fix clnflict

.pre-commit-config.yaml

+repos:
+-   repo: https://github.com/Lucas-C/pre-commit-hooks.git
+    sha: v1.0.1
+    hooks:
+    -   id: remove-crlf
+        files: (?!.*third_party)^.*$ | (?!.*book)^.*$
+-   repo: https://github.com/PaddlePaddle/mirrors-yapf.git
+    sha: 0d79c0c469bab64f7229c9aca2b1186ef47f0e37
+    hooks:
+    -   id: yapf
+        files: (.*\.(py|bzl)|BUILD|.*\.BUILD|WORKSPACE)$
+-   repo: https://github.com/pre-commit/pre-commit-hooks
+    sha: 5bf6c09bfa1297d3692cadd621ef95f1284e33c0
+    hooks:
+    -   id: check-added-large-files
+    -   id: check-merge-conflict
+    -   id: check-symlinks
+    -   id: detect-private-key
+        files: (?!.*third_party)^.*$ | (?!.*book)^.*$
+    -   id: end-of-file-fixer
+-   repo: local
+    hooks:
+    -   id: copyright_checker
+        name: copyright_checker
+        entry: python ./tools/codestyle/copyright.hook
+        language: system
+        files: \.(c|cc|cxx|cpp|cu|h|hpp|hxx|proto|py)$
+        exclude: (?!.*third_party)^.*$ | (?!.*book)^.*$

.travis.yml

+language: generic 
+sudo: required
+dist: trusty
+
+services:
+  - docker
+
+os:
+  - linux
+
+env:
+  - JOB=check_style
+
+before_install:
+  # For pylint dockstring checker
+  - sudo pip install pylint pytest astroid isort pre-commit
+  - |
+    function timeout() { perl -e 'alarm shift; exec @ARGV' "$@"; }
+
+script:
+  - "travis_wait 30 sleep 1800 &"
+  - |
+    # 43min timeout
+    tools/build_script.sh ${JOB}
+    if [ $? -eq 0 ] || [ $? -eq 142 ]; then true; else exit 1; fi;
+
+notifications:
+  email:
+    on_success: change
+    on_failure: always

core/__init__.py

+# 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.

core/engine/__init__.py

+# 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.

core/engine/cluster/__init__.py

+# 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.

core/engine/cluster/cloud/__init__.py

+# 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.

core/engine/cluster/cluster.py

@@ -27,6 +27,7 @@ from paddlerec.core.utils import envs
 class ClusterEngine(Engine):
     def __init_impl__(self):
         abs_dir = os.path.dirname(os.path.abspath(__file__))
+
         backend = envs.get_runtime_environ("engine_backend")
         if backend == "PaddleCloud":
             self.submit_script = os.path.join(abs_dir, "cloud/cluster.sh")
@@ -57,4 +58,5 @@ class ClusterEngine(Engine):
             self.start_worker_procs()
 
         else:
-            raise ValueError("role {} error, must in MASTER/WORKER".format(role))
+            raise ValueError("role {} error, must in MASTER/WORKER".format(
+                role))

core/engine/local_cluster.py

@@ -46,10 +46,13 @@ class LocalClusterEngine(Engine):
                     ports.append(new_port)
                     break
         user_endpoints = ",".join(["127.0.0.1:" + str(x) for x in ports])
-        user_endpoints_ips = [x.split(":")[0]
-                              for x in user_endpoints.split(",")]
-        user_endpoints_port = [x.split(":")[1]
-                               for x in user_endpoints.split(",")]
+
+        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 = "paddlerec.core.factory"
         cmd = [sys.executable, "-u", "-m", factory, self.trainer]
@@ -97,8 +100,10 @@ class LocalClusterEngine(Engine):
             if len(log_fns) > 0:
                 log_fns[i].close()
             procs[i].terminate()
-        print("all workers already completed, you can view logs under the `{}` directory".format(logs_dir),
-              file=sys.stderr)
+        print(
+            "all workers already completed, you can view logs under the `{}` directory".
+            format(logs_dir),
+            file=sys.stderr)
 
     def run(self):
         self.start_procs()

core/engine/local_mpi.py

@@ -26,7 +26,6 @@ from paddlerec.core.engine.engine import Engine
 class LocalMPIEngine(Engine):
     def start_procs(self):
         logs_dir = self.envs["log_dir"]
-
         default_env = os.environ.copy()
         current_env = copy.copy(default_env)
         current_env.pop("http_proxy", None)
@@ -42,7 +41,8 @@ class LocalMPIEngine(Engine):
             os.system("mkdir -p {}".format(logs_dir))
             fn = open("%s/job.log" % logs_dir, "w")
             log_fns.append(fn)
-            proc = subprocess.Popen(cmd, env=current_env, stdout=fn, stderr=fn, cwd=os.getcwd())
+            proc = subprocess.Popen(
+                cmd, env=current_env, stdout=fn, stderr=fn, cwd=os.getcwd())
         else:
             proc = subprocess.Popen(cmd, env=current_env, cwd=os.getcwd())
         procs.append(proc)
@@ -51,7 +51,9 @@ class LocalMPIEngine(Engine):
             if len(log_fns) > 0:
                 log_fns[i].close()
             procs[i].wait()
-        print("all workers and parameter servers already completed", file=sys.stderr)
+        print(
+            "all workers and parameter servers already completed",
+            file=sys.stderr)
 
     def run(self):
         self.start_procs()

core/factory.py

@@ -19,24 +19,23 @@ import yaml
 
 from paddlerec.core.utils import envs
 
-trainer_abs = os.path.join(os.path.dirname(
-    os.path.abspath(__file__)), "trainers")
+trainer_abs = os.path.join(
+    os.path.dirname(os.path.abspath(__file__)), "trainers")
 trainers = {}
 
 
 def trainer_registry():
-    trainers["SingleTrainer"] = os.path.join(
-        trainer_abs, "single_trainer.py")
-    trainers["ClusterTrainer"] = os.path.join(
-        trainer_abs, "cluster_trainer.py")
-    trainers["CtrCodingTrainer"] = os.path.join(
-        trainer_abs, "ctr_coding_trainer.py")
-    trainers["CtrModulTrainer"] = os.path.join(
-        trainer_abs, "ctr_modul_trainer.py")
-    trainers["TDMSingleTrainer"] = os.path.join(
-        trainer_abs, "tdm_single_trainer.py")
-    trainers["TDMClusterTrainer"] = os.path.join(
-        trainer_abs, "tdm_cluster_trainer.py")
+    trainers["SingleTrainer"] = os.path.join(trainer_abs, "single_trainer.py")
+    trainers["ClusterTrainer"] = os.path.join(trainer_abs,
+                                              "cluster_trainer.py")
+    trainers["CtrCodingTrainer"] = os.path.join(trainer_abs,
+                                                "ctr_coding_trainer.py")
+    trainers["CtrModulTrainer"] = os.path.join(trainer_abs,
+                                               "ctr_modul_trainer.py")
+    trainers["TDMSingleTrainer"] = os.path.join(trainer_abs,
+                                                "tdm_single_trainer.py")
+    trainers["TDMClusterTrainer"] = os.path.join(trainer_abs,
+                                                 "tdm_cluster_trainer.py")
 
 
 trainer_registry()
@@ -55,8 +54,8 @@ class TrainerFactory(object):
 
         if trainer_abs is None:
             if not os.path.isfile(train_mode):
-                raise IOError(
-                    "trainer {} can not be recognized".format(train_mode))
+                raise IOError("trainer {} can not be recognized".format(
+                    train_mode))
             trainer_abs = train_mode
             train_mode = "UserDefineTrainer"
 

core/metrics/auc_metrics.py

@@ -22,7 +22,7 @@ from paddlerec.core.metric import Metric
 
 class AUCMetric(Metric):
     """
-    Metric For Paddle Model
+    Metric For Fluid Model
     """
 
     def __init__(self, config, fleet):
@@ -83,7 +83,8 @@ class AUCMetric(Metric):
             if scope.find_var(metric_item['var'].name) is None:
                 result[metric_name] = None
                 continue
-            result[metric_name] = self.get_metric(scope, metric_item['var'].name)
+            result[metric_name] = self.get_metric(scope,
+                                                  metric_item['var'].name)
         return result
 
