fix code style

paddle/fluid/train/custom_trainer/feed/accessor/epoch_accessor.cc

@@ -126,7 +126,7 @@ namespace feed {
             case ModelSaveWay::ModelSaveInferenceBase:
                 return is_last_epoch(epoch_id);
             case ModelSaveWay::ModelSaveTrainCheckpoint:
-                return ((epoch_id / SecondsPerHour) % 8) == 0;
+                return delta_id(epoch_id) % 8 == 0;
         }
         return false;
     }

paddle/fluid/train/custom_trainer/feed/common/pslib_warpper.h

-/* 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. */
-
 #pragma once
 
 // Hide BLOG

paddle/fluid/train/custom_trainer/feed/common/runtime_environment.cc

@@ -139,14 +139,13 @@ public:
         ar.SetCursor(ar.Buffer());
         MPI_Bcast(ar.Buffer(), len, MPI_BYTE, root_id, node_info.mpi_comm);
     }
-
-    virtual double all_reduce_ele(double x) {
-        double tot = 0.0;
-        MPI_Allreduce(&x, &tot, 1, mpi_type_trait<double>::type(), MPI_SUM, MPI_COMM_WORLD);
-        return tot;
-    }
-    virtual void all_reduce_arr(double* x, int n) {
-        MPI_Allreduce(MPI_IN_PLACE, x, n, mpi_type_trait<double>::type(), MPI_SUM, MPI_COMM_WORLD);
+    virtual void all_reduce_in_place(double* x, int n, ReduceOperator op, EnvironmentRole role) {
+        auto& node_info = mpi_node_info(role);
+        if (op == ReduceOperator::SUM) {
+            MPI_Allreduce(MPI_IN_PLACE, x, n, MPI_DOUBLE, MPI_SUM, node_info.mpi_comm);
+        } else {
+            CHECK(false) << "unsupport operator";
+        }
     }
 
 protected:
@@ -201,10 +200,7 @@ public:
     virtual void bcast(paddle::framework::BinaryArchive& ar, int root_id, EnvironmentRole role) {
         return;
     }
-    virtual double all_reduce_ele(double x) {
-        return x;
-    }
-    virtual void all_reduce_arr(double* x, int n) {
+    virtual void all_reduce_in_place(double* x, int n, ReduceOperator op, EnvironmentRole role) {
         return;
     }
 protected:

paddle/fluid/train/custom_trainer/feed/common/runtime_environment.h

@@ -27,7 +27,7 @@ enum class EnvironmentLogType {
     ALL_LOG         = 1         //所有节点都会对外输出
 };
 
-//保持该枚举值的连续递增，且ALL在尾部
+// 保持该枚举值的连续递增，且ALL在尾部
 enum class EnvironmentRole {
     WORKER          = 0,        //训练Worker
     PSERVER         = 1,        //参数服务器
@@ -35,6 +35,11 @@ enum class EnvironmentRole {
     ALL             = 2         //所有角色，请保持在枚举尾部
 };
 
+// Reduce的操作类型
+enum class ReduceOperator {
+    SUM             = 0         //求和
+};
+
 class RuntimeEnvironment {
 public:
     RuntimeEnvironment();
@@ -72,10 +77,15 @@ public:
     virtual void barrier(EnvironmentRole role) = 0;
     // bcast 广播
     virtual void bcast(paddle::framework::BinaryArchive& ar, int root_id, EnvironmentRole role) = 0;
-    // all_reduce sum element 规约元素
-    virtual double all_reduce_ele(double x) = 0;
-    // all_reduce sum array 规约数组
-    virtual void all_reduce_arr(double* x, int n) = 0;
+    // 全局reduce操作, 返回reduce结果
+    virtual double all_reduce(double x, ReduceOperator op, EnvironmentRole role) {
+        double result = x;
+        all_reduce_in_place(&result, 1, op, role);
+        return result;
+    }
+    // 全局reduce，就地执行
+    virtual void all_reduce_in_place(double* x, int n, 
+            ReduceOperator op, EnvironmentRole role) = 0;
     // 接口只允许在主线程调用   End
 protected:
     virtual void print_log(EnvironmentRole role, EnvironmentLogType type, 

paddle/fluid/train/custom_trainer/feed/conf/gflags.conf

 -log_dir=log
--v=4
+-v=2
 -logbufsecs=0
 -pslib_push_dense_merge_limit=1
 -pslib_push_sparse_merge_limit=1

paddle/fluid/train/custom_trainer/feed/executor/multi_thread_executor.cc

 #include "paddle/fluid/train/custom_trainer/feed/io/file_system.h"
+#include "paddle/fluid/train/custom_trainer/feed/monitor/monitor.h"
 #include "paddle/fluid/train/custom_trainer/feed/executor/multi_thread_executor.h"
 
 namespace paddle {
@@ -63,11 +64,23 @@ int MultiThreadExecutor::initialize(YAML::Node exe_config,
         }
     } 
 
