});
    // flatten_out
    nodes[i * kNumFields + kFlattenOutOffset]->LinksFrom(
        {nodes[i * kNumFields + kFlattenOffset]});
    flatten_outs.push_back(nodes[i * kNumFields + kFlattenOutOffset]);
  }

  concat_op->LinksFrom(flatten_outs).LinksTo({concat_out});
  return concat_out;
}

1367 1368 1369
}  // namespace ir
}  // namespace framework
}  // namespace paddle