     def calculate_auc(self, global_pos, global_neg):
@@ -178,14 +179,18 @@ class AUCMetric(Metric):
             self._result['mean_q'] = 0
             return self._result
         if 'stat_pos' in result and 'stat_neg' in result:
-            result['auc'] = self.calculate_auc(result['stat_pos'], result['stat_neg'])
-            result['bucket_error'] = self.calculate_auc(result['stat_pos'], result['stat_neg'])
+            result['auc'] = self.calculate_auc(result['stat_pos'],
+                                               result['stat_neg'])
+            result['bucket_error'] = self.calculate_auc(result['stat_pos'],
+                                                        result['stat_neg'])
         if 'pos_ins_num' in result:
-            result['actual_ctr'] = result['pos_ins_num'] / result['total_ins_num']
+            result['actual_ctr'] = result['pos_ins_num'] / result[
+                'total_ins_num']
         if 'abserr' in result:
             result['mae'] = result['abserr'] / result['total_ins_num']
         if 'sqrerr' in result:
-            result['rmse'] = math.sqrt(result['sqrerr'] / result['total_ins_num'])
+            result['rmse'] = math.sqrt(result['sqrerr'] /
+                                       result['total_ins_num'])
         if 'prob' in result:
             result['predict_ctr'] = result['prob'] / result['total_ins_num']
             if abs(result['predict_ctr']) > 1e-6:

core/model.py

@@ -20,7 +20,7 @@ from paddlerec.core.utils import envs
 
 
 class Model(object):
-    """R
+    """Base Model
     """
     __metaclass__ = abc.ABCMeta
 
@@ -38,6 +38,45 @@ class Model(object):
         self._namespace = "train.model"
         self._platform = envs.get_platform()
 
+    def _init_slots(self):
+        sparse_slots = envs.get_global_env("sparse_slots", None,
+                                           "train.reader")
+        dense_slots = envs.get_global_env("dense_slots", None, "train.reader")
+
+        if sparse_slots is not None or dense_slots is not None:
+            sparse_slots = sparse_slots.strip().split(" ")
+            dense_slots = dense_slots.strip().split(" ")
+            dense_slots_shape = [[
+                int(j) for j in i.split(":")[1].strip("[]").split(",")
+            ] for i in dense_slots]
+            dense_slots = [i.split(":")[0] for i in dense_slots]
+            self._dense_data_var = []
+            for i in range(len(dense_slots)):
+                l = fluid.layers.data(
+                    name=dense_slots[i],
+                    shape=dense_slots_shape[i],
+                    dtype="float32")
+                self._data_var.append(l)
+                self._dense_data_var.append(l)
+            self._sparse_data_var = []
+            for name in sparse_slots:
+                l = fluid.layers.data(
+                    name=name, shape=[1], lod_level=1, dtype="int64")
+                self._data_var.append(l)
+                self._sparse_data_var.append(l)
+
+        dataset_class = envs.get_global_env("dataset_class", None,
+                                            "train.reader")
+        if dataset_class == "DataLoader":
+            self._init_dataloader()
+
+    def _init_dataloader(self):
+        self._data_loader = fluid.io.DataLoader.from_generator(
+            feed_list=self._data_var,
+            capacity=64,
+            use_double_buffer=False,
+            iterable=False)
+
     def get_inputs(self):
         return self._data_var
 
@@ -68,8 +107,8 @@ class Model(object):
                 "configured optimizer can only supported SGD/Adam/Adagrad")
 
         if name == "SGD":
-            reg = envs.get_global_env(
-                "hyper_parameters.reg", 0.0001, self._namespace)
+            reg = envs.get_global_env("hyper_parameters.reg", 0.0001,
+                                      self._namespace)
             optimizer_i = fluid.optimizer.SGD(
                 lr, regularization=fluid.regularizer.L2DecayRegularizer(reg))
         elif name == "ADAM":
@@ -83,10 +122,10 @@ class Model(object):
         return optimizer_i
 
     def optimizer(self):
-        learning_rate = envs.get_global_env(
-            "hyper_parameters.learning_rate", None, self._namespace)
-        optimizer = envs.get_global_env(
-            "hyper_parameters.optimizer", None, self._namespace)
+        learning_rate = envs.get_global_env("hyper_parameters.learning_rate",
+                                            None, self._namespace)
+        optimizer = envs.get_global_env("hyper_parameters.optimizer", None,
+                                        self._namespace)
         print(">>>>>>>>>>>.learnig rate: %s" % learning_rate)
         return self._build_optimizer(optimizer, learning_rate)
 

core/modules/__init__.py

+# 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.

core/modules/coding/__init__.py

+# 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.

core/modules/coding/layers.py

+# 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.

core/modules/modul/__init__.py

+# 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.

core/modules/modul/build.py

@@ -31,6 +31,7 @@ def create(config):
         Model Instance
     """
     model = None
+
     if config['mode'] == 'fluid':
         model = YamlModel(config)
         model.train_net()
@@ -50,7 +51,12 @@ class YamlModel(Model):
         f = open(config['layer_file'], 'r')
         self._build_nodes = yaml.safe_load(f.read())
         self._build_phase = ['input', 'param', 'summary', 'layer']
-        self._build_param = {'layer': {}, 'inner_layer': {}, 'layer_extend': {}, 'model': {}}
+        self._build_param = {
+            'layer': {},
+            'inner_layer': {},
+            'layer_extend': {},
+            'model': {}
+        }
         self._inference_meta = {'dependency': {}, 'params': {}}
 
     def train_net(self):
@@ -76,10 +82,12 @@ class YamlModel(Model):
                     if self._build_nodes[phase] is None:
                         continue
                     for node in self._build_nodes[phase]:
-                        exec("""layer=layer.{}(node)""".format(node['class']))
-                        layer_output, extend_output = layer.generate(self._config['mode'], self._build_param)
+                        exec ("""layer=layer.{}(node)""".format(node['class']))
+                        layer_output, extend_output = layer.generate(
+                            self._config['mode'], self._build_param)
                         self._build_param['layer'][node['name']] = layer_output
-                        self._build_param['layer_extend'][node['name']] = extend_output
+                        self._build_param['layer_extend'][node[
+                            'name']] = extend_output
                         if extend_output is None:
                             continue
                         if 'loss' in extend_output:
@@ -89,17 +97,24 @@ class YamlModel(Model):
                                 self._cost += extend_output['loss']
                         if 'data_var' in extend_output:
                             self._data_var += extend_output['data_var']
-                        if 'metric_label' in extend_output and extend_output['metric_label'] is not None:
-                            self._metrics[extend_output['metric_label']] = extend_output['metric_dict']
+                        if 'metric_label' in extend_output and extend_output[
+                                'metric_label'] is not None:
+                            self._metrics[extend_output[
+                                'metric_label']] = extend_output['metric_dict']
 
                         if 'inference_param' in extend_output:
                             inference_param = extend_output['inference_param']
                             param_name = inference_param['name']
                             if param_name not in self._build_param['table']:
-                                self._build_param['table'][param_name] = {'params': []}
-                                table_meta = table.TableMeta.alloc_new_table(inference_param['table_id'])
-                                self._build_param['table'][param_name]['_meta'] = table_meta
-                            self._build_param['table'][param_name]['params'] += inference_param['params']
+                                self._build_param['table'][param_name] = {
+                                    'params': []
+                                }
+                                table_meta = table.TableMeta.alloc_new_table(
+                                    inference_param['table_id'])
+                                self._build_param['table'][param_name][
+                                    '_meta'] = table_meta
+                            self._build_param['table'][param_name][
+                                'params'] += inference_param['params']
         pass
 
     @classmethod
@@ -114,20 +129,25 @@ class YamlModel(Model):
             metrics = params['metrics']
             for name in metrics:
                 model_metrics = metrics[name]
-                stat_var_names += [model_metrics[metric]['var'].name for metric in model_metrics]
+                stat_var_names += [
+                    model_metrics[metric]['var'].name
+                    for metric in model_metrics
+                ]
             strategy['stat_var_names'] = list(set(stat_var_names))
         optimizer_generator = 'optimizer = fluid.optimizer.' + optimizer_conf['class'] + \
                               '(learning_rate=' + str(optimizer_conf['learning_rate']) + ')'
-        exec(optimizer_generator)
+        exec (optimizer_generator)
         optimizer = fleet.distributed_optimizer(optimizer, strategy=strategy)
         return optimizer
 
     def dump_model_program(self, path):
         """R
         """
-        with open(path + '/' + self._name + '_main_program.pbtxt', "w") as fout:
+        with open(path + '/' + self._name + '_main_program.pbtxt',
+                  "w") as fout:
             print >> fout, self._build_param['model']['train_program']
-        with open(path + '/' + self._name + '_startup_program.pbtxt', "w") as fout:
+        with open(path + '/' + self._name + '_startup_program.pbtxt',
+                  "w") as fout:
             print >> fout, self._build_param['model']['startup_program']
         pass
 
