update 2.0 api (#4921)

* update api 1.8

* fix paddlerec readme

* fix readme

* update 2.0 api

* fix, test=develop

PaddleRec/ctr/deepfm_dygraph/README.md

-
 # DeepFM动态图
 
 以下是本例的简要目录结构及说明：
@@ -65,9 +64,9 @@ CUDA_VISIBLE_DEVICES=0 python infer.py --checkpoint=models/epoch_0
 
 ## 效果
 ```text
-test auc of epoch 0 is 0.802877
+test auc of epoch 0 is 0.78+
 ```
 
-第一轮数据训练结束后，test auc为0.802877。
+第一轮数据训练结束后，test auc为0.78+。
 
 继续训练模型易出现过拟合现象，可以通过评估模型选择效果最好的模型作为最终训练结果。

PaddleRec/ctr/deepfm_dygraph/infer.py

@@ -3,8 +3,7 @@ from __future__ import print_function
 import os
 
 import numpy as np
-import paddle.fluid as fluid
-from paddle.fluid.dygraph.base import to_variable
+import paddle
 import logging
 import time
 
@@ -19,72 +18,74 @@ logger = logging.getLogger(__name__)
 
 def infer(args):
     if args.use_gpu:
-        place = fluid.CUDAPlace(0)
+        place = paddle.CUDAPlace(0)
     else:
-        place = fluid.CPUPlace()
-    with fluid.dygraph.guard(place):
-        deepfm = DeepFM(args)
-
-        test_filelist = [
-            os.path.join(args.test_data_dir, x)
-            for x in os.listdir(args.test_data_dir)
-        ]
-
-        test_reader = data_reader.data_reader(
-            args.batch_size, test_filelist, args.feat_dict, data_type="test")
-
-        # load model
-        if args.checkpoint:
-            model_dict, optimizer_dict = fluid.dygraph.load_dygraph(
-                args.checkpoint)
-            deepfm.set_dict(model_dict)
-            logger.info("load model {} finished.".format(args.checkpoint))
-        else:
-            logger.error("no model to load!")
-            logger.error("please set model to load in --checkpoint first.")
-            exit(1)
-
-        def eval():
-            deepfm.eval()
-            logger.info("start eval model.")
-            total_step = 0
-            batch_begin = time.time()
-            auc_metric_test = fluid.metrics.Auc("ROC")
-            for data in test_reader():
-                total_step += 1
-                raw_feat_idx, raw_feat_value, label = zip(*data)
-                raw_feat_idx = np.array(raw_feat_idx, dtype=np.int64)
-                raw_feat_value = np.array(raw_feat_value, dtype=np.float32)
-                label = np.array(label, dtype=np.int64)
-                raw_feat_idx, raw_feat_value, label = [
-                    to_variable(i)
-                    for i in [raw_feat_idx, raw_feat_value, label]
-                ]
-
-                predict = deepfm(raw_feat_idx, raw_feat_value, label)
-
-                # for auc
-                predict_2d = fluid.layers.concat([1 - predict, predict], 1)
-                auc_metric_test.update(
-                    preds=predict_2d.numpy(), labels=label.numpy())
-
-                if total_step > 0 and total_step % 100 == 0:
-                    logger.info(
-                        "TEST --> batch: {} auc: {:.6f} speed: {:.2f} ins/s".
-                        format(total_step,
-                               auc_metric_test.eval(), 100 * args.batch_size / (
-                                   time.time() - batch_begin)))
-                    batch_begin = time.time()
-
-            logger.info("test auc is %.6f" % auc_metric_test.eval())
-
-        begin = time.time()
-        eval()
-        logger.info("test finished, cost %f s" % (time.time() - begin))
+        place = paddle.CPUPlace()
+    paddle.disable_static(place)
+    deepfm = DeepFM(args)
+
+    test_filelist = [
+        os.path.join(args.test_data_dir, x)
+        for x in os.listdir(args.test_data_dir)
+    ]
+
+    test_reader = data_reader.data_reader(
+        args.batch_size, test_filelist, args.feat_dict, data_type="test")
+
+    # load model
+    if args.checkpoint:
+        model_dict, optimizer_dict = paddle.fluid.dygraph.load_dygraph(
+            args.checkpoint)
+        deepfm.set_dict(model_dict)
+        logger.info("load model {} finished.".format(args.checkpoint))
+    else:
+        logger.error("no model to load!")
+        logger.error("please set model to load in --checkpoint first.")
+        exit(1)
+
+    def eval():
+        deepfm.eval()
+        logger.info("start eval model.")
+        total_step = 0
+        batch_begin = time.time()
+        auc_metric_test = paddle.fluid.metrics.Auc("ROC")
+        for data in test_reader():
+            total_step += 1
+            raw_feat_idx, raw_feat_value, label = zip(*data)
+            raw_feat_idx = np.array(raw_feat_idx, dtype=np.int64)
+            raw_feat_value = np.array(raw_feat_value, dtype=np.float32)
+            label = np.array(label, dtype=np.int64)
+            raw_feat_idx, raw_feat_value, label = [
+                paddle.to_tensor(
+                    data=i, dtype=None, place=None, stop_gradient=True)
+                for i in [raw_feat_idx, raw_feat_value, label]
+            ]
+
+            predict = deepfm(raw_feat_idx, raw_feat_value, label)
+
+            # for auc
+            predict_2d = paddle.concat(x=[1 - predict, predict], axis=1)
+            auc_metric_test.update(
+                preds=predict_2d.numpy(), labels=label.numpy())
+
+            if total_step > 0 and total_step % 100 == 0:
+                logger.info(
+                    "TEST --> batch: {} auc: {:.6f} speed: {:.2f} ins/s".format(
+                        total_step,
+                        auc_metric_test.eval(), 100 * args.batch_size / (
+                            time.time() - batch_begin)))
+                batch_begin = time.time()
+
+        logger.info("test auc is %.6f" % auc_metric_test.eval())
+
+    begin = time.time()
+    eval()
+    logger.info("test finished, cost %f s" % (time.time() - begin))
+    paddle.enable_static()
 
