communicator.h

/* Copyright (c) 2019 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 <ThreadPool.h>
#include <stdint.h>
#include <atomic>
#include <deque>
#include <map>
#include <memory>
#include <numeric>
#include <set>
#include <string>
#include <unordered_map>
#include <unordered_set>
#include <utility>
#include <vector>

#include "gflags/gflags.h"
#include "paddle/fluid/framework/scope.h"
#include "paddle/fluid/framework/variable.h"
#include "paddle/fluid/framework/variable_helper.h"
#include "paddle/fluid/operators/distributed/communicator_common.h"
#include "paddle/fluid/operators/distributed/distributed.h"
#include "paddle/fluid/operators/distributed/large_scale_kv.h"
#include "paddle/fluid/operators/distributed/rpc_client.h"
#include "paddle/fluid/operators/distributed_ops/send_recv_util.h"
#include "paddle/fluid/operators/math/blas.h"
#include "paddle/fluid/operators/math/math_function.h"
#include "paddle/fluid/operators/math/selected_rows_functor.h"
#include "paddle/fluid/platform/device_context.h"
#include "paddle/fluid/platform/enforce.h"
#include "paddle/fluid/platform/place.h"
#include "paddle/fluid/string/split.h"

DECLARE_bool(communicator_is_sgd_optimizer);

namespace paddle {
namespace operators {
namespace distributed {

using Scope = framework::Scope;
using Variable = framework::Variable;

template <typename T>
class BlockingQueue {
 public:
  explicit BlockingQueue(size_t capacity) : capacity_(capacity) {
    PADDLE_ENFORCE_GT(capacity_, 0,
                      platform::errors::InvalidArgument(
                          "The capacity must be greater than 0."));
  }

  bool Push(const T &elem) {
    {
      std::unique_lock<std::mutex> lock(mutex_);
      cv_.wait(lock, [&] { return queue_.size() < capacity_; });
      PADDLE_ENFORCE_LT(
          queue_.size(), capacity_,
          platform::errors::OutOfRange("The queue size: %s out of capacity:%s",
                                       queue_.size(), capacity_));
      queue_.push_back(elem);
    }
    cv_.notify_one();
    return true;
  }

  bool Push(T &&elem) {
    {
      std::unique_lock<std::mutex> lock(mutex_);
      cv_.wait(lock, [&] { return queue_.size() < capacity_; });
      PADDLE_ENFORCE_LT(
          queue_.size(), capacity_,
          platform::errors::OutOfRange("The queue size: %s out of capacity:%s",
                                       queue_.size(), capacity_));
      queue_.emplace_back(std::move(elem));
    }
    cv_.notify_one();
    return true;
  }

  T Pop() {
    std::unique_lock<std::mutex> lock(mutex_);
    cv_.wait(lock, [=] { return !queue_.empty(); });
    T rc(std::move(queue_.front()));
    queue_.pop_front();
    cv_.notify_one();
    return rc;
  }

  size_t Cap() const {
    std::lock_guard<std::mutex> lock(mutex_);
    return capacity_;
  }

  size_t Size() const {
    std::lock_guard<std::mutex> lock(mutex_);
    return queue_.size();
  }

 private:
  const size_t capacity_;
  std::deque<T> queue_;

  mutable std::mutex mutex_;
  std::condition_variable cv_;
};

template <typename T, int MajorType = Eigen::RowMajor,
          typename IndexType = Eigen::DenseIndex>
using EigenVector = framework::EigenVector<T, MajorType, IndexType>;

template <typename T>
inline void MergeVars(const std::string &var_name,
                      const std::vector<std::shared_ptr<Variable>> &vars,
                      Scope *scope, bool merge_add = true) {
  PADDLE_ENFORCE_NE(vars.empty(), true, platform::errors::InvalidArgument(
                                            "vector vars are empty."));
  auto cpu_place = platform::CPUPlace();
  auto &var0 = vars[0];
  auto *out_var = scope->Var(var_name);
  if (var0->IsType<framework::LoDTensor>()) {
    auto dims = var0->Get<framework::LoDTensor>().dims();
    VLOG(3) << "merge " << var_name << " LoDTensor dims " << dims
            << "; merge add: " << merge_add;
    // init output tensor
    auto *out_t = out_var->GetMutable<framework::LoDTensor>();
    out_t->mutable_data<T>(dims, cpu_place);
    // check the input dims
    for (auto &var : vars) {
      auto &var_t = var->Get<framework::LoDTensor>();
      PADDLE_ENFORCE_EQ(
          var_t.dims(), dims,
          platform::errors::InvalidArgument("vars should have the same dims"));
    }