@@ -137,7 +157,8 @@ class YamlModel(Model):
         scope = params['scope']
         decay = params['decay']
         for param_table in self._build_param['table']:
-            table_id = self._build_param['table'][param_table]['_meta']._table_id
+            table_id = self._build_param['table'][param_table][
+                '_meta']._table_id
             fleet.shrink_dense_table(decay, scope=scope, table_id=table_id)
 
     def dump_inference_program(self, inference_layer, path):
@@ -152,17 +173,25 @@ class YamlModel(Model):
         executor = params['executor']
         program = self._build_param['model']['train_program']
         for table_name, table in self._build_param['table'].items():
-            fleet._fleet_ptr.pull_dense(scope, table['_meta']._table_id, table['params'])
+            fleet._fleet_ptr.pull_dense(scope, table['_meta']._table_id,
+                                        table['params'])
         for infernce_item in params['inference_list']:
-            params_name_list = self.inference_params(infernce_item['layer_name'])
-            params_var_list = [program.global_block().var(i) for i in params_name_list]
+            params_name_list = self.inference_params(infernce_item[
+                'layer_name'])
+            params_var_list = [
+                program.global_block().var(i) for i in params_name_list
+            ]
             params_file_name = infernce_item['save_file_name']
             with fluid.scope_guard(scope):
                 if params['save_combine']:
                     fluid.io.save_vars(executor, "./", \
                                        program, vars=params_var_list, filename=params_file_name)
                 else:
-                    fluid.io.save_vars(executor, params_file_name, program, vars=params_var_list)
+                    fluid.io.save_vars(
+                        executor,
+                        params_file_name,
+                        program,
+                        vars=params_var_list)
 
     def inference_params(self, inference_layer):
         """
@@ -177,11 +206,13 @@ class YamlModel(Model):
             return self._inference_meta['params'][layer]
 
         self._inference_meta['params'][layer] = []
-        self._inference_meta['dependency'][layer] = self.get_dependency(self._build_param['inner_layer'], layer)
+        self._inference_meta['dependency'][layer] = self.get_dependency(
+            self._build_param['inner_layer'], layer)
         for node in self._build_nodes['layer']:
             if node['name'] not in self._inference_meta['dependency'][layer]:
                 continue
-            if 'inference_param' in self._build_param['layer_extend'][node['name']]:
+            if 'inference_param' in self._build_param['layer_extend'][node[
+                    'name']]:
                 self._inference_meta['params'][layer] += \
                     self._build_param['layer_extend'][node['name']]['inference_param']['params']
         return self._inference_meta['params'][layer]
@@ -199,5 +230,6 @@ class YamlModel(Model):
             dependencys = copy.deepcopy(layer_graph[dest_layer]['input'])
             dependency_list = copy.deepcopy(dependencys)
             for dependency in dependencys:
-                dependency_list = dependency_list + self.get_dependency(layer_graph, dependency)
+                dependency_list = dependency_list + self.get_dependency(
+                    layer_graph, dependency)
         return list(set(dependency_list))

core/modules/modul/layers.py

@@ -18,7 +18,7 @@ from paddlerec.core.layer import Layer
 
 
 class EmbeddingFuseLayer(Layer):
-    """R
+    """embedding + sequence + concat
     """
 
     def __init__(self, config):
@@ -40,7 +40,8 @@ class EmbeddingFuseLayer(Layer):
         show_clk.stop_gradient = True
         data_var = []
         for slot in self._slots:
-            l = fluid.layers.data(name=slot, shape=[1], dtype="int64", lod_level=1)
+            l = fluid.layers.data(
+                name=slot, shape=[1], dtype="int64", lod_level=1)
             data_var.append(l)
             emb = fluid.layers.embedding(input=l, size=[10, self._emb_dim], \
                                          is_sparse=True, is_distributed=True,
@@ -48,7 +49,8 @@ class EmbeddingFuseLayer(Layer):
             emb = fluid.layers.sequence_pool(input=emb, pool_type='sum')
             emb = fluid.layers.continuous_value_model(emb, show_clk, self._cvm)
             self._emb_layers.append(emb)
-        output = fluid.layers.concat(input=self._emb_layers, axis=1, name=self._name)
+        output = fluid.layers.concat(
+            input=self._emb_layers, axis=1, name=self._name)
         return output, {'data_var': data_var}
 
 
@@ -111,7 +113,13 @@ class ParamLayer(Layer):
     def generate(self, param):
         """R
         """
-        return self._config, {'inference_param': {'name': 'param', 'params': [], 'table_id': self._table_id}}
+        return self._config, {
+            'inference_param': {
+                'name': 'param',
+                'params': [],
+                'table_id': self._table_id
+            }
+        }
 
 
 class SummaryLayer(Layer):
@@ -129,10 +137,16 @@ class SummaryLayer(Layer):
     def generate(self, param):
         """R
         """
-        return self._config, {'inference_param': {'name': 'summary', 'params': [], 'table_id': self._table_id}}
+        return self._config, {
+            'inference_param': {
+                'name': 'summary',
+                'params': [],
+                'table_id': self._table_id
+            }
+        }
 
 
-class NormalizetionLayer(Layer):
+class NormalizationLayer(Layer):
     """R
     """
 
@@ -152,9 +166,19 @@ class NormalizetionLayer(Layer):
         if len(self._input) > 0:
             input_list = [param['layer'][i] for i in self._input]
             input_layer = fluid.layers.concat(input=input_list, axis=1)
-        bn = fluid.layers.data_norm(input=input_layer, name=self._name, epsilon=1e-4, param_attr={
-            "batch_size": 1e4, "batch_sum_default": 0.0, "batch_square": 1e4})
-        inference_param = [self._name + '.batch_size', self._name + '.batch_sum', self._name + '.batch_square_sum']
+        bn = fluid.layers.data_norm(
+            input=input_layer,
+            name=self._name,
+            epsilon=1e-4,
+            param_attr={
+                "batch_size": 1e4,
+                "batch_sum_default": 0.0,
+                "batch_square": 1e4
+            })
+        inference_param = [
+            self._name + '.batch_size', self._name + '.batch_sum',
+            self._name + '.batch_square_sum'
+        ]
         return bn, {'inference_param': {'name': 'summary', \
                                         'params': inference_param, 'table_id': summary_layer.get('table_id', -1)}}
 
@@ -181,11 +205,13 @@ class FCLayer(Layer):
             input_list = [param['layer'][i] for i in self._input]
             input_layer = fluid.layers.concat(input=input_list, axis=1)
         input_coln = input_layer.shape[1]
-        scale = param_layer['init_range'] / (input_coln ** 0.5)
+        scale = param_layer['init_range'] / (input_coln**0.5)
         bias = None
         if self._bias:
-            bias = fluid.ParamAttr(learning_rate=1.0,
-                                   initializer=fluid.initializer.NormalInitializer(loc=0.0, scale=scale))
+            bias = fluid.ParamAttr(
+                learning_rate=1.0,
+                initializer=fluid.initializer.NormalInitializer(
+                    loc=0.0, scale=scale))
         fc = fluid.layers.fc(
             name=self._name,
             input=input_layer,
@@ -216,18 +242,46 @@ class LogLossLayer(Layer):
         self._extend_output = {
             'metric_label': self._metric_label,
             'metric_dict': {
-                'auc': {'var': None},
-                'batch_auc': {'var': None},
-                'stat_pos': {'var': None, 'data_type': 'int64'},
-                'stat_neg': {'var': None, 'data_type': 'int64'},
-                'batch_stat_pos': {'var': None, 'data_type': 'int64'},
-                'batch_stat_neg': {'var': None, 'data_type': 'int64'},
-                'pos_ins_num': {'var': None},
-                'abserr': {'var': None},
-                'sqrerr': {'var': None},
-                'prob': {'var': None},
-                'total_ins_num': {'var': None},
-                'q': {'var': None}
+                'auc': {
+                    'var': None
+                },
+                'batch_auc': {
+                    'var': None
+                },
+                'stat_pos': {
+                    'var': None,
+                    'data_type': 'int64'
+                },
+                'stat_neg': {
+                    'var': None,
+                    'data_type': 'int64'
+                },
+                'batch_stat_pos': {
+                    'var': None,
+                    'data_type': 'int64'
+                },
+                'batch_stat_neg': {
+                    'var': None,
+                    'data_type': 'int64'
+                },
+                'pos_ins_num': {
+                    'var': None
+                },
+                'abserr': {
+                    'var': None
+                },
+                'sqrerr': {
+                    'var': None
+                },
+                'prob': {
+                    'var': None
+                },
+                'total_ins_num': {
+                    'var': None
+                },
+                'q': {
+                    'var': None
+                }
             }
         }
 
