xref: /external/tensorflow/tensorflow/core/distributed_runtime/rpc/grpc_session.cc

/* Copyright 2016 The TensorFlow 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 "tensorflow/core/distributed_runtime/rpc/grpc_session.h"

#include <unordered_map>

#include "tensorflow/core/common_runtime/session_factory.h"
#include "tensorflow/core/distributed_runtime/call_options.h"
#include "tensorflow/core/distributed_runtime/local_master.h"
#include "tensorflow/core/distributed_runtime/master_interface.h"
#include "tensorflow/core/distributed_runtime/request_id.h"
#include "tensorflow/core/distributed_runtime/rpc/grpc_channel.h"
#include "tensorflow/core/distributed_runtime/rpc/grpc_remote_master.h"
#include "tensorflow/core/framework/attr_value.pb.h"
#include "tensorflow/core/framework/node_def.pb.h"
#include "tensorflow/core/lib/core/errors.h"
#include "tensorflow/core/lib/strings/str_util.h"
#include "tensorflow/core/platform/mutex.h"
#include "tensorflow/core/protobuf/master.pb.h"

namespace tensorflow {

const char* const kSchemePrefix = "grpc://";
const size_t kSchemePrefixLength = strlen(kSchemePrefix);

GrpcSession::GrpcSession(const SessionOptions& options)
    : options_(options), current_graph_version_(-1) {}

GrpcSession::~GrpcSession() {}

/* static */
Status GrpcSession::Create(const SessionOptions& options,
                           std::unique_ptr<GrpcSession>* out_session) {
  std::unique_ptr<GrpcSession> session(new GrpcSession(options));
  std::unique_ptr<MasterInterface> master;
  // For testing, we enable the client to disable the use of the local
  // master registry, so that the RPC stack is exercised.
  if (!options.config.rpc_options().use_rpc_for_inprocess_master()) {
    master = LocalMaster::Lookup(options.target);
  }
  if (!master) {
    SharedGrpcChannelPtr master_channel;
    TF_RETURN_IF_ERROR(
        NewHostPortGrpcChannel(options.target.substr(kSchemePrefixLength),
                               &options.config.rpc_options(), &master_channel));
    master.reset(NewGrpcMaster(master_channel));
  }
  session->SetRemoteMaster(std::move(master));
  *out_session = std::move(session);
  return Status::OK();
}

namespace {
// Re-encodes constant represented in tensor proto into
// tensor_content, which is slightly better (less copies and lower peak
// memory usage) when used with rpc subsystems.
void ReEncodeConsts(GraphDef* gdef) {
  for (NodeDef& ndef : *(gdef->mutable_node())) {
    if (ndef.op() == "Const") {
      TensorProto* proto = nullptr;
      for (auto& attr : *ndef.mutable_attr()) {
        if (attr.first == "value") {
          proto = attr.second.mutable_tensor();
        }
      }
      if (proto != nullptr && proto->tensor_content().empty() &&
          proto->ByteSizeLong() > 64) {
        // If the constant is encoded with repeated proto fields and
        // it is moderate large, we re-encode it in tensor_content as
        // a Cord. This is mildly helpful for reducing the peak memory
        // usage on the server side where GraphDef/NodeDef are copied
        // quite often.
        Tensor parsed(proto->dtype());
        if (parsed.FromProto(*proto)) {
          parsed.AsProtoTensorContent(proto);
        }
      }
    }
  }
}
}  // namespace

void GrpcSession::SetHandleAndGraphVersion(string handle, int64 graph_version) {
  mutex_lock l(mu_);
  handle_ = std::move(handle);
  current_graph_version_ = graph_version;
}

Status GrpcSession::Handle(string* out_handle) {
  mutex_lock l(mu_);
  if (handle_.empty()) {
    return errors::InvalidArgument("A session is not created yet....");
  }
  *out_handle = handle_;
  return Status::OK();
}

Status GrpcSession::CreateImpl(CallOptions* call_options,
                               const GraphDef& graph) {
  {
    mutex_lock l(mu_);
    if (!handle_.empty()) {
      return errors::InvalidArgument("A session is alive.");
    }
  }
  CreateSessionRequest req;
  *req.mutable_config() = options_.config;
  *req.mutable_graph_def() = graph;
  req.set_target(options_.target);
  ReEncodeConsts(req.mutable_graph_def());
  CreateSessionResponse resp;
  Status s = master_->CreateSession(call_options, &req, &resp);
  if (s.ok()) {
    SetHandleAndGraphVersion(resp.session_handle(), resp.graph_version());
  }
  return s;
}