    // set output tensor to 0.
    auto cpu_ctx = paddle::platform::CPUDeviceContext();
    math::SetConstant<paddle::platform::CPUDeviceContext, T> constant_functor;
    constant_functor(cpu_ctx, out_t, static_cast<T>(0));
    // sum all vars to out
    auto result = EigenVector<T>::Flatten(*out_t);
    for (auto &var : vars) {
      auto &in_t = var->Get<framework::LoDTensor>();
      auto in = EigenVector<T>::Flatten(in_t);
      result.device(*cpu_ctx.eigen_device()) = result + in;
    }
    if (!merge_add) {
      result.device(*cpu_ctx.eigen_device()) =
          result / static_cast<T>(vars.size());
    }
  } else if (var0->IsType<framework::SelectedRows>()) {
    auto &slr0 = var0->Get<framework::SelectedRows>();
    auto *out_slr = out_var->GetMutable<framework::SelectedRows>();
    out_slr->mutable_rows()->clear();
    out_slr->mutable_value()->mutable_data<T>({{}}, cpu_place);
    std::vector<const paddle::framework::SelectedRows *> inputs;
    inputs.reserve(vars.size());
    for (auto &var : vars) {
      inputs.push_back(&var->Get<framework::SelectedRows>());
    }
    auto dev_ctx = paddle::platform::CPUDeviceContext();
    if (merge_add) {
      math::scatter::MergeAdd<paddle::platform::CPUDeviceContext, T> merge_add;
      merge_add(dev_ctx, inputs, out_slr);
    } else {
      math::scatter::MergeAverage<paddle::platform::CPUDeviceContext, T>
          merge_average;
      merge_average(dev_ctx, inputs, out_slr);
    }

    VLOG(3) << "merge " << var_name << " SelectedRows height: " << slr0.height()
            << " dims: " << slr0.value().dims() << "; merge add: " << merge_add;
  } else {
    PADDLE_THROW(platform::errors::InvalidArgument("unsupported var type: %s!",
                                                   var0->Type()));
  }
}

using RpcCtxMap = std::unordered_map<std::string, CommContext>;
using SparseValue = std::unordered_map<int64_t, std::vector<float>>;

class Communicator {
 public:
  Communicator();

  explicit Communicator(const std::map<std::string, std::string> &envs_) {
    for (auto &iter : envs_) {
      envs[iter.first] = iter.second;
    }
  }

  virtual ~Communicator() {}

  virtual void Start() = 0;

  virtual void Stop() = 0;

  virtual bool IsRunning() { return running_; }

  virtual void Clean() {}

  virtual void Send(const std::vector<std::string> &var_names,
                    const std::vector<std::string> &var_tables,
                    const framework::Scope &scope) = 0;

  virtual void RecvNoBarrier() {}

  virtual void Barrier() {}

  virtual void BarrierTriggerDecrement() {}

  virtual void BarrierTriggerReset(int init_counter) {}

  virtual void InitEnvs() = 0;

  virtual void InitImpl(const RpcCtxMap &send_varname_to_ctx,
                        const RpcCtxMap &recv_varname_to_ctx,
                        Scope *recv_scope) {}

  static Communicator *GetInstance() { return communicator_.get(); }

  static std::shared_ptr<Communicator> GetInstantcePtr() {
    return communicator_;
  }

  template <typename T>
  static Communicator *InitInstance(
      const RpcCtxMap &send_ctx, const RpcCtxMap &recv_ctx, Scope *recv_scope,
      const std::map<std::string, std::string> &envs) {
    std::call_once(init_flag_, &Communicator::InitWithRpcCtx<T>, send_ctx,
                   recv_ctx, recv_scope, std::ref(envs));
    return communicator_.get();
  }