@@ -236,9 +290,12 @@ class LogLossLayer(Layer):
         """
         input_layer = param['layer'][self._input[0]]
         label_layer = param['layer'][self._label]
-        output = fluid.layers.clip(input_layer, self._bound[0], self._bound[1], name=self._name)
+        output = fluid.layers.clip(
+            input_layer, self._bound[0], self._bound[1], name=self._name)
         norm = fluid.layers.sigmoid(output, name=self._name)
-        output = fluid.layers.log_loss(norm, fluid.layers.cast(x=label_layer, dtype='float32'))
+        output = fluid.layers.log_loss(
+            norm, fluid.layers.cast(
+                x=label_layer, dtype='float32'))
         if self._weight:
             weight_layer = param['layer'][self._weight]
             output = fluid.layers.elementwise_mul(output, weight_layer)
@@ -248,7 +305,11 @@ class LogLossLayer(Layer):
         # For AUC Metric
         metric = self._extend_output['metric_dict']
         binary_predict = fluid.layers.concat(
-            input=[fluid.layers.elementwise_sub(fluid.layers.ceil(norm), norm), norm], axis=1)
+            input=[
+                fluid.layers.elementwise_sub(fluid.layers.ceil(norm), norm),
+                norm
+            ],
+            axis=1)
         metric['auc']['var'], metric['batch_auc']['var'], [metric['batch_stat_pos']['var'], \
                                                            metric['batch_stat_neg']['var'], metric['stat_pos']['var'],
                                                            metric['stat_neg']['var']] = \

core/reader.py

@@ -11,6 +11,7 @@
 # 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.
+
 from __future__ import print_function
 
 import abc
@@ -44,3 +45,65 @@ class Reader(dg.MultiSlotDataGenerator):
     @abc.abstractmethod
     def generate_sample(self, line):
         pass
+
+
+class SlotReader(dg.MultiSlotDataGenerator):
+    __metaclass__ = abc.ABCMeta
+
+    def __init__(self, config):
+        dg.MultiSlotDataGenerator.__init__(self)
+        if os.path.isfile(config):
+            with open(config, 'r') as rb:
+                _config = yaml.load(rb.read(), Loader=yaml.FullLoader)
+        else:
+            raise ValueError("reader config only support yaml")
+        envs.set_global_envs(_config)
+        envs.update_workspace()
+
+    def init(self, sparse_slots, dense_slots, padding=0):
+        from operator import mul
+        self.sparse_slots = sparse_slots.strip().split(" ")
+        self.dense_slots = dense_slots.strip().split(" ")
+        self.dense_slots_shape = [
+            reduce(mul,
+                   [int(j) for j in i.split(":")[1].strip("[]").split(",")])
+            for i in self.dense_slots
+        ]
+        self.dense_slots = [i.split(":")[0] for i in self.dense_slots]
+        self.slots = self.dense_slots + self.sparse_slots
+        self.slot2index = {}
+        self.visit = {}
+        for i in range(len(self.slots)):
+            self.slot2index[self.slots[i]] = i
+            self.visit[self.slots[i]] = False
+        self.padding = padding
+
+    def generate_sample(self, l):
+        def reader():
+            line = l.strip().split(" ")
+            output = [(i, []) for i in self.slots]
+            for i in line:
+                slot_feasign = i.split(":")
+                slot = slot_feasign[0]
+                if slot not in self.slots:
+                    continue
+                if slot in self.sparse_slots:
+                    feasign = int(slot_feasign[1])
+                else:
+                    feasign = float(slot_feasign[1])
+                output[self.slot2index[slot]][1].append(feasign)
+                self.visit[slot] = True
+            for i in self.visit:
+                slot = i
+                if not self.visit[slot]:
+                    if i in self.dense_slots:
+                        output[self.slot2index[i]][1].extend(
+                            [self.padding] *
+                            self.dense_slots_shape[self.slot2index[i]])
+                    else:
+                        output[self.slot2index[i]][1].extend([self.padding])
+                else:
+                    self.visit[slot] = False
+            yield output
+
+        return reader

core/trainer.py

@@ -30,8 +30,10 @@ class Trainer(object):
 
     def __init__(self, config=None):
         self._status_processor = {}
+
         self._place = fluid.CPUPlace()
         self._exe = fluid.Executor(self._place)
+
         self._exector_context = {}
         self._context = {'status': 'uninit', 'is_exit': False}
         self._config_yaml = config
@@ -95,6 +97,6 @@ def user_define_engine(engine_yaml):
     train_dirname = os.path.dirname(train_location)
     base_name = os.path.splitext(os.path.basename(train_location))[0]
     sys.path.append(train_dirname)
-    trainer_class = envs.lazy_instance_by_fliename(
-        base_name, "UserDefineTraining")
+    trainer_class = envs.lazy_instance_by_fliename(base_name,
+                                                   "UserDefineTraining")
     return trainer_class

core/trainers/__init__.py

@@ -11,7 +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.
-
 """
 trainer implement.
 
@@ -22,5 +21,3 @@ Trainer
                                ↘  (for online learning training) OnlineLearningTrainer
 
 """
-
-

core/trainers/cluster_trainer.py

@@ -11,7 +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.
-
 """
 Training use fluid with one node only.
 """
@@ -43,11 +42,14 @@ class ClusterTrainer(TranspileTrainer):
             self.regist_context_processor('uninit', self.instance)
             self.regist_context_processor('init_pass', self.init)
             self.regist_context_processor('startup_pass', self.startup)
-            if envs.get_platform() == "LINUX" and envs.get_global_env("dataset_class", None, "train.reader") != "DataLoader":
+
+            if envs.get_platform() == "LINUX" and envs.get_global_env(
+                    "dataset_class", None, "train.reader") != "DataLoader":
                 self.regist_context_processor('train_pass', self.dataset_train)
             else:
-                self.regist_context_processor(
-                    'train_pass', self.dataloader_train)
+                self.regist_context_processor('train_pass',
+                                              self.dataloader_train)
+
             self.regist_context_processor('infer_pass', self.infer)
             self.regist_context_processor('terminal_pass', self.terminal)
 
@@ -75,8 +77,8 @@ class ClusterTrainer(TranspileTrainer):
     def init(self, context):
         self.model.train_net()
         optimizer = self.model.optimizer()
-        optimizer_name = envs.get_global_env(
-            "hyper_parameters.optimizer", None, "train.model")
+        optimizer_name = envs.get_global_env("hyper_parameters.optimizer",
+                                             None, "train.model")
         if optimizer_name not in ["", "sgd", "SGD", "Sgd"]:
             os.environ["FLAGS_communicator_is_sgd_optimizer"] = '0'
 
