xgboost/plugin/federated/engine_federated.cc
Rong Ou a2686543a9
Common interface for collective communication (#8057)
* implement broadcast for federated communicator

* implement allreduce

* add communicator factory

* add device adapter

* add device communicator to factory

* add rabit communicator

* add rabit communicator to the factory

* add nccl device communicator

* add synchronize to device communicator

* add back print and getprocessorname

* add python wrapper and c api

* clean up types

* fix non-gpu build

* try to fix ci

* fix std::size_t

* portable string compare ignore case

* c style size_t

* fix lint errors

* cross platform setenv

* fix memory leak

* fix lint errors

* address review feedback

* add python test for rabit communicator

* fix failing gtest

* use json to configure communicators

* fix lint error

* get rid of factories

* fix cpu build

* fix include

* fix python import

* don't export collective.py yet

* skip collective communicator pytest on windows

* add review feedback

* update documentation

* remove mpi communicator type

* fix tests

* shutdown the communicator separately

Co-authored-by: Hyunsu Cho <chohyu01@cs.washington.edu>
2022-09-12 15:21:12 -07:00

198 lines
6.6 KiB
C++

/*!
* Copyright 2022 XGBoost contributors
*/
#include <cstdio>
#include <fstream>
#include <sstream>
#include "federated_client.h"
#include "rabit/internal/engine.h"
#include "rabit/internal/utils.h"
namespace MPI { // NOLINT
// MPI data type to be compatible with existing MPI interface
class Datatype {
public:
size_t type_size;
explicit Datatype(size_t type_size) : type_size(type_size) {}
};
} // namespace MPI
namespace rabit {
namespace engine {
/*! \brief implementation of engine using federated learning */
class FederatedEngine : public IEngine {
public:
void Init(int argc, char *argv[]) {
// Parse environment variables first.
for (auto const &env_var : env_vars_) {
char const *value = getenv(env_var.c_str());
if (value != nullptr) {
SetParam(env_var, value);
}
}
// Command line argument overrides.
for (int i = 0; i < argc; ++i) {
std::string const key_value = argv[i];
auto const delimiter = key_value.find('=');
if (delimiter != std::string::npos) {
SetParam(key_value.substr(0, delimiter), key_value.substr(delimiter + 1));
}
}
utils::Printf("Connecting to federated server %s, world size %d, rank %d",
server_address_.c_str(), world_size_, rank_);
if (server_cert_.empty() || client_key_.empty() || client_cert_.empty()) {
utils::Printf("Certificates not specified, turning off SSL.");
client_.reset(new xgboost::federated::FederatedClient(server_address_, rank_));
} else {
client_.reset(new xgboost::federated::FederatedClient(server_address_, rank_, server_cert_,
client_key_, client_cert_));
}
}
void Finalize() { client_.reset(); }
void Allgather(void *sendrecvbuf, size_t total_size, size_t slice_begin, size_t slice_end,
size_t size_prev_slice) override {
throw std::logic_error("FederatedEngine:: Allgather is not supported");
}
std::string Allgather(void *sendbuf, size_t total_size) {
std::string const send_buffer(reinterpret_cast<char *>(sendbuf), total_size);
return client_->Allgather(send_buffer);
}
void Allreduce(void *sendrecvbuf, size_t type_nbytes, size_t count, ReduceFunction reducer,
PreprocFunction prepare_fun, void *prepare_arg) override {
throw std::logic_error("FederatedEngine:: Allreduce is not supported, use Allreduce_ instead");
}
void Allreduce(void *sendrecvbuf, size_t size, mpi::DataType dtype, mpi::OpType op) {
auto *buffer = reinterpret_cast<char *>(sendrecvbuf);
std::string const send_buffer(buffer, size);
auto const receive_buffer =