Status GrpcSession::Create(const GraphDef& graph) {
  CallOptions call_options;
  call_options.SetTimeout(options_.config.operation_timeout_in_ms());
  return CreateImpl(&call_options, graph);
}

Status GrpcSession::Create(const RunOptions& run_options,
                           const GraphDef& graph) {
  CallOptions call_options;
  call_options.SetTimeout(run_options.timeout_in_ms());
  return CreateImpl(&call_options, graph);
}

Status GrpcSession::ExtendImpl(CallOptions* call_options,
                               const GraphDef& graph) {
  bool handle_is_empty;
  {
    mutex_lock l(mu_);
    handle_is_empty = handle_.empty();
  }
  if (handle_is_empty) {
    // Session was unitialized, so simply initialize the session with 'graph'.
    return Create(graph);
  }
  mutex_lock l(mu_);
  ExtendSessionRequest req;
  req.set_session_handle(handle_);
  *req.mutable_graph_def() = graph;
  req.set_current_graph_version(current_graph_version_);
  ExtendSessionResponse resp;
  Status s = master_->ExtendSession(call_options, &req, &resp);
  if (s.ok()) {
    current_graph_version_ = resp.new_graph_version();
  }
  return s;
}

Status GrpcSession::Extend(const GraphDef& graph) {
  CallOptions call_options;
  call_options.SetTimeout(options_.config.operation_timeout_in_ms());
  return ExtendImpl(&call_options, graph);
}

Status GrpcSession::Extend(const RunOptions& run_options,
                           const GraphDef& graph) {
  CallOptions call_options;
  call_options.SetTimeout(run_options.timeout_in_ms());
  return ExtendImpl(&call_options, graph);
}

Status GrpcSession::RunHelper(
    const RunOptions& run_options,
    const std::vector<std::pair<string, Tensor>>& inputs,
    const std::vector<string>& output_tensor_names,
    const std::vector<string>& target_node_names, std::vector<Tensor>* outputs,
    RunMetadata* run_metadata, const string& prun_handle) {
  // Convert to proto
  std::unique_ptr<MutableRunStepRequestWrapper> req(
      master_->CreateRunStepRequest());
  std::unique_ptr<MutableRunStepResponseWrapper> resp(
      master_->CreateRunStepResponse());

  *req->mutable_options() = run_options;

  if (run_options.timeout_in_ms() == 0) {
    req->mutable_options()->set_timeout_in_ms(
        options_.config.operation_timeout_in_ms());
  }

  if (!prun_handle.empty()) {
    req->set_partial_run_handle(prun_handle);
  }

  for (const auto& it : inputs) {
    req->add_feed(it.first, it.second);
  }

  // Support long error messages by storing the error code in the response body.
  req->set_store_errors_in_response_body(true);

  // Build an index from fetch tensor name to first index in
  // output_tensor_names.
  std::unordered_map<string, int> output_name_to_offset;
  for (int i = 0; i < output_tensor_names.size(); ++i) {
    const string& name = output_tensor_names[i];
    if (output_name_to_offset.insert(std::make_pair(name, i)).second) {
      req->add_fetch(name);
    }
  }
  for (const string& target : target_node_names) {
    req->add_target(target);
  }

  CallOptions call_options;
  call_options.SetTimeout(req->options().timeout_in_ms());
  TF_RETURN_IF_ERROR(RunProto(&call_options, req.get(), resp.get()));

  // Look for an extended error returned in the response body.
  if (resp->status_code() != error::Code::OK) {
    return Status(resp->status_code(), resp->status_error_message());
  }

  if (!output_tensor_names.empty()) {
    outputs->resize(output_tensor_names.size());
  }

  // Convert response back to Tensors in the correct order.
  for (size_t i = 0; i < resp->num_tensors(); ++i) {
    auto fetch_it = output_name_to_offset.find(resp->tensor_name(i));
    if (fetch_it == output_name_to_offset.end()) {
      return errors::Internal("Received response for unrequested fetch: ",
                              resp->tensor_name(i));
    }

