Add aitm model

models/rank/aitm/__init__.py

@@ -0,0 +1,13 @@
+# 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.

models/rank/aitm/config.yaml

@@ -0,0 +1,43 @@
+# 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.
+
+# global settings
+
+runner:
+  train_data_dir: "./data/sample_data/train"
+  train_reader_path: "reader" # importlib format
+  use_gpu: True
+  train_batch_size: 10
+  epochs: 2
+  print_interval: 10
+  #model_init_path: "output_model/0" # init model
+  model_save_path: "output_model_aitm_test/"
+  test_data_dir: "./data/sample_data/test"
+  infer_reader_path: "reader" # importlib format
+  infer_batch_size: 10
+  infer_load_path: "output_model_aitm_test/"
+  infer_start_epoch: 0
+  infer_end_epoch: 2
+
+# hyper parameters of user-defined network
+hyper_parameters:
+  # optimizer config
+  optimizer:
+    class: Adam
+    learning_rate: 0.0001
+  # user-defined <key, value> pairs
+  embedding_size: 5
+  dims: [128, 64, 32]
+  drop_prob: [0.1, 0.3, 0.3]
+  feature_vocabulary: [['101', 238635], ['121', 98], ['122', 14], ['124', 3], ['125', 8], ['126', 4], ['127', 4], ['128', 3], ['129', 5], ['205', 467298], ['206', 6929], ['207', 263942], ['216', 106399], ['508', 5888], ['509', 104830], ['702', 51878], ['853', 37148], ['301', 4]]

models/rank/aitm/config_bigdata.yaml

@@ -0,0 +1,43 @@
+# 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.
+
+# global settings 
+
+runner:
+  train_data_dir: "./data/whole_data/train"
+  train_reader_path: "reader" # importlib format
+  use_gpu: True
+  train_batch_size: 2000
+  epochs: 6
+  print_interval: 500
+  #model_init_path: "output_model/0" # init model
+  model_save_path: "output_model_aitm/"
+  test_data_dir: "./data/whole_data/test"
+  infer_reader_path: "reader" # importlib format
+  infer_batch_size: 2000
+  infer_load_path: "output_model_aitm/"
+  infer_start_epoch: 0
+  infer_end_epoch: 6
+
+# hyper parameters of user-defined network
+hyper_parameters:
+  # optimizer config
+  optimizer:
+    class: Adam
+    learning_rate: 0.0002
+  # user-defined <key, value> pairs
+  embedding_size: 5
+  dims: [128, 64, 32]
+  drop_prob: [0.1, 0.3, 0.3]
+  feature_vocabulary: [['101', 238635], ['121', 98], ['122', 14], ['124', 3], ['125', 8], ['126', 4], ['127', 4], ['128', 3], ['129', 5], ['205', 467298], ['206', 6929], ['207', 263942], ['216', 106399], ['508', 5888], ['509', 104830], ['702', 51878], ['853', 37148], ['301', 4]]