@@ -114,9 +116,9 @@ class ClusterTrainer(TranspileTrainer):
 
         program = fluid.compiler.CompiledProgram(
             fleet.main_program).with_data_parallel(
-            loss_name=self.model.get_avg_cost().name,
-            build_strategy=self.strategy.get_build_strategy(),
-            exec_strategy=self.strategy.get_execute_strategy())
+                loss_name=self.model.get_avg_cost().name,
+                build_strategy=self.strategy.get_build_strategy(),
+                exec_strategy=self.strategy.get_execute_strategy())
 
         metrics_varnames = []
         metrics_format = []
@@ -135,9 +137,8 @@ class ClusterTrainer(TranspileTrainer):
             batch_id = 0
             try:
                 while True:
-                    metrics_rets = self._exe.run(
-                        program=program,
-                        fetch_list=metrics_varnames)
+                    metrics_rets = self._exe.run(program=program,
+                                                 fetch_list=metrics_varnames)
 
                     metrics = [epoch, batch_id]
                     metrics.extend(metrics_rets)
@@ -162,14 +163,16 @@ class ClusterTrainer(TranspileTrainer):
 
         for i in range(epochs):
             begin_time = time.time()
-            self._exe.train_from_dataset(program=fluid.default_main_program(),
-                                         dataset=dataset,
-                                         fetch_list=self.fetch_vars,
-                                         fetch_info=self.fetch_alias,
-                                         print_period=self.fetch_period)
+            self._exe.train_from_dataset(
+                program=fluid.default_main_program(),
+                dataset=dataset,
+                fetch_list=self.fetch_vars,
+                fetch_info=self.fetch_alias,
+                print_period=self.fetch_period)
             end_time = time.time()
-            times = end_time-begin_time
-            print("epoch {} using time {}, speed {:.2f} lines/s".format(i, times, ins/times))
+            times = end_time - begin_time
+            print("epoch {} using time {}, speed {:.2f} lines/s".format(
+                i, times, ins / times))
 
             self.save(i, "train", is_fleet=True)
         fleet.stop_worker()

core/trainers/ctr_coding_trainer.py

@@ -59,8 +59,10 @@ class CtrTrainer(Trainer):
         reader_class = envs.get_global_env("class", None, namespace)
         abs_dir = os.path.dirname(os.path.abspath(__file__))
         reader = os.path.join(abs_dir, '../utils', 'dataset_instance.py')
-        pipe_cmd = "python {} {} {} {}".format(reader, reader_class, "TRAIN", self._config_yaml)
-        train_data_path = envs.get_global_env("train_data_path", None, namespace)
+        pipe_cmd = "python {} {} {} {}".format(reader, reader_class, "TRAIN",
+                                               self._config_yaml)
+        train_data_path = envs.get_global_env("train_data_path", None,
+                                              namespace)
 
         dataset = fluid.DatasetFactory().create_dataset()
         dataset.set_use_var(inputs)
@@ -87,7 +89,8 @@ class CtrTrainer(Trainer):
         self.model.train_net()
         optimizer = self.model.optimizer()
 
-        optimizer = fleet.distributed_optimizer(optimizer, strategy={"use_cvm": False})
+        optimizer = fleet.distributed_optimizer(
+            optimizer, strategy={"use_cvm": False})
         optimizer.minimize(self.model.get_avg_cost())
 
         if fleet.is_server():
@@ -118,16 +121,18 @@ class CtrTrainer(Trainer):
         gs = shuf * 0
         fleet._role_maker._node_type_comm.Allreduce(shuf, gs)
 
-        print("trainer id: {}, trainers: {}, gs: {}".format(fleet.worker_index(), fleet.worker_num(), gs))
+        print("trainer id: {}, trainers: {}, gs: {}".format(fleet.worker_index(
+        ), fleet.worker_num(), gs))
 
         epochs = envs.get_global_env("train.epochs")
 
         for i in range(epochs):
-            self._exe.train_from_dataset(program=fluid.default_main_program(),
-                                         dataset=dataset,
-                                         fetch_list=self.fetch_vars,
-                                         fetch_info=self.fetch_alias,
-                                         print_period=self.fetch_period)
+            self._exe.train_from_dataset(
+                program=fluid.default_main_program(),
+                dataset=dataset,
+                fetch_list=self.fetch_vars,
+                fetch_info=self.fetch_alias,
+                print_period=self.fetch_period)
 
         context['status'] = 'terminal_pass'
         fleet.stop_worker()

core/trainers/online_learning_trainer.py

@@ -11,7 +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.
-
 """
 Training use fluid with one node only.
 """
@@ -44,11 +43,14 @@ class OnlineLearningTrainer(TranspileTrainer):
             self.regist_context_processor('uninit', self.instance)
             self.regist_context_processor('init_pass', self.init)
             self.regist_context_processor('startup_pass', self.startup)
-            if envs.get_platform() == "LINUX" and envs.get_global_env("dataset_class", None, "train.reader") != "DataLoader":
+
+            if envs.get_platform() == "LINUX" and envs.get_global_env(
+                    "dataset_class", None, "train.reader") != "DataLoader":
                 self.regist_context_processor('train_pass', self.dataset_train)
             else:
-                self.regist_context_processor(
-                    'train_pass', self.dataloader_train)
+                self.regist_context_processor('train_pass',
+                                              self.dataloader_train)
+
             self.regist_context_processor('infer_pass', self.infer)
             self.regist_context_processor('terminal_pass', self.terminal)
 
@@ -110,27 +112,27 @@ class OnlineLearningTrainer(TranspileTrainer):
         if state == "TRAIN":
             inputs = self.model.get_inputs()
             namespace = "train.reader"
-            train_data_path = envs.get_global_env(
-                "train_data_path", None, namespace)
+            train_data_path = envs.get_global_env("train_data_path", None,
+                                                  namespace)
         else:
             inputs = self.model.get_infer_inputs()
             namespace = "evaluate.reader"
-            train_data_path = envs.get_global_env(
-                "test_data_path", None, namespace)
+            train_data_path = envs.get_global_env("test_data_path", None,
+                                                  namespace)
 
         threads = int(envs.get_runtime_environ("train.trainer.threads"))
         batch_size = envs.get_global_env("batch_size", None, namespace)
         reader_class = envs.get_global_env("class", None, namespace)
         abs_dir = os.path.dirname(os.path.abspath(__file__))
         reader = os.path.join(abs_dir, '../utils', 'dataset_instance.py')
-        pipe_cmd = "python {} {} {} {}".format(
-            reader, reader_class, state, self._config_yaml)
+        pipe_cmd = "python {} {} {} {}".format(reader, reader_class, state,
+                                               self._config_yaml)
 
         if train_data_path.startswith("paddlerec::"):
             package_base = envs.get_runtime_environ("PACKAGE_BASE")
             assert package_base is not None
-            train_data_path = os.path.join(
-                package_base, train_data_path.split("::")[1])
+            train_data_path = os.path.join(package_base,
+                                           train_data_path.split("::")[1])
 
         dataset = fluid.DatasetFactory().create_dataset()
         dataset.set_use_var(inputs)
@@ -166,14 +168,16 @@ class OnlineLearningTrainer(TranspileTrainer):
                 ins = self._get_dataset_ins()
 
                 begin_time = time.time()
-                self._exe.train_from_dataset(program=fluid.default_main_program(),
-                                             dataset=dataset,
-                                             fetch_list=self.fetch_vars,
-                                             fetch_info=self.fetch_alias,
-                                             print_period=self.fetch_period)
+                self._exe.train_from_dataset(
+                    program=fluid.default_main_program(),
+                    dataset=dataset,
+                    fetch_list=self.fetch_vars,
+                    fetch_info=self.fetch_alias,
+                    print_period=self.fetch_period)
                 end_time = time.time()
-                times = end_time-begin_time
-                print("epoch {} using time {}, speed {:.2f} lines/s".format(i, times, ins/times))
+                times = end_time - begin_time
+                print("epoch {} using time {}, speed {:.2f} lines/s".format(
+                    i, times, ins / times))
                 self.save(i, "train", is_fleet=True)
 
         fleet.stop_worker()

core/trainers/single_trainer.py

@@ -11,7 +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.
-
 """
 Training use fluid with one node only.
 """
@@ -36,8 +35,9 @@ class SingleTrainer(TranspileTrainer):
         self.regist_context_processor('uninit', self.instance)
         self.regist_context_processor('init_pass', self.init)
         self.regist_context_processor('startup_pass', self.startup)
-        if envs.get_platform() == "LINUX" and envs.get_global_env("dataset_class", None,
-                                                                  "train.reader") != "DataLoader":
+
+        if envs.get_platform() == "LINUX" and envs.get_global_env(
+                "dataset_class", None, "train.reader") != "DataLoader":
             self.regist_context_processor('train_pass', self.dataset_train)
         else:
             self.regist_context_processor('train_pass', self.dataloader_train)
@@ -73,9 +73,8 @@ class SingleTrainer(TranspileTrainer):
         reader = self._get_dataloader("TRAIN")
         epochs = envs.get_global_env("train.epochs")
 
