fix comments

deepes/benchmark/cartpole_config.prototxt

@@ -11,5 +11,5 @@ optimizer {
   epsilon: 1e-08
 }
 async_es {
-  model_iter_id: 0
+  model_iter_id: 99
 }

deepes/include/paddle/async_es_agent.h

@@ -12,8 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#ifndef _ASYNC_ES_AGENT_H
-#define _ASYNC_ES_AGENT_H
+#ifndef ASYNC_ES_AGENT_H
+#define ASYNC_ES_AGENT_H
 
 #include "es_agent.h"
 #include <map>
@@ -49,7 +49,10 @@ class AsyncESAgent: public ESAgent {
     std::shared_ptr<AsyncESAgent> clone();
 
     /**
-     * @brief: Clone an agent for sampling.
+     * @brief: update parameters given data collected during evaluation.
+     * @args:
+     *   noisy_info: sampling information returned by add_noise function.
+     *   noisy_reward: evaluation rewards.
      */
     bool update(
         std::vector<SamplingInfo>& noisy_info,

deepes/include/paddle/es_agent.h

@@ -21,15 +21,13 @@
 #include "gaussian_sampling.h"
 #include "deepes.pb.h"
 #include <vector>
-using namespace paddle::lite_api;
 
+using namespace paddle::lite_api;
 
 namespace DeepES {
 
 int64_t ShapeProduction(const shape_t& shape);
 
-typedef paddle::lite_api::PaddlePredictor PaddlePredictor;
-
 /**
  * @brief DeepES agent with PaddleLite as backend.
  * Users mainly focus on the following functions:

deepes/include/utils.h

@@ -40,7 +40,7 @@ bool load_proto_conf(const std::string& config_file, T& proto_config) {
   bool success = true;
   std::ifstream fin(config_file);
   if (!fin || fin.fail()) {
-    LOG(FATAL) << "open prototxt config failed: " << config_file;
+    LOG(ERROR) << "open prototxt config failed: " << config_file;
     success = false;
   } else {
     fin.seekg(0, std::ios::end);
@@ -52,7 +52,7 @@ bool load_proto_conf(const std::string& config_file, T& proto_config) {
 
     std::string proto_str(file_content_buffer, file_size);
     if (!google::protobuf::TextFormat::ParseFromString(proto_str, &proto_config)) {
-      LOG(FATAL) << "Failed to load config: " << config_file;
+      LOG(ERROR) << "Failed to load config: " << config_file;
       success = false;
     }
     delete[] file_content_buffer;
@@ -66,7 +66,7 @@ bool save_proto_conf(const std::string& config_file, T&proto_config) {
   bool success = true;
   std::ofstream ofs(config_file, std::ofstream::out);
   if (!ofs || ofs.fail()) {
-    LOG(FATAL) << "open prototxt config failed: " << config_file;
+    LOG(ERROR) << "open prototxt config failed: " << config_file;
     success = false;
   } else {
     std::string config_str;
@@ -76,6 +76,7 @@ bool save_proto_conf(const std::string& config_file, T&proto_config) {
     }
     ofs << config_str;
   }
+  return success;
 }
 
 std::vector<std::string> list_all_model_dirs(std::string path);

deepes/scripts/build.sh

@@ -32,8 +32,6 @@ else
   exit 0
 fi
 
-#export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
-
 #----------------protobuf-------------#
 cp ./src/proto/deepes.proto ./
 protoc deepes.proto --cpp_out ./

deepes/src/paddle/async_es_agent.cc

@@ -30,7 +30,7 @@ AsyncESAgent::~AsyncESAgent() {
 bool AsyncESAgent::_save() {
   bool success = true;
   if (_is_sampling_agent) {
-    LOG(ERROR) << "[DeepES] Original AsyncESAgent cannot call add_noise function, please use cloned AsyncESAgent.";
+    LOG(ERROR) << "[DeepES] Original AsyncESAgent cannot call `save`.Please use cloned AsyncESAgent.";
     success = false;
     return success;
   }
@@ -49,7 +49,7 @@ bool AsyncESAgent::_save() {
   model_name = "model_iter_id-"+ std::to_string(model_iter_id);
   std::string model_path = _config->async_es().model_warehouse() + "/" + model_name;
   LOG(INFO) << "[save]model_path: " << model_path;
-  _predictor->SaveOptimizedModel(model_path, LiteModelType::kProtobuf);
+  _predictor->SaveOptimizedModel(model_path, paddle::lite_api::LiteModelType::kProtobuf);
   // save config
   auto async_es = _config->mutable_async_es();
   async_es->set_model_iter_id(model_iter_id);
@@ -93,15 +93,17 @@ bool AsyncESAgent::_compute_model_diff() {
     std::shared_ptr<PaddlePredictor> old_predictor = kv.second;
     float* diff = new float[_param_size];
     memset(diff, 0, _param_size * sizeof(float));
-    for (std::string param_name: _param_names) {
+    int offset = 0;
+    for (const std::string& param_name: _param_names) {
       auto des_tensor = old_predictor->GetTensor(param_name);
       auto src_tensor = _predictor->GetTensor(param_name);
       const float* des_data = des_tensor->data<float>();
       const float* src_data = src_tensor->data<float>();
       int64_t tensor_size = ShapeProduction(src_tensor->shape());
       for (int i = 0; i < tensor_size; ++i) {
-        diff[i] = des_data[i] - src_data[i];
+        diff[i + offset] = des_data[i] - src_data[i];
       }
+      offset += tensor_size;
     }
     _param_delta[model_iter_id] = diff;
   }
@@ -206,6 +208,7 @@ bool AsyncESAgent::update(
     float reward = noisy_rewards[i];
     int model_iter_id = noisy_info[i].model_iter_id();
     bool success = _sampling_method->resampling(key, _noise, _param_size);
+    CHECK(success) << "[DeepES] resampling error occurs at sample: " << i;
     float* delta = _param_delta[model_iter_id];
     // compute neg_gradients
     if (model_iter_id == current_model_iter_id) {

deepes/src/paddle/es_agent.cc

@@ -17,9 +17,6 @@
 
 namespace DeepES {
 
-typedef paddle::lite_api::Tensor Tensor;
-typedef paddle::lite_api::shape_t shape_t;
-
 int64_t ShapeProduction(const shape_t& shape) {
   int64_t res = 1;
   for (auto i : shape) res *= i;