+    // Monitor组件
+    for (const auto& monitor_config : _model_config["monitor"]) {
+        auto monitor_class = monitor_config["class"].as<std::string>();
+        auto* monitor_ptr = CREATE_INSTANCE(Monitor, monitor_class);
+        _monitors.emplace_back(monitor_ptr);
+        CHECK(monitor_ptr->initialize(monitor_config, context_ptr) == 0)
+            << "Monitor init Failed, class:" << monitor_class;
+    }
     return ret;
 }
 
 paddle::framework::Channel<DataItem> MultiThreadExecutor::run(
     paddle::framework::Channel<DataItem> input, const DataParser* parser) {
+
+    uint64_t epoch_id = _trainer_context->epoch_accessor->current_epoch_id();
+
+    // 输入流
     PipelineOptions input_pipe_option;
     input_pipe_option.need_hold_input_data = true;
     input_pipe_option.batch_size = 1;
@@ -97,6 +110,7 @@ paddle::framework::Channel<DataItem> MultiThreadExecutor::run(
             return 0;
         });
     
+    // 训练流
     PipelineOptions train_pipe_option;
     train_pipe_option.input_output_rate = 1;
     train_pipe_option.thread_num = _train_thread_num;
@@ -108,19 +122,20 @@ paddle::framework::Channel<DataItem> MultiThreadExecutor::run(
             auto* executor = _thread_executors[thread_idx].get();
             size_t& out_idx = *out_num;
             for (out_idx = 0; out_idx < in_num; ++out_idx) {
-                //CHECK(executor->run(in_items[out_idx].get()) == 0);
+                CHECK(executor->run(in_items[out_idx].get()) == 0);
                 out_items[out_idx] = std::move(in_items[out_idx]);
             }
             return 0;
         });
 
+    // 梯度回传流
     PipelineOptions gradient_pipe_option;
     gradient_pipe_option.input_output_rate = 1;
     gradient_pipe_option.thread_num = _push_gradient_thread_num;
     gradient_pipe_option.buffer_batch_count = 2 * _train_thread_num;
     auto gradient_pipe = std::make_shared<Pipeline<ScopePoolObj, int>>();
     gradient_pipe->connect_to(*train_pipe, gradient_pipe_option, 
-        [this] (ScopePoolObj* in_items, size_t in_num, 
+        [epoch_id, this] (ScopePoolObj* in_items, size_t in_num, 
             int* out_items, size_t* out_num, size_t thread_idx) -> int {
             size_t& out_idx = *out_num;
             for (out_idx = 0; out_idx < in_num; ++out_idx) {
@@ -134,14 +149,27 @@ paddle::framework::Channel<DataItem> MultiThreadExecutor::run(
                     out_items[out_idx] = _input_accessors[i]->
                         backward(samples, sample_num, scope);
                 }
+                for (auto& monitor : _monitors) {
+                    monitor->add_data(epoch_id, this, samples, sample_num);
+                }
                 delete[] samples; // 所有pipe完成后，再回收sample
             }
             return 0;
         });
 
+    // 等待训练流结束
     std::vector<int> gradient_status;
     while (gradient_pipe->read(gradient_status) > 0) {
     }
+    // 输出相关监控&统计项
+    for (auto& monitor : _monitors) {
+        if (monitor->need_compute_result(epoch_id)) {
+            monitor->compute_result();
+            VLOG(2) << "[Monitor]" << _train_exe_name << ", monitor:" << monitor->get_name()
+                << ", result:" << monitor->format_result();
+            monitor->reset();
+        }
+    }
     return input_pipe->backup_channel();
 }
 

paddle/fluid/train/custom_trainer/feed/executor/multi_thread_executor.h

@@ -8,6 +8,7 @@ namespace paddle {
 namespace custom_trainer {
 namespace feed {
 
+class Monitor;
 typedef paddle::ps::ObjectPool<::paddle::framework::Scope>::PooledObject ScopePoolObj;
 