-        program = fluid.compiler.CompiledProgram(
-            fluid.default_main_program()).with_data_parallel(
-            loss_name=self.model.get_avg_cost().name)
+        program = fluid.compiler.CompiledProgram(fluid.default_main_program(
+        )).with_data_parallel(loss_name=self.model.get_avg_cost().name)
 
         metrics_varnames = []
         metrics_format = []
@@ -94,9 +93,8 @@ class SingleTrainer(TranspileTrainer):
             batch_id = 0
             try:
                 while True:
-                    metrics_rets = self._exe.run(
-                        program=program,
-                        fetch_list=metrics_varnames)
+                    metrics_rets = self._exe.run(program=program,
+                                                 fetch_list=metrics_varnames)
 
                     metrics = [epoch, batch_id]
                     metrics.extend(metrics_rets)
@@ -117,14 +115,16 @@ class SingleTrainer(TranspileTrainer):
         epochs = envs.get_global_env("train.epochs")
         for i in range(epochs):
             begin_time = time.time()
-            self._exe.train_from_dataset(program=fluid.default_main_program(),
-                                         dataset=dataset,
-                                         fetch_list=self.fetch_vars,
-                                         fetch_info=self.fetch_alias,
-                                         print_period=self.fetch_period)
+            self._exe.train_from_dataset(
+                program=fluid.default_main_program(),
+                dataset=dataset,
+                fetch_list=self.fetch_vars,
+                fetch_info=self.fetch_alias,
+                print_period=self.fetch_period)
             end_time = time.time()
             times = end_time - begin_time
-            print("epoch {} using time {}, speed {:.2f} lines/s".format(i, times, ins / times))
+            print("epoch {} using time {}, speed {:.2f} lines/s".format(
+                i, times, ins / times))
 
             self.save(i, "train", is_fleet=False)
         context['status'] = 'infer_pass'

core/trainers/tdm_cluster_trainer.py

@@ -11,7 +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.
-
 """
 Training use fluid with one node only.
 """
@@ -36,8 +35,8 @@ special_param = ["TDM_Tree_Travel", "TDM_Tree_Layer", "TDM_Tree_Info"]
 class TDMClusterTrainer(ClusterTrainer):
     def server(self, context):
         namespace = "train.startup"
-        init_model_path = envs.get_global_env(
-            "cluster.init_model_path", "", namespace)
+        init_model_path = envs.get_global_env("cluster.init_model_path", "",
+                                              namespace)
         assert init_model_path != "", "Cluster train must has init_model for TDM"
         fleet.init_server(init_model_path)
         logger.info("TDM: load model from {}".format(init_model_path))
@@ -48,24 +47,27 @@ class TDMClusterTrainer(ClusterTrainer):
         self._exe.run(fleet.startup_program)
 
         namespace = "train.startup"
-        load_tree = envs.get_global_env(
-            "tree.load_tree", True, namespace)
-        self.tree_layer_path = envs.get_global_env(
-            "tree.tree_layer_path", "", namespace)
-        self.tree_travel_path = envs.get_global_env(
-            "tree.tree_travel_path", "", namespace)
-        self.tree_info_path = envs.get_global_env(
-            "tree.tree_info_path", "", namespace)
-
-        save_init_model = envs.get_global_env(
-            "cluster.save_init_model", False, namespace)
-        init_model_path = envs.get_global_env(
-            "cluster.init_model_path", "", namespace)
+        load_tree = envs.get_global_env("tree.load_tree", True, namespace)
+
+        self.tree_layer_path = envs.get_global_env("tree.tree_layer_path", "",
+                                                   namespace)
+
+        self.tree_travel_path = envs.get_global_env("tree.tree_travel_path",
+                                                    "", namespace)
+
+        self.tree_info_path = envs.get_global_env("tree.tree_info_path", "",
+                                                  namespace)
+
+        save_init_model = envs.get_global_env("cluster.save_init_model", False,
+                                              namespace)
+        init_model_path = envs.get_global_env("cluster.init_model_path", "",
+                                              namespace)
 
         if load_tree:
             # covert tree to tensor, set it into Fluid's variable.
             for param_name in special_param:
-                param_t = fluid.global_scope().find_var(param_name).get_tensor()
+                param_t = fluid.global_scope().find_var(param_name).get_tensor(
+                )
                 param_array = self._tdm_prepare(param_name)
                 param_t.set(param_array.astype('int32'), self._place)
 
@@ -93,8 +95,8 @@ class TDMClusterTrainer(ClusterTrainer):
     def _tdm_travel_prepare(self):
         """load tdm tree param from npy/list file"""
         travel_array = np.load(self.tree_travel_path)
-        logger.info("TDM Tree leaf node nums: {}".format(
-            travel_array.shape[0]))
+        logger.info("TDM Tree leaf node nums: {}".format(travel_array.shape[
+            0]))
         return travel_array
 
     def _tdm_layer_prepare(self):

core/trainers/tdm_single_trainer.py

@@ -11,7 +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.
-
 """
 Training use fluid with one node only.
 """
@@ -27,33 +26,38 @@ from paddlerec.core.utils import envs
 logging.basicConfig(format="%(asctime)s - %(levelname)s - %(message)s")
 logger = logging.getLogger("fluid")
 logger.setLevel(logging.INFO)
-special_param = ["TDM_Tree_Travel", "TDM_Tree_Layer",
-                 "TDM_Tree_Info", "TDM_Tree_Emb"]
+special_param = [
+    "TDM_Tree_Travel", "TDM_Tree_Layer", "TDM_Tree_Info", "TDM_Tree_Emb"
+]
 
 
 class TDMSingleTrainer(SingleTrainer):
     def startup(self, context):
         namespace = "train.startup"
-        load_persistables = envs.get_global_env(
-            "single.load_persistables", False, namespace)
+        load_persistables = envs.get_global_env("single.load_persistables",
+                                                False, namespace)
+
         persistables_model_path = envs.get_global_env(
             "single.persistables_model_path", "", namespace)
 
-        load_tree = envs.get_global_env(
-            "tree.load_tree", False, namespace)
-        self.tree_layer_path = envs.get_global_env(
-            "tree.tree_layer_path", "", namespace)
-        self.tree_travel_path = envs.get_global_env(
-            "tree.tree_travel_path", "", namespace)
-        self.tree_info_path = envs.get_global_env(
-            "tree.tree_info_path", "", namespace)
-        self.tree_emb_path = envs.get_global_env(
-            "tree.tree_emb_path", "", namespace)
-
-        save_init_model = envs.get_global_env(
-            "single.save_init_model", False, namespace)
-        init_model_path = envs.get_global_env(
-            "single.init_model_path", "", namespace)
+        load_tree = envs.get_global_env("tree.load_tree", False, namespace)
+
+        self.tree_layer_path = envs.get_global_env("tree.tree_layer_path", "",
+                                                   namespace)
+
+        self.tree_travel_path = envs.get_global_env("tree.tree_travel_path",
+                                                    "", namespace)
+
+        self.tree_info_path = envs.get_global_env("tree.tree_info_path", "",
+                                                  namespace)
+
+        self.tree_emb_path = envs.get_global_env("tree.tree_emb_path", "",
+                                                 namespace)
+
+        save_init_model = envs.get_global_env("single.save_init_model", False,
+                                              namespace)
+        init_model_path = envs.get_global_env("single.init_model_path", "",
+                                              namespace)
         self._exe.run(fluid.default_startup_program())
 
         if load_persistables:
@@ -68,7 +72,8 @@ class TDMSingleTrainer(SingleTrainer):
         if load_tree:
             # covert tree to tensor, set it into Fluid's variable.
             for param_name in special_param:
-                param_t = fluid.global_scope().find_var(param_name).get_tensor()
+                param_t = fluid.global_scope().find_var(param_name).get_tensor(
+                )
                 param_array = self._tdm_prepare(param_name)
                 if param_name == 'TDM_Tree_Emb':
                     param_t.set(param_array.astype('float32'), self._place)
@@ -102,15 +107,15 @@ class TDMSingleTrainer(SingleTrainer):
     def _tdm_travel_prepare(self):
         """load tdm tree param from npy/list file"""
         travel_array = np.load(self.tree_travel_path)
-        logger.info("TDM Tree leaf node nums: {}".format(
-            travel_array.shape[0]))
+        logger.info("TDM Tree leaf node nums: {}".format(travel_array.shape[
+            0]))
         return travel_array
 
     def _tdm_emb_prepare(self):
         """load tdm tree param from npy/list file"""
         emb_array = np.load(self.tree_emb_path)
-        logger.info("TDM Tree node nums from emb: {}".format(
-            emb_array.shape[0]))
+        logger.info("TDM Tree node nums from emb: {}".format(emb_array.shape[
+            0]))
         return emb_array
 
