fix bugs (#4549)

PaddleRec/multi-task/MMoE/README.md

@@ -50,7 +50,7 @@ train_path="data/census-income.data"
 test_path="data/census-income.test"
 train_data_path="train_data/"
 test_data_path="test_data/"
-
+pip install -r requirements.txt
 wget -P data/ https://archive.ics.uci.edu/ml/machine-learning-databases/census-income-mld/census.tar.gz
 tar -zxvf data/census.tar.gz -C data/
 

PaddleRec/multi-task/MMoE/args.py

@@ -30,9 +30,9 @@ def parse_args():
     parser.add_argument("--epochs", type=int, default=400, help="epochs")
     parser.add_argument("--batch_size", type=int, default=32, help="batch_size")
     parser.add_argument('--use_gpu', type=int, default=0, help='whether using gpu')
-    parser.add_argument('--model_dir',type=str, default='./model_dir', help="model_dir")
-    parser.add_argument('--train_data_path',type=str, default='./train_data', help="train_data_path")
-    parser.add_argument('--test_data_path',type=str, default='./test_data', help="test_data_path")
+    parser.add_argument('--model_dir',type=str, default='model_dir', help="model_dir")
+    parser.add_argument('--train_data_path',type=str, default='train_data', help="train_data_path")
+    parser.add_argument('--test_data_path',type=str, default='test_data', help="test_data_path")
     args = parser.parse_args()
     return args
     
@@ -43,6 +43,5 @@ def data_preparation_args():
 
     parser.add_argument('--train_data_path',type=str, default='', help="train_data_path")
     parser.add_argument('--test_data_path',type=str, default='', help="test_data_path")
-    parser.add_argument('--validation_data_path',type=str, default='', help="validation_data_path")
     args = parser.parse_args()
     return args

PaddleRec/multi-task/MMoE/create_data.sh

@@ -5,7 +5,6 @@ train_path="data/census-income.data"
 test_path="data/census-income.test"
 train_data_path="train_data/"
 test_data_path="test_data/"
-
 pip install -r requirements.txt
 
 wget -P data/ https://archive.ics.uci.edu/ml/machine-learning-databases/census-income-mld/census.tar.gz

PaddleRec/multi-task/MMoE/data_preparation.py

@@ -18,8 +18,7 @@ def fun2(x):
         return 0
 
 
-def data_preparation(train_path, test_path, train_data_path, test_data_path,
-                     validation_data_path):
+def data_preparation(train_path, test_path, train_data_path, test_data_path):
     # The column names are from
     # https://www2.1010data.com/documentationcenter/prod/Tutorials/MachineLearningExamples/CensusIncomeDataSet.html
     column_names = [
@@ -102,4 +101,4 @@ def data_preparation(train_path, test_path, train_data_path, test_data_path,
 
 args = data_preparation_args()
 data_preparation(args.train_path, args.test_path, args.train_data_path,
-                 args.test_data_path, args.validation_data_path)
+                 args.test_data_path)

PaddleRec/multi-task/MMoE/mmoe_train.py

@@ -69,8 +69,11 @@ def MMOE(feature_size=499,expert_num=8, gate_num=2, expert_size=16, tower_size=8
     label_income_1 = fluid.layers.slice(label_income, axes=[1], starts=[1], ends=[2])
     label_marital_1 = fluid.layers.slice(label_marital, axes=[1], starts=[1], ends=[2])
     
-    auc_income, batch_auc_1, auc_states_1  = fluid.layers.auc(input=output_layers[0], label=fluid.layers.cast(x=label_income_1, dtype='int64'))
-    auc_marital, batch_auc_2, auc_states_2 = fluid.layers.auc(input=output_layers[1], label=fluid.layers.cast(x=label_marital_1, dtype='int64'))
+    pred_income = fluid.layers.clip(output_layers[0], min=1e-10, max=1.0 - 1e-10)
+    pred_marital = fluid.layers.clip(output_layers[1], min=1e-10, max=1.0 - 1e-10)
+    
+    auc_income, batch_auc_1, auc_states_1  = fluid.layers.auc(input=pred_income, label=fluid.layers.cast(x=label_income_1, dtype='int64'))
+    auc_marital, batch_auc_2, auc_states_2 = fluid.layers.auc(input=pred_marital, label=fluid.layers.cast(x=label_marital_1, dtype='int64'))
     
     avg_cost_income = fluid.layers.mean(x=cost_income)
     avg_cost_marital = fluid.layers.mean(x=cost_marital)
@@ -116,7 +119,6 @@ test_loader = fluid.io.DataLoader.from_generator(feed_list=data_list, capacity=b
 test_loader.set_sample_list_generator(test_reader, places=place)
 auc_income_list = []
 auc_marital_list = []
-mmoe_res_file = open('mmoe_res.txt', 'w',encoding='utf-8')
 for epoch in range(epochs):
     for var in auc_states_1:  # reset auc states
         set_zero(var.name,place=place)