  // Init is called by InitInstance.
  template <typename T>
  static void InitWithRpcCtx(const RpcCtxMap &send_ctx,
                             const RpcCtxMap &recv_ctx, Scope *recv_scope,
                             const std::map<std::string, std::string> &envs) {
    if (communicator_.get() == nullptr) {
      communicator_.reset(new T(std::ref(envs)));
      communicator_->InitEnvs();
      communicator_->InitImpl(send_ctx, recv_ctx, recv_scope);
    }
  }

 protected:
  bool running_ = false;
  bool waiting_ = true;
  static std::shared_ptr<Communicator> communicator_;
  static std::once_flag init_flag_;
  std::unordered_map<std::string, std::string> envs;
};

class AsyncCommunicator : public Communicator {
 public:
  AsyncCommunicator() : Communicator() {}

  explicit AsyncCommunicator(const std::map<std::string, std::string> &envs)
      : Communicator(envs) {}

  ~AsyncCommunicator();

  void InitEnvs() {
    min_send_grad_num_before_recv_ =
        std::stoi(envs.at("communicator_min_send_grad_num_before_recv"));
    thread_pool_size_ = std::stoi(envs.at("communicator_thread_pool_size"));
    max_merge_var_num_ = std::stoi(envs.at("communicator_max_merge_var_num"));
    send_wait_times_ = std::stoi(envs.at("communicator_send_wait_times"));
    send_queue_size_ = std::stoi(envs.at("communicator_send_queue_size"));
    need_global_step_ =
        static_cast<bool>(std::stoi(envs.at("need_global_step")));
    VLOG(0) << "AsyncCommunicator Initialized";
  }

  void Start() override;

  void Stop() override;

  void InitImpl(const RpcCtxMap &send_varname_to_ctx,
                const RpcCtxMap &recv_varname_to_ctx,
                Scope *recv_scope) override;

  void InitParams();

  virtual void MainThread();

  void Send(const std::vector<std::string> &var_names,
            const std::vector<std::string> &var_tables,
            const framework::Scope &scope) override;

  virtual void SendByCommunicator(int batches);

  virtual void SendGlobalStep(int batches);

  virtual void RecvByCommunicator();

  virtual void RecvNoBarrier();

  virtual int BatchesCounter();

  virtual void BarrierSend() {}

  virtual void BarrierRecv() {}

  virtual void BarrierWeakUp() {}

 protected:
  int min_send_grad_num_before_recv_;
  int thread_pool_size_;
  int max_merge_var_num_;
  int send_wait_times_;
  int send_queue_size_;
  int trainer_id_ = 0;
  bool need_global_step_ = false;

  std::unordered_map<std::string,
                     std::shared_ptr<BlockingQueue<std::shared_ptr<Variable>>>>
      send_varname_to_queue_;
  RpcCtxMap send_varname_to_ctx_;
  RpcCtxMap recv_varname_to_ctx_;
  std::unique_ptr<std::thread> main_thread_{nullptr};
  Scope *recv_scope_;                  // should be global scope
  std::unique_ptr<Scope> send_scope_;  // an independent scope
  std::unique_ptr<::ThreadPool> send_threadpool_{nullptr};
  std::unique_ptr<::ThreadPool> recv_threadpool_{nullptr};
  std::atomic_uint grad_num_{0};  // the num of gradient sent since last recv
};

class HalfAsyncCommunicator : public AsyncCommunicator {
 public:
  HalfAsyncCommunicator() {}

  explicit HalfAsyncCommunicator(const std::map<std::string, std::string> &envs)
      : AsyncCommunicator(envs) {}

  void InitEnvs() {
    min_send_grad_num_before_recv_ = 0;

    max_merge_var_num_ = std::stoi(envs.at("communicator_max_merge_var_num"));
    send_wait_times_ = std::stoi(envs.at("communicator_send_wait_times"));
    thread_pool_size_ = std::stoi(envs.at("communicator_thread_pool_size"));
    send_queue_size_ = std::stoi(envs.at("communicator_send_queue_size"));
    need_global_step_ =
        static_cast<bool>(std::stoi(envs.at("need_global_step")));
    VLOG(0) << "HalfAsyncCommunicator Initialized";
  }