 class MultiThreadExecutor {
@@ -50,6 +51,7 @@ protected:
     YAML::Node _model_config;
     std::string _train_exe_name;
     TrainerContext* _trainer_context = nullptr;
+    std::vector<std::shared_ptr<Monitor>> _monitors;
     std::vector<std::shared_ptr<Executor>> _thread_executors;
     std::vector<std::shared_ptr<DataInputAccessor>> _input_accessors;
     std::map<uint32_t, std::vector<DataInputAccessor*>> _table_to_accessors;

paddle/fluid/train/custom_trainer/feed/monitor/auc_monitor.cc

@@ -6,8 +6,8 @@ namespace feed {
 
 int AucMonitor::initialize(const YAML::Node& config, std::shared_ptr<TrainerContext> context_ptr) {
     Monitor::initialize(config, context_ptr);
+    _target_idx = config["target_idx"].as<int32_t>();
     _target_name = config["target"].as<std::string>();
-    _label_name = config["label"].as<std::string>();
     _table_size = 1000000;
     if (config["table_size"]) {
         _table_size = config["table_size"].as<int>();
@@ -15,45 +15,34 @@ int AucMonitor::initialize(const YAML::Node& config, std::shared_ptr<TrainerCont
     set_table_size(_table_size);
     _compute_interval = 3600;
     if (config["compute_interval"]) {
-        uint32_t interval = config["compute_interval"].as<uint32_t>();
-        if (interval != 3600 || interval != 86400) {
-            LOG(FATAL) << " AucMonitor config compute_interval just support hour: 3600 or day: 86400. ";
-            return -1;
-        }
-        _compute_interval = interval;
+        _compute_interval = config["compute_interval"].as<uint32_t>();
+        CHECK(_compute_interval % 60 == 0);
     }
+    return 0;
 }
 
-void AucMonitor::add_data(int epoch_id, const Executor* executor, SampleInstance* instance, size_t num) {
-    if (executor == nullptr
-            || instance == nullptr 
-            || instance->predicts.empty()
-            || instance->labels.empty()
-            || num <= 0
-            || instance->predicts.size() < num
-            || instance->labels.size() < num) {
-        LOG(FATAL) << "AucMonitor add predict data is invalid, predicts or labels is empty, num[" << num << "]";
-        return;
-    }
+void AucMonitor::add_data(int epoch_id, 
+    const MultiThreadExecutor* executor, SampleInstance* samples, size_t num) {
+    CHECK(num > 0);
     std::lock_guard<std::mutex> lock(_mutex);
     for (int i = 0; i < num; ++i) {
-        add_unlocked(instance->predicts[i], std::lround(instance->labels[i]));
+        auto& instance = samples[i]; 
+        add_unlocked(instance.predicts[_target_idx], std::lround(instance.labels[_target_idx]));
     }
 }
 
-bool AucMonitor::need_compute_result(int epoch_id, EpochAccessor* accessor) {
-    CHECK(accessor != nullptr);
-    uint64_t epoch_time = accessor->epoch_timestamp(epoch_id);
-    if (epoch_time % _compute_interval != 0) {
-        return false;
-    }
-    return true;
+bool AucMonitor::need_compute_result(int epoch_id) {
+    CHECK(_epoch_accessor != nullptr);
+    uint64_t epoch_time = _epoch_accessor->epoch_timestamp(epoch_id);
+    return epoch_time % _compute_interval == 0;
 }
 
 void AucMonitor::compute_result() {
+    auto* environment = Monitor::_context_ptr->environment.get();
     double* table[2] = {&_table[0][0], &_table[1][0]};
     for (int i = 0; i < 2; i++) {
-        Monitor::_context_ptr->environment->all_reduce_arr(table[i], _table_size);
+        environment->all_reduce_in_place(table[i], 
+            _table_size, ReduceOperator::SUM, EnvironmentRole::WORKER);
     }
     double area = 0;
     double fp = 0;
@@ -66,11 +55,14 @@ void AucMonitor::compute_result() {
         tp = newtp;
     }
     _auc = area / (fp * tp);
-    _mae = Monitor::_context_ptr->environment->all_reduce_ele(_local_abserr) / (fp + tp);
-    _rmse = sqrt(Monitor::_context_ptr->environment->all_reduce_ele(_local_sqrerr) / (fp + tp));
+    _mae = environment->all_reduce(_local_abserr, 
+        ReduceOperator::SUM, EnvironmentRole::WORKER) / (fp + tp);
+    _rmse = sqrt(environment->all_reduce(_local_sqrerr,
+        ReduceOperator::SUM, EnvironmentRole::WORKER) / (fp + tp));
     _rmse = sqrt(_rmse / (fp + tp));
     _actual_ctr = tp / (fp + tp);
-    _predicted_ctr = Monitor::_context_ptr->environment->all_reduce_ele(_local_pred) / (fp + tp);
+    _predicted_ctr = environment->all_reduce(_local_pred,
+        ReduceOperator::SUM, EnvironmentRole::WORKER) / (fp + tp);
     _size = fp + tp;
     calculate_bucket_error();
 }
@@ -81,9 +73,8 @@ std::string AucMonitor::format_result() {
         copc = _actual_ctr / _predicted_ctr;
     }
     char buf[10240];
-    snprintf(buf, 10240 * sizeof(char), "%s: AUC=%.6f BUCKET_ERROR=%.6f MAE=%.6f RMSE=%.6f "
+    snprintf(buf, 10240 * sizeof(char), "AUC=%.6f BUCKET_ERROR=%.6f MAE=%.6f RMSE=%.6f "
         "Actual CTR=%.6f Predicted CTR=%.6f COPC=%.6f INS Count=%.0f",
-        Monitor::_name.c_str(), 
         _auc,
         _bucket_error,
         _mae, 
@@ -157,6 +148,8 @@ void AucMonitor::reset() {
     _local_pred = 0;
 }
 
