update & support

models/rank/dnn/benchmark.yaml

@@ -33,20 +33,19 @@ runner:
   use_gpu: 0
 
   model_path: "static_model.py"
-  reader_type: "QueueDataset"  # DataLoader / QueueDataset
-  pipe_command: "python static_reader.py"
+  reader_type: "RecDataset"  # DataLoader / QueueDataset / RecDataset
+  pipe_command: "python benchmark_reader.py"
   dataset_debug: False
   split_file_list: False
 
-  train_batch_size: 1000
-  train_data_dir: "train_data"
-  train_reader_path: "static_reader.py"
+  train_batch_size: 10
+  train_data_dir: "data/sample_data/train"
+  train_reader_path: "reader"
   model_save_path: "model"
-  warmup_model_path: ""
 
   infer_batch_size: 2
   test_data_dir: "test_data"
-  infer_reader_path: "static_reader.py"
+  infer_reader_path: "benchmark_reader"
   infer_load_path: "model"
   infer_start_epoch: 0
   infer_end_epoch: 1

models/rank/dnn/static_reader.py → models/rank/dnn/benchmark_reader.py

@@ -16,7 +16,7 @@
 import six
 import os
 import copy
-import paddle.fluid.incubate.data_generator as dg
+import paddle.distributed.fleet as fleet
 import logging
 
 cont_min_ = [0, -3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
@@ -31,7 +31,7 @@
 logger = logging.getLogger(__name__)
 
 
-class Reader(dg.MultiSlotDataGenerator):
+class Reader(fleet.MultiSlotDataGenerator):
     def init(self, config):
         self.config = config
 
@@ -65,7 +65,7 @@ def reader():
         return reader
 
     def dataloader(self, file_list):
-        "DataLoader Generator"
+        "DataLoader Pyreader Generator"
 
         def reader():
             for file in file_list:

models/recall/word2vec/benchmark.yaml

@@ -11,8 +11,6 @@
 # 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.
-workspace: "./"
-
 hyper_parameters:
   optimizer:
     learning_rate: 1.0
@@ -24,27 +22,28 @@ hyper_parameters:
   neg_num: 5
   window_size: 5
 
-static_benchmark:
-  epochs: 5
-  batch_size: 100
-  print_period: 1000
+runner:
+  epochs: 20
+  print_interval: 10
 
-  sync_mode: "async"  # sync / async /geo / heter
+  sync_mode: "geo"  # sync / async /geo / heter
   geo_step: 400
   thread_num: 16
-  use_cuda: 0
+  use_gpu: 0
+
+  model_path: "static_model.py"
+  train_reader_path: "static_reader.py"
   reader_type: "QueueDataset"  # DataLoader / QueueDataset
   dataset_debug: False
-
-  model_path: "{workspace}/static_model.py"
-  reader_path: "{workspace}/static_reader.py"
-  train_data_path: "{workspace}/train_data"
-  test_data_path: "{workspace}/test_data"
-  save_model_path: "{workspace}/model"
-  split_file_list: False
-  pipe_command: "python {workspace}/static_reader.py"
+  pipe_command: "python static_reader.py"
   example_count_method: "word"
-
+  dataset_debug: False
+  split_file_list: False
+
+  train_batch_size: 100
+  train_data_dir: "train_data"
+  train_reader_path: "static_ps_reader.py"
+  model_save_path: "model"
+
   word_count_dict_path: "{workspace}/dict/word_count_dict.txt"
   word_id_dict_path: "{workspace}/dict/word_id_dict.txt"
-