     def _tdm_layer_prepare(self):

core/trainers/transpiler_trainer.py

@@ -11,7 +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.
-
 """
 Training use fluid with DistributeTranspiler
 """
@@ -23,6 +22,7 @@ from paddle.fluid.incubate.fleet.parameter_server.distribute_transpiler import f
 from paddlerec.core.trainer import Trainer
 from paddlerec.core.utils import envs
 from paddlerec.core.utils import dataloader_instance
+from paddlerec.core.reader import SlotReader
 
 
 class TranspileTrainer(Trainer):
@@ -38,9 +38,12 @@ class TranspileTrainer(Trainer):
         self.increment_models = []
 
     def processor_register(self):
-        print("Need implement by trainer, `self.regist_context_processor('uninit', self.instance)` must be the first")
+        print(
+            "Need implement by trainer, `self.regist_context_processor('uninit', self.instance)` must be the first"
+        )
 
     def _get_dataloader(self, state="TRAIN"):
+
         if state == "TRAIN":
             dataloader = self.model._data_loader
             namespace = "train.reader"
@@ -50,14 +53,24 @@ class TranspileTrainer(Trainer):
             namespace = "evaluate.reader"
             class_name = "EvaluateReader"
 
+        sparse_slots = envs.get_global_env("sparse_slots", None, namespace)
+        dense_slots = envs.get_global_env("dense_slots", None, namespace)
+
         batch_size = envs.get_global_env("batch_size", None, namespace)
-        reader_class = envs.get_global_env("class", None, namespace)
         print("batch_size: {}".format(batch_size))
-        reader = dataloader_instance.dataloader(
-            reader_class, state, self._config_yaml)
 
-        reader_class = envs.lazy_instance_by_fliename(reader_class, class_name)
-        reader_ins = reader_class(self._config_yaml)
+        if sparse_slots is None and dense_slots is None:
+            reader_class = envs.get_global_env("class", None, namespace)
+            reader = dataloader_instance.dataloader(reader_class, state,
+                                                    self._config_yaml)
+            reader_class = envs.lazy_instance_by_fliename(reader_class,
+                                                          class_name)
+            reader_ins = reader_class(self._config_yaml)
+        else:
+            reader = dataloader_instance.slotdataloader("", state,
+                                                        self._config_yaml)
+            reader_ins = SlotReader(self._config_yaml)
+
         if hasattr(reader_ins, 'generate_batch_from_trainfiles'):
             dataloader.set_sample_list_generator(reader)
         else:
@@ -85,27 +98,37 @@ class TranspileTrainer(Trainer):
         if state == "TRAIN":
             inputs = self.model.get_inputs()
             namespace = "train.reader"
-            train_data_path = envs.get_global_env(
-                "train_data_path", None, namespace)
+            train_data_path = envs.get_global_env("train_data_path", None,
+                                                  namespace)
         else:
             inputs = self.model.get_infer_inputs()
             namespace = "evaluate.reader"
-            train_data_path = envs.get_global_env(
-                "test_data_path", None, namespace)
+            train_data_path = envs.get_global_env("test_data_path", None,
+                                                  namespace)
+
+        sparse_slots = envs.get_global_env("sparse_slots", None, namespace)
+        dense_slots = envs.get_global_env("dense_slots", None, namespace)
 
         threads = int(envs.get_runtime_environ("train.trainer.threads"))
         batch_size = envs.get_global_env("batch_size", None, namespace)
         reader_class = envs.get_global_env("class", None, namespace)
         abs_dir = os.path.dirname(os.path.abspath(__file__))
         reader = os.path.join(abs_dir, '../utils', 'dataset_instance.py')
-        pipe_cmd = "python {} {} {} {}".format(
-            reader, reader_class, state, self._config_yaml)
+
+        if sparse_slots is None and dense_slots is None:
+            pipe_cmd = "python {} {} {} {}".format(reader, reader_class, state,
+                                                   self._config_yaml)
+        else:
+            padding = envs.get_global_env("padding", 0, namespace)
+            pipe_cmd = "python {} {} {} {} {} {} {} {}".format(
+                reader, "slot", "slot", self._config_yaml, namespace, \
+                sparse_slots.replace(" ", "#"), dense_slots.replace(" ", "#"), str(padding))
 
         if train_data_path.startswith("paddlerec::"):
             package_base = envs.get_runtime_environ("PACKAGE_BASE")
             assert package_base is not None
-            train_data_path = os.path.join(
-                package_base, train_data_path.split("::")[1])
+            train_data_path = os.path.join(package_base,
+                                           train_data_path.split("::")[1])
 
         dataset = fluid.DatasetFactory().create_dataset()
         dataset.set_use_var(inputs)
@@ -121,11 +144,11 @@ class TranspileTrainer(Trainer):
 
         debug_mode = envs.get_global_env("reader_debug_mode", False, namespace)
         if debug_mode:
-            print(
-                "--- Dataset Debug Mode Begin , show pre 10 data of {}---".format(file_list[0]))
+            print("--- Dataset Debug Mode Begin , show pre 10 data of {}---".
+                  format(file_list[0]))
             os.system("cat {} | {} | head -10".format(file_list[0], pipe_cmd))
-            print(
-                "--- Dataset Debug Mode End , show pre 10 data of {}---".format(file_list[0]))
+            print("--- Dataset Debug Mode End , show pre 10 data of {}---".
+                  format(file_list[0]))
             exit(0)
 
         return dataset
@@ -147,30 +170,29 @@ class TranspileTrainer(Trainer):
             if not need_save(epoch_id, save_interval, False):
                 return
 
-            #  print("save inference model is not supported now.")
-            #  return
-
-            feed_varnames = envs.get_global_env(
-                "save.inference.feed_varnames", None, namespace)
+            feed_varnames = envs.get_global_env("save.inference.feed_varnames",
+                                                None, namespace)
             fetch_varnames = envs.get_global_env(
                 "save.inference.fetch_varnames", None, namespace)
             if feed_varnames is None or fetch_varnames is None:
                 return
 
-            fetch_vars = [fluid.default_main_program().global_block().vars[varname]
-                          for varname in fetch_varnames]
-            dirname = envs.get_global_env(
-                "save.inference.dirname", None, namespace)
+            fetch_vars = [
+                fluid.default_main_program().global_block().vars[varname]
+                for varname in fetch_varnames
+            ]
+            dirname = envs.get_global_env("save.inference.dirname", None,
+                                          namespace)
 
             assert dirname is not None
             dirname = os.path.join(dirname, str(epoch_id))
 
             if is_fleet:
-                fleet.save_inference_model(
-                    self._exe, dirname, feed_varnames, fetch_vars)
+                fleet.save_inference_model(self._exe, dirname, feed_varnames,
+                                           fetch_vars)
             else:
-                fluid.io.save_inference_model(
-                    dirname, feed_varnames, fetch_vars, self._exe)
+                fluid.io.save_inference_model(dirname, feed_varnames,
+                                              fetch_vars, self._exe)
             self.inference_models.append((epoch_id, dirname))
 
         def save_persistables():
@@ -180,8 +202,8 @@ class TranspileTrainer(Trainer):
             if not need_save(epoch_id, save_interval, False):
                 return
 