models/rank/aitm/data/sample_data/test/test.csv

@@ -0,0 +1,50 @@
+0,0,180467,4,1,2,3,1,1,2,1,157266,5673,164308,73943,753,0,0,32454,2
+0,0,180467,4,1,2,3,1,1,2,1,426254,5673,216630,0,753,82564,0,24700,2
+1,0,180467,4,1,2,3,1,1,2,1,172023,2114,63355,0,2261,80294,0,0,2
+0,0,225145,39,3,2,4,1,1,2,2,70837,1668,46734,61858,1952,0,0,36223,2
+0,0,225145,39,3,2,4,1,1,2,2,45118,6423,137020,39048,0,0,0,0,2
+0,0,225145,39,3,2,4,1,1,2,2,0,3384,55310,13430,287,0,0,35494,2
+0,1,225145,39,3,2,4,1,1,2,2,0,2015,0,81649,467,0,10904,0,2
+0,0,225145,39,3,2,4,1,1,2,2,27259,4109,88551,49425,4417,0,39516,0,2
+1,0,225145,39,3,2,4,1,1,2,2,297113,4109,219686,5769,4417,0,0,0,2
+0,0,225145,39,3,2,4,1,1,2,2,151749,3228,148375,89123,2599,47614,10053,0,2
+0,0,225145,39,3,2,4,1,1,2,2,0,1081,47185,0,3865,0,0,0,2
+0,0,225145,39,3,2,4,1,1,2,2,0,2015,0,81649,467,0,10904,0,2
+0,0,225145,39,3,2,4,1,1,2,2,26903,4630,159418,0,0,80716,0,24471,2
+0,0,225145,39,3,2,4,1,1,2,2,282069,2967,230832,84386,5400,0,0,0,2
+0,1,186821,75,10,1,3,2,1,2,2,62621,4544,58151,70939,0,0,0,0,3
+0,0,186821,75,10,1,3,2,1,2,2,93138,532,115823,87562,3213,0,0,16010,3
+0,0,186821,75,10,1,3,2,1,2,2,45206,2820,183822,13990,3138,63787,37851,4086,3
+0,0,146759,60,11,2,6,0,2,2,0,368243,2215,123885,92072,5027,0,0,0,1
+0,1,146759,60,11,2,6,0,2,2,0,215196,959,195714,0,3110,0,0,9556,1
+0,0,131084,23,11,2,6,3,1,2,2,239163,1830,99024,90467,2743,0,0,0,1
+0,0,131084,23,11,2,6,3,1,2,2,264944,2096,111223,67585,4293,0,0,0,1
+0,0,131084,23,11,2,6,3,1,2,2,233483,598,197331,53113,0,0,0,0,1
+0,0,131084,23,11,2,6,3,1,2,2,312012,3757,205654,22313,1892,0,0,0,1
+0,0,131084,23,11,2,6,3,1,2,2,198046,4010,46170,0,567,0,0,0,1
+0,0,25124,60,5,2,5,0,1,2,2,314484,6224,138638,0,533,37733,0,0,1
+0,0,25124,60,5,2,5,0,1,2,2,258277,5670,74133,72066,1197,72538,20372,0,1
+0,0,25124,60,5,2,5,0,1,2,2,0,4770,137189,37029,1778,0,0,0,1
+0,0,25124,60,5,2,5,0,1,2,2,424602,2114,130310,60811,2261,0,0,32948,1
+0,0,25124,60,5,2,5,0,1,2,2,209585,3746,60855,34322,3483,87904,26605,0,1
+1,0,25124,60,5,2,5,0,1,2,2,433251,572,36404,0,319,103340,0,0,1
+0,0,139588,60,10,1,3,0,1,2,2,319344,6048,176737,0,0,0,0,0,1
+0,0,139588,60,10,1,3,0,1,2,2,170111,5294,255332,26976,0,0,0,0,1
+0,0,139588,60,10,1,3,0,1,2,2,35,5674,100397,21139,0,0,0,0,1
+0,0,233279,60,13,1,5,0,1,2,3,225154,5748,229278,0,4488,0,0,28266,3
+0,0,233279,60,13,1,5,0,1,2,3,0,3612,158255,46162,5182,0,0,0,3
+0,0,233279,60,13,1,5,0,1,2,3,86976,2541,234649,75046,4702,0,0,0,3
+0,0,233279,60,13,1,5,0,1,2,3,161648,2748,15385,0,3270,0,0,12825,3
+0,0,233279,60,13,1,5,0,1,2,3,71718,2371,74049,83673,4259,36871,46702,9076,3
+0,0,233279,60,13,1,5,0,1,2,3,108993,2139,107889,0,5313,0,0,0,3
+0,0,57653,60,10,1,3,0,1,2,1,405019,5473,238445,86159,1489,0,31263,14752,3
+0,0,57653,60,10,1,3,0,1,2,1,103212,5473,62663,13073,1489,0,0,0,3
+0,0,57653,60,10,1,3,0,1,2,1,464808,887,212115,0,3249,0,0,14835,3
+0,0,94847,0,0,0,0,0,0,0,0,10251,4770,1730,0,1778,0,0,989,1
+0,0,94847,0,0,0,0,0,0,0,0,206244,2139,246373,97164,5313,0,0,11014,1
+0,0,94847,0,0,0,0,0,0,0,0,222670,4770,91229,13685,1778,0,0,8594,1
+0,0,94847,0,0,0,0,0,0,0,0,308509,4010,50918,104006,0,0,0,0,1
+0,0,94847,0,0,0,0,0,0,0,0,60836,3639,47235,65915,662,0,0,0,1
+0,0,94847,0,0,0,0,0,0,0,0,437045,2697,6369,105353,0,0,0,0,1
+0,0,94847,0,0,0,0,0,0,0,0,28761,1005,91855,91674,5425,0,0,0,1
+0,0,94847,0,0,0,0,0,0,0,0,0,990,0,0,0,0,0,0,1

