structure rebuild

fleetrec/core/trainer/details/local_engine.py

@@ -38,7 +38,7 @@ def start_procs(args, yaml):
     user_endpoints_ips = [x.split(":")[0] for x in user_endpoints.split(",")]
     user_endpoints_port = [x.split(":")[1] for x in user_endpoints.split(",")]
 
-    factory = "fleetrec.trainer.factory"
+    factory = "fleetrec.core.factory"
     cmd = [sys.executable, "-u", "-m", factory, yaml]
 
     for i in range(server_num):
@@ -91,11 +91,11 @@ def start_procs(args, yaml):
         procs[i].terminate()
     print("all parameter server are killed", file=sys.stderr)
 
-class Launch():
+
+class Launch:
     def __init__(self, envs, trainer):
         self.envs = envs
         self.trainer = trainer
 
     def run(self):
         start_procs(self.envs, self.trainer)
-

fleetrec/examples/built_in/cluster_training_local.yaml

+
+trainer:  "LocalClusterTraining"
+
+pserver_num: 2
+trainer_num: 2
+start_port: 36001
+log_dirname: "logs"
+
+strategy:
+  mode: "async"

fleetrec/examples/built_in/cluster_training_mpi.yaml

+
+trainer:  "MPIClusterTraining"
+
+pserver_num: 2
+trainer_num: 2
+start_port: 36001
+log_dirname: "logs"
+
+strategy:
+  mode: "async"

fleetrec/examples/built_in/cluster_training_user_define.yaml

+trainer:  "UserDefineTrainer"
+location: "/root/FleetRec/fleetrec/examples/user_define_trainer.py"

fleetrec/examples/ctr-dnn_train_single.yaml → fleetrec/examples/built_in/ctr-dnn_train.yaml

@@ -15,9 +15,10 @@
 train:
   threads: 12
   epochs: 10
-  trainer:  "SingleTraining"
+  trainer:  "single_training.yaml"
 
   reader:
+    mode: "dataset"
     batch_size: 2
     class: "fleetrec.models.ctr_dnn.data_generator"
     train_data_path: "/root/FleetRec/fleetrec/models/ctr_dnn/data/train/"
@@ -40,6 +41,8 @@ train:
     inference:
       dirname: "models_for_inference"
       epoch_interval: 4
+      feed_varnames: ["C1", "C2", "C3"]
+      fetch_varnames: "predict"
       save_last: True
 
 evaluate:

fleetrec/examples/train.py → fleetrec/examples/built_in/run.py

@@ -14,7 +14,7 @@
 
 import os
 
-from fleetrec.trainer.factory import TrainerFactory
+from fleetrec.core.factory import TrainerFactory
 
 if __name__ == "__main__":
 

fleetrec/examples/built_in/single_training.yaml

+
+trainer:  "SingleTraining"
\ No newline at end of file

fleetrec/examples/user_define_trainer.py → fleetrec/examples/built_in/user_define_trainer.py

@@ -21,6 +21,7 @@ from fleetrec.utils import envs
 class UserDefineTrainer(TranspileTrainer):
     def __init__(self, config=None):
         TranspileTrainer.__init__(self, config)
+        print("this is a demo about how to use user define trainer in fleet-rec")
 
     def processor_register(self):
         self.regist_context_processor('uninit', self.instance)

fleetrec/examples/ctr-dnn_train_cluster.yaml

-# Copyright (c) 2020 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.
-
-train:
-  threads: 12
-  epochs: 10
-  trainer:  "ClusterTraining"
-  container: "local"
-
-  pserver_num: 2
-  trainer_num: 2
-  start_port: 36001
-  log_dirname: "logs"
-
-  strategy:
-    mode: "async"
-
-  reader:
-    mode: "dataset"
-    batch_size: 2
-    pipe_command: "python /paddle/eleps/fleetrec/models/ctr_dnn/dataset.py"
-    train_data_path: "/paddle/eleps/fleetrec/models/ctr_dnn/data/train"
-
-  model:
-    models: "fleetrec.models.ctr_dnn.model"
-    hyper_parameters:
-      sparse_inputs_slots: 27
-      sparse_feature_number: 1000001
-      sparse_feature_dim: 8
-      dense_input_dim: 13
-      fc_sizes: [512, 256, 128, 32]
-      learning_rate: 0.001
-
-  save:
-    increment:
-      dirname: "models_for_increment"
-      epoch_interval: 2
-      save_last: True
-    inference:
-      dirname: "models_for_inference"
-      epoch_interval: 4
-      feed_varnames: ["C1", "C2", "C3"]
-      fetch_varnames: "predict"
-      save_last: True
-
-evaluate:
-  batch_size: 32
-  train_thread_num: 12
-  reader: "reader.py"
-
-

fleetrec/examples/ctr-dnn_train_userdefine.yaml

-# Copyright (c) 2020 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.
-
-
-train:
-  threads: 12
-  epochs: 10
-  trainer:  "UserDefineTrainer"
-  location: "/root/FleetRec/fleetrec/examples/user_define_trainer.py"
-
-  reader:
-    batch_size: 2
-    class: "fleetrec.models.ctr_dnn.data_generator"
-    train_data_path: "/root/FleetRec/fleetrec/models/ctr_dnn/data/train/"
-
-  model:
-    models: "fleetrec.models.ctr_dnn.model"
-    hyper_parameters:
-      sparse_inputs_slots: 27
-      sparse_feature_number: 1000001
-      sparse_feature_dim: 8
-      dense_input_dim: 13
-      fc_sizes: [512, 256, 128, 32] 
-      learning_rate: 0.001
-
-  save:
-    increment:
-      dirname: "models_for_increment"
-      epoch_interval: 2
-      save_last: True
-    inference:
-      dirname: "models_for_inference"
-      epoch_interval: 4
-      save_last: True
-
-evaluate:
-  batch_size: 32
-  train_thread_num: 12
-  reader: "reader.py"
-
-