+REGIST_CLASS(Monitor, AucMonitor);
+
 }  // namespace feed
 }  // namespace custom_trainer
 }  // namespace paddle

paddle/fluid/train/custom_trainer/feed/monitor/auc_monitor.h

@@ -18,13 +18,11 @@ public:
         std::shared_ptr<TrainerContext> context_ptr) override;
 
     //添加一项记录，统计内容Monitor自行从Executor按需获取
-    virtual void add_data(int epoch_id, 
-            const Executor* executor, 
-            SampleInstance* instance, 
-            size_t num);
+    virtual void add_data(int epoch_id, const MultiThreadExecutor* executor, 
+            SampleInstance* samples, size_t num);
     
     //是否开始结果统计
-    virtual bool need_compute_result(int epoch_id, EpochAccessor* accessor);
+    virtual bool need_compute_result(int epoch_id);
     //统计当前结果
     virtual void compute_result();
     //基于现有结果，输出格式化的统计信息
@@ -33,7 +31,7 @@ public:
     virtual void reset();
 
 protected:
-    std::string _label_name;
+    uint32_t _target_idx;
     std::string _target_name;
     std::string _name;
     std::string _output_var;
@@ -47,7 +45,7 @@ protected:
     double _bucket_error;
     int _table_size;
     void add_unlocked(double pred, int label);
-
+    
 private:
     void calculate_bucket_error();
     void set_table_size(int table_size);

paddle/fluid/train/custom_trainer/feed/monitor/monitor.h

@@ -9,6 +9,7 @@
 namespace paddle {
 namespace custom_trainer {
 namespace feed {
+class MultiThreadExecutor;
 
 class Monitor {
 public:
@@ -19,14 +20,16 @@ public:
         std::shared_ptr<TrainerContext> context_ptr) {
         _name = config["name"].as<std::string>();
         _context_ptr = context_ptr;
+        _epoch_accessor = _context_ptr->epoch_accessor.get();
         return 0;
     }
 
     //添加一项记录，统计内容Monitor自行从Executor按需获取
-    virtual void add_data(int epoch_id, const Executor* executor, SampleInstance* instance, size_t num) = 0;
+    virtual void add_data(int epoch_id, const MultiThreadExecutor* executor, 
+            SampleInstance* samples, size_t num) = 0;
     
     //是否对于当前epoch_id进行结果统计
-    virtual bool need_compute_result(int epoch_id, EpochAccessor* accessor) = 0;
+    virtual bool need_compute_result(int epoch_id) = 0;
     //统计当前结果
     virtual void compute_result() = 0;
     //基于现有结果，输出格式化的统计信息
@@ -40,6 +43,7 @@ public:
 
 protected:
     std::string _name;
+    EpochAccessor* _epoch_accessor = nullptr;
     std::shared_ptr<TrainerContext> _context_ptr;
 };
 

paddle/fluid/train/custom_trainer/feed/scripts/create_programs.py

@@ -95,7 +95,8 @@ class ModelBuilder:
         main_program = fluid.Program()
         startup_program = fluid.Program()
         with fluid.program_guard(main_program, startup_program):
-            input_accessor, sparses, inputs, outputs = self._inference()
+            #TODO return dict maybe better ?
+            input_accessor, sparses, inputs, outputs, monitors = self._inference()
             test_program = main_program.clone(for_test=True)
             loss, labels = self._loss_function(*outputs)
 