models/rank/aitm/data/sample_data/train/train.csv

@@ -0,0 +1,49 @@
+click,purchase,101,121,122,124,125,126,127,128,129,205,206,207,216,508,509,702,853,301
+0,0,85077,60,5,2,5,0,1,2,1,303983,3751,249750,4124,5565,0,0,6260,2
+0,0,85077,60,5,2,5,0,1,2,1,200850,1005,156972,315,5425,0,0,31950,2
+0,0,85077,60,5,2,5,0,1,2,1,361912,2748,116642,103627,3270,0,0,0,2
+0,0,85077,60,5,2,5,0,1,2,1,235052,4791,18014,92858,1575,18172,39447,12632,2
+0,0,85077,60,5,2,5,0,1,2,1,236563,1464,49011,27526,2400,0,40351,0,2
+1,1,85077,60,5,2,5,0,1,2,1,0,2668,73088,0,1349,36448,0,3202,2
+0,0,85077,60,5,2,5,0,1,2,1,868,2114,134232,27903,2261,0,42361,8186,2
+0,1,85077,60,5,2,5,0,1,2,1,278769,887,252320,35421,0,0,0,0,2
+0,0,85077,60,5,2,5,0,1,2,1,378778,959,237727,24961,3110,0,12437,0,2
+0,0,85077,60,5,2,5,0,1,2,1,286476,2668,36149,50475,1349,0,17247,24785,2
+1,0,85077,60,5,2,5,0,1,2,1,37325,1219,210814,0,732,0,0,0,2
+0,0,85077,60,5,2,5,0,1,2,1,0,6396,145671,104006,0,0,10562,0,2
+0,1,85077,60,5,2,5,0,1,2,1,0,6396,187630,3695,0,0,39953,0,2
+0,0,85077,60,5,2,5,0,1,2,1,174938,2542,121978,25017,2752,0,0,16432,2
+0,0,85077,60,5,2,5,0,1,2,1,126218,2748,99220,3695,3270,1641,39953,0,2
+0,0,85077,60,5,2,5,0,1,2,1,137678,2668,156813,76812,1349,0,0,0,2
+0,0,85077,60,5,2,5,0,1,2,1,0,6639,75496,0,1641,0,0,0,2
+0,0,85077,60,5,2,5,0,1,2,1,0,3358,28909,24961,1715,33178,12437,0,2
+0,0,85077,60,5,2,5,0,1,2,1,463355,2321,143602,0,2827,0,0,0,2
+0,0,85077,60,5,2,5,0,1,2,1,208394,5441,165858,100800,3175,0,0,14526,2
+0,0,85077,60,5,2,5,0,1,2,1,0,4770,160421,49203,1778,104211,12570,0,2
+0,0,85077,60,5,2,5,0,1,2,1,0,5130,152781,34023,3305,0,0,0,2
+0,0,85077,60,5,2,5,0,1,2,1,0,1219,71745,37124,732,0,43897,0,2
+0,0,85077,60,5,2,5,0,1,2,1,0,4770,78413,0,1778,68142,0,0,2
+0,0,85077,60,5,2,5,0,1,2,1,0,4770,0,51246,1778,0,12494,0,2
+0,0,85077,60,5,2,5,0,1,2,1,0,4770,12306,99058,1778,94838,23330,0,2
+0,0,126936,75,10,1,3,1,1,2,2,436088,5342,251939,55848,5686,0,43802,891,3
+0,0,225894,36,5,2,5,1,1,2,1,236983,2114,100033,45859,2261,104041,37050,32948,2
+0,0,225894,36,5,2,5,1,1,2,1,0,2668,81805,94691,1349,80860,23323,36346,2
+1,0,225894,36,5,2,5,1,1,2,1,236983,2114,100033,45859,2261,104041,37050,32948,2
+0,0,225894,36,5,2,5,1,1,2,1,76908,1323,2845,4224,0,0,0,0,2
+0,0,225894,36,5,2,5,1,1,2,1,0,2139,27250,65082,5313,86671,24195,0,2
+0,0,214292,60,13,1,5,0,1,2,2,405168,4408,200625,53253,2913,0,0,0,3
+0,0,214292,60,13,1,5,0,1,2,2,0,3711,105033,0,3877,0,0,0,3
+0,0,214292,60,13,1,5,0,1,2,2,342042,3346,230129,0,610,0,0,0,3
+0,0,214292,60,13,1,5,0,1,2,2,140466,3711,160865,33806,3877,0,0,19879,3
+0,0,214292,60,13,1,5,0,1,2,2,243201,3924,130907,30524,1289,0,0,0,3
+0,0,214292,60,13,1,5,0,1,2,2,229158,3643,123318,92849,0,0,0,0,3
+0,0,214292,60,13,1,5,0,1,2,2,438610,4782,215293,100092,0,0,0,0,3
+0,0,214292,60,13,1,5,0,1,2,2,106945,5334,16167,34467,4967,0,0,0,3
+0,0,214292,60,13,1,5,0,1,2,2,72148,2177,182249,0,5689,93690,0,13168,3
+0,0,214292,60,13,1,5,0,1,2,2,145491,989,248896,65379,0,0,0,0,3
+0,0,214292,60,13,1,5,0,1,2,2,49677,5673,57894,0,0,0,0,0,3
+0,0,214292,60,13,1,5,0,1,2,2,401746,3334,54001,104747,0,83894,0,0,3
+0,0,214292,60,13,1,5,0,1,2,2,293581,1685,30366,47952,2631,0,0,12116,3
+0,0,214292,60,13,1,5,0,1,2,2,211989,2815,162936,92766,0,0,0,0,3
+0,0,214292,60,13,1,5,0,1,2,2,398227,4782,12082,65379,0,0,0,0,3
+0,0,214292,60,13,1,5,0,1,2,2,0,3713,0,68249,1813,0,27110,0,3