    Tensor output;
    TF_RETURN_IF_ERROR(resp->TensorValue(i, &output));
    (*outputs)[fetch_it->second] = output;
  }
  // In the unlikely event that output_tensor_names contains duplicates, fill in
  // the duplicate values.
  if (output_name_to_offset.size() != output_tensor_names.size()) {
    for (int i = 0; i < output_tensor_names.size(); ++i) {
      const string& name = output_tensor_names[i];
      int offset = output_name_to_offset[name];
      if (offset != i) {
        (*outputs)[i] = (*outputs)[offset];
      }
    }
  }

  if (run_metadata) {
    run_metadata->Swap(resp->mutable_metadata());
  }

  return Status::OK();
}

Status GrpcSession::Run(const RunOptions& run_options,
                        const std::vector<std::pair<string, Tensor>>& inputs,
                        const std::vector<string>& output_tensor_names,
                        const std::vector<string>& target_node_names,
                        std::vector<Tensor>* outputs,
                        RunMetadata* run_metadata) {
  return RunHelper(run_options, inputs, output_tensor_names, target_node_names,
                   outputs, run_metadata, /* prun_handle */ "");
}

Status GrpcSession::Run(const std::vector<std::pair<string, Tensor>>& inputs,
                        const std::vector<string>& output_tensor_names,
                        const std::vector<string>& target_node_names,
                        std::vector<Tensor>* outputs) {
  RunOptions run_options;
  run_options.set_timeout_in_ms(options_.config.operation_timeout_in_ms());
  return Run(run_options, inputs, output_tensor_names, target_node_names,
             outputs, nullptr);
}

Status GrpcSession::RunProto(CallOptions* call_options,
                             MutableRunStepRequestWrapper* req,
                             MutableRunStepResponseWrapper* resp) {
  string handle;
  TF_RETURN_IF_ERROR(Handle(&handle));
  req->set_session_handle(handle);
  return master_->RunStep(call_options, req, resp);
}

Status GrpcSession::PRunSetup(const std::vector<string>& input_names,
                              const std::vector<string>& output_names,
                              const std::vector<string>& target_nodes,
                              string* handle) {
  // Convert to proto
  PartialRunSetupRequest req;
  PartialRunSetupResponse resp;
  CallOptions call_options;
  TF_RETURN_IF_ERROR(Handle(req.mutable_session_handle()));
  for (const string& feed : input_names) {
    req.add_feed(feed);
  }
  for (const string& fetch : output_names) {
    req.add_fetch(fetch);
  }
  for (const string& target : target_nodes) {
    req.add_target(target);
  }
  req.set_request_id(GetUniqueRequestId());
  call_options.SetTimeout(options_.config.operation_timeout_in_ms());
  TF_RETURN_IF_ERROR(master_->PartialRunSetup(&call_options, &req, &resp));
  *handle = resp.partial_run_handle();
  return Status::OK();
}

Status GrpcSession::PRun(const string& handle,
                         const std::vector<std::pair<string, Tensor>>& inputs,
                         const std::vector<string>& output_names,
                         std::vector<Tensor>* outputs) {
  RunOptions run_options;
  run_options.set_timeout_in_ms(options_.config.operation_timeout_in_ms());
  return RunHelper(run_options, inputs, output_names, /* targets */ {}, outputs,
                   /* run_metadata */ nullptr, handle);
}

Status GrpcSession::Close() {
  CloseSessionRequest req;
  {
    mutex_lock l(mu_);
    if (handle_.empty()) {
      return Status::OK();
    }
    req.set_session_handle(handle_);
    handle_.clear();
  }
  CloseSessionResponse resp;
  CallOptions call_options;
  call_options.SetTimeout(options_.config.operation_timeout_in_ms());
  return master_->CloseSession(&call_options, &req, &resp);
}

Status GrpcSession::ListDevices(std::vector<DeviceAttributes>* response) {
  ListDevicesRequest req;
  {
    mutex_lock l(mu_);
    req.set_session_handle(handle_);
  }
  if (req.session_handle().empty()) {
    LOG(WARNING) << "GrpcSession::ListDevices will initialize the session with "
                    "an empty graph and other defaults because the session has "
                    "not yet been created.";
    GraphDef graph_def;
    TF_RETURN_IF_ERROR(Create(graph_def));
    {
      mutex_lock l(mu_);
      req.set_session_handle(handle_);
    }
  }
  ListDevicesResponse resp;
  CallOptions call_options;
  call_options.SetTimeout(options_.config.operation_timeout_in_ms());
  Status s = master_->ListDevices(&call_options, &req, &resp);
  if (!s.ok()) {
    LOG(ERROR) << "Could not list devices: " << s;
    return s;
  }

  response->clear();
  response->reserve(resp.local_device_size() + resp.remote_device_size());
  for (const auto& device_attr : resp.local_device()) {
    response->emplace_back(device_attr);
  }
  for (const auto& device_attr : resp.remote_device()) {
    response->emplace_back(device_attr);
  }
  return Status::OK();
}

void GrpcSession::SetRemoteMaster(std::unique_ptr<MasterInterface> master) {
  master_ = std::move(master);
}