PaddleRec/multi-task/Share_bottom/README.md

@@ -50,7 +50,7 @@ train_path="data/census-income.data"
 test_path="data/census-income.test"
 train_data_path="train_data/"
 test_data_path="test_data/"
-
+pip install -r requirements.txt
 wget -P data/ https://archive.ics.uci.edu/ml/machine-learning-databases/census-income-mld/census.tar.gz
 tar -zxvf data/census.tar.gz -C data/
 

PaddleRec/multi-task/Share_bottom/args.py

@@ -40,15 +40,7 @@ def data_preparation_args():
     parser = argparse.ArgumentParser(description=__doc__)
     parser.add_argument("--train_path", type=str, default='', help="train_path")
     parser.add_argument("--test_path", type=str, default='', help="test_path")
-
-    parser.add_argument(
-        '--train_data_path', type=str, default='train_data', help="train_data_path")
-    parser.add_argument(
-        '--test_data_path', type=str, default='test_data', help="test_data_path")
-    parser.add_argument(
-        '--validation_data_path',
-        type=str,
-        default='',
-        help="validation_data_path")
+    parser.add_argument('--train_data_path', type=str, default='train_data', help="train_data_path")
+    parser.add_argument('--test_data_path', type=str, default='test_data', help="test_data_path")
     args = parser.parse_args()
     return args

PaddleRec/multi-task/Share_bottom/data_preparation.py

@@ -18,8 +18,7 @@ def fun2(x):
         return 0
 
 
-def data_preparation(train_path, test_path, train_data_path, test_data_path,
-                     validation_data_path):
+def data_preparation(train_path, test_path, train_data_path, test_data_path):
     # The column names are from
     # https://www2.1010data.com/documentationcenter/prod/Tutorials/MachineLearningExamples/CensusIncomeDataSet.html
     column_names = [
@@ -101,5 +100,4 @@ def data_preparation(train_path, test_path, train_data_path, test_data_path,
 
 
 args = data_preparation_args()
-data_preparation(args.train_path, args.test_path, args.train_data_path,
-                 args.test_data_path, args.validation_data_path)
+data_preparation(args.train_path, args.test_path, args.train_data_path,args.test_data_path)

PaddleRec/multi-task/Share_bottom/share_bottom.py

@@ -56,8 +56,11 @@ def share_bottom(feature_size=499,bottom_size=117,tower_nums=2,tower_size=8):
     label_income_1 = fluid.layers.slice(label_income, axes=[1], starts=[1], ends=[2])
     label_marital_1 = fluid.layers.slice(label_marital, axes=[1], starts=[1], ends=[2])
     
-    auc_income, batch_auc_1, auc_states_1  = fluid.layers.auc(input=output_layers[0], label=fluid.layers.cast(x=label_income_1, dtype='int64'))
-    auc_marital, batch_auc_2, auc_states_2 = fluid.layers.auc(input=output_layers[1], label=fluid.layers.cast(x=label_marital_1, dtype='int64'))
+    pred_income = fluid.layers.clip(output_layers[0], min=1e-10, max=1.0 - 1e-10)
+    pred_marital = fluid.layers.clip(output_layers[1], min=1e-10, max=1.0 - 1e-10)
+    
+    auc_income, batch_auc_1, auc_states_1  = fluid.layers.auc(input=pred_income, label=fluid.layers.cast(x=label_income_1, dtype='int64'))
+    auc_marital, batch_auc_2, auc_states_2 = fluid.layers.auc(input=pred_marital, label=fluid.layers.cast(x=label_marital_1, dtype='int64'))
     
     avg_cost_income = fluid.layers.mean(x=cost_income)
     avg_cost_marital = fluid.layers.mean(x=cost_marital)

PaddleRec/multi-task/Share_bottom/train_gpu.sh

-python share_bottom.py --use_gpu 1 \
+CUDA_VISIBLE_DEVICES=0 python share_bottom.py --use_gpu 1 \
                        --epochs 100 \
+                       --train_data_path 'train_data' \
+                       --test_data_path 'test_data' \
                        --train_data_path '.train_data' \
                        --test_data_path '.test_data' \
                        --model_dir 'model_dir' \