models/rank/aitm/dygraph_model.py

@@ -0,0 +1,108 @@
+# 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.
+
+import paddle
+import paddle.nn as nn
+import paddle.nn.functional as F
+import net
+
+
+class DygraphModel():
+    def create_model(self, config):
+        feature_vocabulary = config.get("hyper_parameters.feature_vocabulary")
+        embedding_size = config.get("hyper_parameters.embedding_size")
+        tower_dims = config.get("hyper_parameters.dims")
+        drop_prob = config.get('hyper_parameters.drop_prob')
+        feature_vocabulary = dict(feature_vocabulary)
+        model = net.AITM(feature_vocabulary, embedding_size, tower_dims,
+                         drop_prob)
+        return model
+
+    # define feeds which convert numpy of batch data to paddle.tensor 
+    def create_feeds(self, batch_data, config):
+        click, conversion, features = batch_data
+        return click.astype('float32'), conversion.astype('float32'), features
+
+    # define loss function by predicts and label
+    def create_loss(self,
+                    click_pred,
+                    conversion_pred,
+                    click_label,
+                    conversion_label,
+                    constraint_weight=0.6):
+        click_loss = F.binary_cross_entropy(click_pred, click_label)
+        conversion_loss = F.binary_cross_entropy(conversion_pred,
+                                                 conversion_label)
+
+        label_constraint = paddle.maximum(conversion_pred - click_pred,
+                                          paddle.zeros_like(click_label))
+        constraint_loss = paddle.sum(label_constraint)
+
+        loss = click_loss + conversion_loss + constraint_weight * constraint_loss
+        return loss
+
+    # define optimizer 
+    def create_optimizer(self, dy_model, config):
+        lr = config.get("hyper_parameters.optimizer.learning_rate", 0.0001)
+        optimizer = paddle.optimizer.Adam(
+            learning_rate=lr,
+            parameters=dy_model.parameters(),
+            weight_decay=1e-6)
+        return optimizer
+
+    # define metrics such as auc/acc
+    # multi-task need to define multi metric
+    def create_metrics(self):
+        metrics_list_name = ["click_auc", "purchase_auc"]
+        metrics_list = [
+            paddle.metric.Auc("ROC", num_thresholds=100000),
+            paddle.metric.Auc("ROC", num_thresholds=100000)
+        ]
+        return metrics_list, metrics_list_name
+
+    # construct train forward phase  
+    def train_forward(self, dy_model, metrics_list, batch_data, config):
+        click, conversion, features = self.create_feeds(batch_data, config)
+        click_pred, conversion_pred = dy_model.forward(features)
+        loss = self.create_loss(click_pred, conversion_pred, click, conversion)
+        # update metrics
+
+        self.update_auc(click_pred, click, metrics_list[0])
+        self.update_auc(conversion_pred, conversion, metrics_list[1])
+        print_dict = {'loss': loss}
+        return loss, metrics_list, print_dict
+
+    @staticmethod
+    def update_auc(prob, label, metrics):
+        if prob.ndim == 1:
+            prob = prob.unsqueeze(-1)
+        assert prob.ndim == 2
+        predict_2d = paddle.concat(x=[1 - prob, prob], axis=1)
+        metrics.update(predict_2d, label)
+
+    def infer_forward(self, dy_model, metrics_list, batch_data, config):
+        click, conversion, features = self.create_feeds(batch_data, config)
+        with paddle.no_grad():
+            click_pred, conversion_pred = dy_model.forward(features)
+        # update metrics
+        self.update_auc(click_pred, click, metrics_list[0])
+        self.update_auc(conversion_pred, conversion, metrics_list[1])
+        return metrics_list, None
+
+    def forward(self, dy_model, batch_data, config):
+        click, conversion, features = self.create_feeds(batch_data, config)
+        with paddle.no_grad():
+            click_pred, conversion_pred = dy_model.forward(features)
+        # update metrics
+        return click, click_pred, conversion, conversion_pred