@@ -134,7 +135,11 @@ class ModelBuilder:
                 accessor["input"] = [
                     {"label_name": label.name, "shape": label.shape, "output_name": output.name } 
                     for (label, output) in zip(labels, outputs) ]
-            
+       
+        for monitor in monitors:
+            idx = outputs.index(monitor['target'])
+            monitor["target_idx"] = idx
+            monitor["target"] = outputs[idx].name
 
         model_desc_path = os.path.join(self._save_path, 'model.yaml')
         model_desc = {
@@ -142,7 +147,9 @@ class ModelBuilder:
             'outputs': [{"name": var.name, "shape": var.shape} for var in outputs],
             'labels': [{"name": var.name, "shape": var.shape} for var in labels],
             'loss': loss.name,
-            'input_accessor': input_accessor
+            'input_accessor': input_accessor,
+            'monitor': monitors,
+            'aa_Attention' : 'Do Not Modify This File Manually, Unless You Really Know It'
         }
 
         with open(model_desc_path, 'w') as f:

paddle/fluid/train/custom_trainer/feed/scripts/example.py

@@ -32,7 +32,7 @@ def inference():
     net = fluid.layers.fc(net, 128, act='relu', name='fc_7')
 
     ctr_output = fluid.layers.fc(net, 1, act='sigmoid', name='ctr')
-    return [cvm_input], [ctr_output]
+    return [], [], [cvm_input], [ctr_output], monitors
 
 def loss_function(ctr_output):
     """

paddle/fluid/train/custom_trainer/feed/scripts/join.py

@@ -46,7 +46,11 @@ def inference():
         { "class": "DenseInputAccessor", "input": "sums", "table_id": 2, "need_gradient": True, "async_pull": True},
         { "class": "LabelInputAccessor", "input": "labels"}
     ]
-    return accessors, [sparse_cvm], [cvm_input], [ctr_output]
+    monitors = [
+        { "name": "epoch_auc", "class": "AucMonitor", "target": ctr_output, "compute_interval": 600 },
+        { "name": "day_auc", "class": "AucMonitor", "target": ctr_output, "compute_interval": 86400 }
+    ]
+    return accessors, [sparse_cvm], [cvm_input], [ctr_output], monitors
 
 def loss_function(ctr_output):
     """

paddle/fluid/train/custom_trainer/feed/scripts/model/join/model.yaml

+aa_Attention: Do Not Modify This File Manually, Unless You Really Know It
 input_accessor:
 - class: AbacusSparseUpdateAccessor
   input:
@@ -98,6 +99,10 @@ labels:
 - name: label_ctr
   shape: [-1, 1]
 loss: loss_ctr
+monitor:
+- {class: AucMonitor, compute_interval: 3600, name: epoch_auc, target: ctr.tmp_2,
+  target_idx: 0}
+- {class: AucMonitor, compute_interval: 86400, name: day_auc, target: ctr.tmp_2, target_idx: 0}
 outputs:
 - name: ctr.tmp_2
   shape: [-1, 1]

paddle/fluid/train/custom_trainer/feed/scripts/model/update/model.yaml

+aa_Attention: Do Not Modify This File Manually, Unless You Really Know It
 input_accessor:
 - class: AbacusSparseUpdateAccessor
   input:
@@ -79,6 +80,10 @@ labels:
 - name: label_ctr
   shape: [-1, 1]
 loss: loss_ctr
+monitor:
+- {class: AucMonitor, compute_interval: 3600, name: epoch_auc, target: ctr.tmp_2,
+  target_idx: 0}
+- {class: AucMonitor, compute_interval: 86400, name: day_auc, target: ctr.tmp_2, target_idx: 0}
 outputs:
 - name: ctr.tmp_2
   shape: [-1, 1]

paddle/fluid/train/custom_trainer/feed/scripts/update.py

@@ -38,8 +38,12 @@ def inference():
         { "class": "DenseInputAccessor", "input": "vars", "table_id": 3, "need_gradient": True, "async_pull": True},
         { "class": "LabelInputAccessor", "input": "labels"}
     ]
+    monitors = [
+        { "name": "epoch_auc", "class": "AucMonitor", "target": ctr_output, "compute_interval": 600 },
+        { "name": "day_auc", "class": "AucMonitor", "target": ctr_output, "compute_interval": 86400 }
+    ]
 
-    return accessors, [sparse_cvm], [cvm_input], [ctr_output]
+    return accessors, [sparse_cvm], [cvm_input], [ctr_output], monitors
 
 def loss_function(ctr_output):
     """