client_->Allreduce(send_buffer, static_cast<xgboost::federated::DataType>(dtype),
static_cast<xgboost::federated::ReduceOperation>(op));
receive_buffer.copy(buffer, size);
}
int GetRingPrevRank() const override {
throw std::logic_error("FederatedEngine:: GetRingPrevRank is not supported");
}
void Broadcast(void *sendrecvbuf, size_t size, int root) override {
if (world_size_ == 1) return;
auto *buffer = reinterpret_cast<char *>(sendrecvbuf);
std::string const send_buffer(buffer, size);
auto const receive_buffer = client_->Broadcast(send_buffer, root);
if (rank_ != root) {
receive_buffer.copy(buffer, size);
}
}
int LoadCheckPoint() override { return 0; }
void CheckPoint() override { version_number_ += 1; }
int VersionNumber() const override { return version_number_; }
/*! \brief get rank of current node */
int GetRank() const override { return rank_; }
/*! \brief get total number of */
int GetWorldSize() const override { return world_size_; }
/*! \brief whether it is distributed */
bool IsDistributed() const override { return true; }
/*! \brief get the host name of current node */
std::string GetHost() const override { return "rank" + std::to_string(rank_); }
void TrackerPrint(const std::string &msg) override {
// simply print information into the tracker
utils::Printf("%s", msg.c_str());
}
private:
void SetParam(std::string const &name, std::string const &val) {
if (!strcasecmp(name.c_str(), "FEDERATED_SERVER_ADDRESS")) {
server_address_ = val;
} else if (!strcasecmp(name.c_str(), "FEDERATED_WORLD_SIZE")) {
world_size_ = std::stoi(val);
} else if (!strcasecmp(name.c_str(), "FEDERATED_RANK")) {
rank_ = std::stoi(val);
} else if (!strcasecmp(name.c_str(), "FEDERATED_SERVER_CERT")) {
server_cert_ = ReadFile(val);
} else if (!strcasecmp(name.c_str(), "FEDERATED_CLIENT_KEY")) {
client_key_ = ReadFile(val);
} else if (!strcasecmp(name.c_str(), "FEDERATED_CLIENT_CERT")) {
client_cert_ = ReadFile(val);
}
}
static std::string ReadFile(std::string const &path) {
auto stream = std::ifstream(path.data());
std::ostringstream out;
out << stream.rdbuf();
return out.str();
}
// clang-format off
std::vector<std::string> const env_vars_{
"FEDERATED_SERVER_ADDRESS",
"FEDERATED_WORLD_SIZE",
"FEDERATED_RANK",
"FEDERATED_SERVER_CERT",
"FEDERATED_CLIENT_KEY",
"FEDERATED_CLIENT_CERT" };
// clang-format on
std::string server_address_{"localhost:9091"};
int world_size_{1};
int rank_{0};
std::string server_cert_{};
std::string client_key_{};
std::string client_cert_{};
std::unique_ptr<xgboost::federated::FederatedClient> client_{};
int version_number_{0};
};
// Singleton federated engine.
FederatedEngine engine; // NOLINT(cert-err58-cpp)
/*! \brief initialize the synchronization module */
bool Init(int argc, char *argv[]) {
try {
engine.Init(argc, argv);
return true;
} catch (std::exception const &e) {
fprintf(stderr, " failed in federated Init %s\n", e.what());
return false;
}
}
/*! \brief finalize synchronization module */
bool Finalize() {
try {
engine.Finalize();
return true;
} catch (const std::exception &e) {
fprintf(stderr, "failed in federated shutdown %s\n", e.what());
return false;
}
}
/*! \brief singleton method to get engine */
IEngine *GetEngine() { return &engine; }
// perform in-place allreduce, on sendrecvbuf
void Allreduce_(void *sendrecvbuf, size_t type_nbytes, size_t count, IEngine::ReduceFunction red,
mpi::DataType dtype, mpi::OpType op, IEngine::PreprocFunction prepare_fun,
void *prepare_arg) {
if (prepare_fun != nullptr) prepare_fun(prepare_arg);
if (engine.GetWorldSize() == 1) return;
engine.Allreduce(sendrecvbuf, type_nbytes * count, dtype, op);
}
} // namespace engine
} // namespace rabit