 
 if __name__ == '__main__':
-    
+
     args = utils.parse_args()
     utils.print_arguments(args)
 

PaddleRec/ctr/deepfm_dygraph/network.py

@@ -2,7 +2,6 @@ import math
 import paddle
 
 
-
 class DeepFM(paddle.nn.Layer):
     def __init__(self, args):
         super(DeepFM, self).__init__()
@@ -13,8 +12,8 @@ class DeepFM(paddle.nn.Layer):
         self.dnn = DNN(args)
 
     def forward(self, raw_feat_idx, raw_feat_value, label):
-        feat_idx = paddle.fluid.layers.reshape(raw_feat_idx,
-                                        [-1, 1])  # (None * num_field) * 1
+        feat_idx = paddle.fluid.layers.reshape(
+            raw_feat_idx, [-1, 1])  # (None * num_field) * 1
         feat_value = paddle.fluid.layers.reshape(
             raw_feat_value,
             [-1, self.args.num_field, 1])  # None * num_field * 1
@@ -23,7 +22,8 @@ class DeepFM(paddle.nn.Layer):
                                                                  feat_value)
         y_dnn = self.dnn(feat_embeddings)
 
-        predict = paddle.nn.functional.sigmoid(y_first_order + y_second_order + y_dnn)
+        predict = paddle.nn.functional.sigmoid(y_first_order + y_second_order +
+                                               y_dnn)
 
         return predict
 
@@ -39,7 +39,7 @@ class FM(paddle.nn.Layer):
             padding_idx=0,
             param_attr=paddle.ParamAttr(
                 initializer=paddle.nn.initializer.TruncatedNormal(
-                    loc=0.0, scale=self.init_value_),
+                    mean=0.0, std=self.init_value_),
                 regularizer=paddle.fluid.regularizer.L1DecayRegularizer(
                     self.args.reg)))
         self.embedding = paddle.fluid.dygraph.nn.Embedding(
@@ -48,8 +48,8 @@ class FM(paddle.nn.Layer):
             padding_idx=0,
             param_attr=paddle.ParamAttr(
                 initializer=paddle.nn.initializer.TruncatedNormal(
-                    loc=0.0,
-                    scale=self.init_value_ /
+                    mean=0.0,
+                    std=self.init_value_ /
                     math.sqrt(float(self.args.embedding_size)))))
 
     def forward(self, feat_idx, feat_value):
@@ -69,16 +69,16 @@ class FM(paddle.nn.Layer):
         feat_embeddings = feat_embeddings * feat_value  # None * num_field * embedding_size
 
         # sum_square part
-        summed_features_emb = paddle.reduce_sum(
-            feat_embeddings, 1)  # None * embedding_size
+        summed_features_emb = paddle.reduce_sum(feat_embeddings,
+                                                1)  # None * embedding_size
         summed_features_emb_square = paddle.square(
             summed_features_emb)  # None * embedding_size
 
         # square_sum part
         squared_features_emb = paddle.square(
             feat_embeddings)  # None * num_field * embedding_size
-        squared_sum_features_emb = paddle.reduce_sum(
-            squared_features_emb, 1)  # None * embedding_size
+        squared_sum_features_emb = paddle.reduce_sum(squared_features_emb,
+                                                     1)  # None * embedding_size
 
         y_second_order = 0.5 * paddle.reduce_sum(
             summed_features_emb_square - squared_sum_features_emb,
@@ -93,7 +93,8 @@ class DNN(paddle.nn.Layer):
         super(DNN, self).__init__()
         self.args = args
         self.init_value_ = 0.1
-        sizes = [self.args.num_field * self.args.embedding_size] + self.args.layer_sizes + [1]
+        sizes = [self.args.num_field * self.args.embedding_size
+                 ] + self.args.layer_sizes + [1]
         acts = [self.args.act
                 for _ in range(len(self.args.layer_sizes))] + [None]
         w_scales = [
@@ -107,10 +108,10 @@ class DNN(paddle.nn.Layer):
                 out_features=sizes[i + 1],
                 weight_attr=paddle.ParamAttr(
                     initializer=paddle.nn.initializer.TruncatedNormal(
-                        loc=0.0, scale=w_scales[i])),
+                        mean=0.0, std=w_scales[i])),
                 bias_attr=paddle.ParamAttr(
                     initializer=paddle.nn.initializer.TruncatedNormal(
-                        loc=0.0, scale=self.init_value_)))
+                        mean=0.0, std=self.init_value_)))
             #linear = getattr(paddle.nn.functional, acts[i])(linear) if acts[i] else linear
             if acts[i] == 'relu':
                 act = paddle.nn.ReLU()