// Static method.
Status GrpcSession::Reset(const SessionOptions& options,
                          const std::vector<string>& containers) {
  SharedGrpcChannelPtr master_channel;
  TF_RETURN_IF_ERROR(
      NewHostPortGrpcChannel(options.target.substr(kSchemePrefixLength),
                             /*rpc_options=*/nullptr, &master_channel));
  auto master = NewGrpcMaster(master_channel);
  ResetRequest req;
  for (const auto& c : containers) req.add_container(c);
  ResetResponse resp;
  CallOptions call_options;
  call_options.SetTimeout(options.config.operation_timeout_in_ms());
  Status ret = master->Reset(&call_options, &req, &resp);
  delete master;
  return ret;
}

Status GrpcSession::MakeCallable(const CallableOptions& callable_options,
                                 CallableHandle* out_handle) {
  MakeCallableRequest req;
  TF_RETURN_IF_ERROR(Handle(req.mutable_session_handle()));
  *req.mutable_options() = callable_options;
  req.set_request_id(GetUniqueRequestId());
  MakeCallableResponse resp;
  CallOptions call_options;
  call_options.SetTimeout(options_.config.operation_timeout_in_ms());
  TF_RETURN_IF_ERROR(master_->MakeCallable(&call_options, &req, &resp));
  *out_handle = resp.handle();
  return Status::OK();
}

Status GrpcSession::RunCallable(CallableHandle handle,
                                const std::vector<Tensor>& feed_tensors,
                                std::vector<Tensor>* fetch_tensors,
                                RunMetadata* run_metadata) {
  RunCallableRequest req;
  TF_RETURN_IF_ERROR(Handle(req.mutable_session_handle()));
  req.set_handle(handle);
  req.set_request_id(GetUniqueRequestId());
  for (const Tensor& feed : feed_tensors) {
    feed.AsProtoTensorContent(req.mutable_feed()->Add());
  }

  RunCallableResponse resp;
  CallOptions call_options;
  call_options.SetTimeout(options_.config.operation_timeout_in_ms());
  TF_RETURN_IF_ERROR(master_->RunCallable(&call_options, &req, &resp));
  for (const TensorProto& fetch : resp.fetch()) {
    Tensor fetch_tensor;
    if (!fetch_tensor.FromProto(cpu_allocator(), fetch)) {
      return errors::Internal(
          "Could not parse fetched tensor data in response from master.");
    }
    fetch_tensors->push_back(std::move(fetch_tensor));
  }
  return Status::OK();
}

Status GrpcSession::ReleaseCallable(CallableHandle handle) {
  ReleaseCallableRequest req;
  TF_RETURN_IF_ERROR(Handle(req.mutable_session_handle()));
  req.set_handle(handle);
  ReleaseCallableResponse resp;
  CallOptions call_options;
  call_options.SetTimeout(options_.config.operation_timeout_in_ms());
  return master_->ReleaseCallable(&call_options, &req, &resp);
}

class GrpcSessionFactory : public SessionFactory {
 public:
  bool AcceptsOptions(const SessionOptions& options) override {
    return str_util::StartsWith(options.target, kSchemePrefix);
  }

  Status NewSession(const SessionOptions& options,
                    Session** out_session) override {
    std::unique_ptr<GrpcSession> session;
    TF_RETURN_IF_ERROR(GrpcSession::Create(options, &session));
    *out_session = session.release();
    return Status::OK();
  }

  // Invokes the session specific static method to reset containers.
  Status Reset(const SessionOptions& options,
               const std::vector<string>& containers) override {
    return GrpcSession::Reset(options, containers);
  }
};

class GrpcSessionRegistrar {
 public:
  GrpcSessionRegistrar() {
    SessionFactory::Register("GRPC_SESSION", new GrpcSessionFactory());
  }
};
static GrpcSessionRegistrar registrar;

}  // namespace tensorflow