  void Clean() override;

  void Barrier() override;

  void BarrierTriggerDecrement() override;

  void BarrierTriggerReset(int initial_val) override;

  int BatchesCounter();

  void BarrierWeakUp();

 protected:
  // mutex for Wait for barrier
  std::mutex barrier_mutex_;
  std::condition_variable barrier_cond_;
  std::atomic<int64_t> barrier_trigger_{0};
  std::atomic<int64_t> barrier_counter_{0};
};

class SyncCommunicator : public HalfAsyncCommunicator {
 public:
  SyncCommunicator() : HalfAsyncCommunicator() {}

  explicit SyncCommunicator(const std::map<std::string, std::string> &envs)
      : HalfAsyncCommunicator(envs) {}

  void InitEnvs() {
    min_send_grad_num_before_recv_ = 0;

    max_merge_var_num_ = std::stoi(envs.at("communicator_max_merge_var_num"));
    send_wait_times_ = std::stoi(envs.at("communicator_send_wait_times"));
    thread_pool_size_ = std::stoi(envs.at("communicator_thread_pool_size"));
    send_queue_size_ = std::stoi(envs.at("communicator_send_queue_size"));
    need_global_step_ =
        static_cast<bool>(std::stoi(envs.at("need_global_step")));

    trainer_id_ = std::stoi(envs.at("trainer_id"));
    auto pserver_strings = envs.at("pserver_endpoints");
    pserver_endpoints_ = paddle::string::Split(pserver_strings, ',');
    VLOG(0) << "SyncCommunicator Initialized";
  }

  void BarrierSend();

  void BarrierRecv();

 private:
  std::vector<std::string> pserver_endpoints_{};
};

class GeoCommunicator : public AsyncCommunicator {
 public:
  GeoCommunicator() : AsyncCommunicator() {}

  explicit GeoCommunicator(const std::map<std::string, std::string> &envs)
      : AsyncCommunicator(envs) {}

  void InitImpl(const RpcCtxMap &send_varname_to_ctx,
                const RpcCtxMap &recv_varname_to_ctx,
                Scope *recv_scope) override;
  void MainThread() override;
  void InitEnvs() {
    min_send_grad_num_before_recv_ = 0;

    max_merge_var_num_ = std::stoi(envs.at("communicator_max_merge_var_num"));
    send_wait_times_ = std::stoi(envs.at("communicator_send_wait_times"));
    thread_pool_size_ = std::stoi(envs.at("communicator_thread_pool_size"));

    send_queue_size_ = max_merge_var_num_;
    trainers_ = std::stoi(envs.at("trainers"));
    sparse_attrs_ = envs.at("sparse_attrs");
    VLOG(0) << "GeoCommunicator Initialized";
  }

  void Send(const std::vector<std::string> &var_names,
            const std::vector<std::string> &var_tables,
            const framework::Scope &scope) override;

  void SendByCommunicator(int batches) { return; }

  std::vector<int64_t> MergeSparseIds(const std::string &send_varname);

  void SendSparse(const std::string &varname, int ep_idx,
                  const std::vector<int64_t> &sparse_ids);

  void SendDense(const std::string &varname);

  void SendGlobalStep(int batches) override {}

  void RecvByCommunicator() override;

  void RecvSparse(const std::string &varname, int ep_idx);

  void RecvDense(const std::string &varname);

  void InitParams();

  void InitSparse();

  void InitDense(const std::string varname);

 private:
  int trainers_;
  std::string sparse_attrs_;

  // parameter for delta calc and send
  std::shared_ptr<Scope> delta_scope_;

  // parameter for storage the pserver param after last recv
  std::shared_ptr<Scope> old_scope_;

  // parameter on pserver
  std::shared_ptr<Scope> pserver_scope_;

  int send_var_nums_ = 0;
  std::unordered_map<std::string, std::shared_ptr<SparseValue>> old_sparses_;

  std::unordered_map<
      std::string,
      std::shared_ptr<BlockingQueue<std::shared_ptr<std::vector<int64_t>>>>>
      sparse_id_queues_;
};

}  // namespace distributed
}  // namespace operators
}  // namespace paddle