-            dirname = envs.get_global_env(
-                "save.increment.dirname", None, namespace)
+            dirname = envs.get_global_env("save.increment.dirname", None,
+                                          namespace)
 
             assert dirname is not None
             dirname = os.path.join(dirname, str(epoch_id))
@@ -259,10 +281,9 @@ class TranspileTrainer(Trainer):
             batch_id = 0
             try:
                 while True:
-                    metrics_rets = self._exe.run(
-                        program=program,
-                        fetch_list=metrics_varnames,
-                        return_numpy=is_return_numpy)
+                    metrics_rets = self._exe.run(program=program,
+                                                 fetch_list=metrics_varnames,
+                                                 return_numpy=is_return_numpy)
 
                     metrics = [epoch, batch_id]
                     metrics.extend(metrics_rets)

core/utils/dataloader_instance.py

@@ -18,6 +18,7 @@ import os
 from paddlerec.core.utils.envs import lazy_instance_by_fliename
 from paddlerec.core.utils.envs import get_global_env
 from paddlerec.core.utils.envs import get_runtime_environ
+from paddlerec.core.reader import SlotReader
 
 
 def dataloader(readerclass, train, yaml_file):
@@ -62,3 +63,49 @@ def dataloader(readerclass, train, yaml_file):
     if hasattr(reader, 'generate_batch_from_trainfiles'):
         return gen_batch_reader()
     return gen_reader
+
+
+def slotdataloader(readerclass, train, yaml_file):
+    if train == "TRAIN":
+        reader_name = "SlotReader"
+        namespace = "train.reader"
+        data_path = get_global_env("train_data_path", None, namespace)
+    else:
+        reader_name = "SlotReader"
+        namespace = "evaluate.reader"
+        data_path = get_global_env("test_data_path", None, namespace)
+
+    if data_path.startswith("paddlerec::"):
+        package_base = get_runtime_environ("PACKAGE_BASE")
+        assert package_base is not None
+        data_path = os.path.join(package_base, data_path.split("::")[1])
+
+    files = [str(data_path) + "/%s" % x for x in os.listdir(data_path)]
+
+    sparse = get_global_env("sparse_slots", None, namespace)
+    dense = get_global_env("dense_slots", None, namespace)
+    padding = get_global_env("padding", 0, namespace)
+    reader = SlotReader(yaml_file)
+    reader.init(sparse, dense, int(padding))
+
+    def gen_reader():
+        for file in files:
+            with open(file, 'r') as f:
+                for line in f:
+                    line = line.rstrip('\n')
+                    iter = reader.generate_sample(line)
+                    for parsed_line in iter():
+                        if parsed_line is None:
+                            continue
+                        else:
+                            values = []
+                            for pased in parsed_line:
+                                values.append(pased[1])
+                            yield values
+
+    def gen_batch_reader():
+        return reader.generate_batch_from_trainfiles(files)
+
+    if hasattr(reader, 'generate_batch_from_trainfiles'):
+        return gen_batch_reader()
+    return gen_reader

core/utils/dataset_holder.py

@@ -24,7 +24,7 @@ from paddlerec.core.utils import util as util
 
 class DatasetHolder(object):
     """
-    Dataset Base
+    Dataset Holder
     """
     __metaclass__ = abc.ABCMeta
 
@@ -74,11 +74,17 @@ class TimeSplitDatasetHolder(DatasetHolder):
         Dataset.__init__(self, config)
         if 'data_donefile' not in config or config['data_donefile'] is None:
             config['data_donefile'] = config['data_path'] + "/to.hadoop.done"
-        self._path_generator = util.PathGenerator({'templates': [
-            {'name': 'data_path', 'template': config['data_path']},
-            {'name': 'donefile_path', 'template': config['data_donefile']}
-        ]})
-        self._split_interval = config['split_interval']  # data split N mins per dir
+        self._path_generator = util.PathGenerator({
+            'templates': [{
+                'name': 'data_path',
+                'template': config['data_path']
+            }, {
+                'name': 'donefile_path',
+                'template': config['data_donefile']
+            }]
+        })
+        self._split_interval = config[
+            'split_interval']  # data split N mins per dir
         self._data_file_handler = fs.FileHandler(config)
 
     def _format_data_time(self, daytime_str, time_window_mins):
@@ -91,7 +97,8 @@ class TimeSplitDatasetHolder(DatasetHolder):
             return None, 0
 
         if mins_of_day % self._split_interval != 0:
-            skip_mins = self._split_interval - (mins_of_day % self._split_interval)
+            skip_mins = self._split_interval - (mins_of_day %
+                                                self._split_interval)
             data_time = data_time + datetime.timedelta(minutes=skip_mins)
             time_window_mins = time_window_mins - skip_mins
         return data_time, time_window_mins
@@ -106,17 +113,24 @@ class TimeSplitDatasetHolder(DatasetHolder):
             True/False
         """
         is_ready = True
-        data_time, windows_mins = self._format_data_time(daytime_str, time_window_mins)
+        data_time, windows_mins = self._format_data_time(daytime_str,
+                                                         time_window_mins)
         while time_window_mins > 0:
-            file_path = self._path_generator.generate_path('donefile_path', {'time_format': data_time})
+            file_path = self._path_generator.generate_path(
+                'donefile_path', {'time_format': data_time})
             if not self._data_file_handler.is_exist(file_path):
                 is_ready = False
                 break
             time_window_mins = time_window_mins - self._split_interval
-            data_time = data_time + datetime.timedelta(minutes=self._split_interval)
+            data_time = data_time + datetime.timedelta(
+                minutes=self._split_interval)
         return is_ready
 
-    def get_file_list(self, daytime_str, time_window_mins, node_num=1, node_idx=0):
+    def get_file_list(self,
+                      daytime_str,
+                      time_window_mins,
+                      node_num=1,
+                      node_idx=0):
         """
         data in  [daytime_str, daytime_str + time_window_mins], random shard to node_num, return shard[node_idx]
         Args:
@@ -128,26 +142,32 @@ class TimeSplitDatasetHolder(DatasetHolder):
             list, data_shard[node_idx]
         """
         data_file_list = []
-        data_time, windows_mins = self._format_data_time(daytime_str, time_window_mins)
+        data_time, windows_mins = self._format_data_time(daytime_str,
+                                                         time_window_mins)
         while time_window_mins > 0:
-            file_path = self._path_generator.generate_path('data_path', {'time_format': data_time})
+            file_path = self._path_generator.generate_path(
+                'data_path', {'time_format': data_time})
             sub_file_list = self._data_file_handler.ls(file_path)
             for sub_file in sub_file_list:
                 sub_file_name = self._data_file_handler.get_file_name(sub_file)
-                if not sub_file_name.startswith(self._config['filename_prefix']):
+                if not sub_file_name.startswith(self._config[
+                        'filename_prefix']):
                     continue
                 if hash(sub_file_name) % node_num == node_idx:
                     data_file_list.append(sub_file)
             time_window_mins = time_window_mins - self._split_interval
-            data_time = data_time + datetime.timedelta(minutes=self._split_interval)
+            data_time = data_time + datetime.timedelta(
+                minutes=self._split_interval)
         return data_file_list
 
     def _alloc_dataset(self, file_list):
         """ """
-        dataset = fluid.DatasetFactory().create_dataset(self._config['dataset_type'])
+        dataset = fluid.DatasetFactory().create_dataset(self._config[
+            'dataset_type'])
         dataset.set_batch_size(self._config['batch_size'])
         dataset.set_thread(self._config['load_thread'])
-        dataset.set_hdfs_config(self._config['fs_name'], self._config['fs_ugi'])
+        dataset.set_hdfs_config(self._config['fs_name'],
+                                self._config['fs_ugi'])
         dataset.set_pipe_command(self._config['data_converter'])
         dataset.set_filelist(file_list)
         dataset.set_use_var(self._config['data_vars'])
@@ -163,7 +183,9 @@ class TimeSplitDatasetHolder(DatasetHolder):
             while self.check_ready(begin_time, windown_min) == False:
                 print("dataset not ready, time:" + begin_time)
                 time.sleep(30)
-            file_list = self.get_file_list(begin_time, windown_min, params['node_num'], params['node_idx'])
+            file_list = self.get_file_list(begin_time, windown_min,
+                                           params['node_num'],
+                                           params['node_idx'])
             self._datasets[begin_time] = self._alloc_dataset(file_list)
             self._datasets[begin_time].load_into_memory()
         else:
@@ -176,9 +198,12 @@ class TimeSplitDatasetHolder(DatasetHolder):
         windown_min = params['time_window_min']
         if begin_time not in self._datasets:
             if self.check_ready(begin_time, windown_min):
-                file_list = self.get_file_list(begin_time, windown_min, params['node_num'], params['node_idx'])
+                file_list = self.get_file_list(begin_time, windown_min,
+                                               params['node_num'],
+                                               params['node_idx'])
                 self._datasets[begin_time] = self._alloc_dataset(file_list)
-                self._datasets[begin_time].preload_into_memory(self._config['preload_thread'])
+                self._datasets[begin_time].preload_into_memory(self._config[
+                    'preload_thread'])
                 return True
         return False
 

core/utils/dataset_instance.py

@@ -16,19 +16,34 @@ from __future__ import print_function
 import sys
 
 from paddlerec.core.utils.envs import lazy_instance_by_fliename
+from paddlerec.core.reader import SlotReader
 
-if len(sys.argv) != 4:
-    raise ValueError("reader only accept 3 argument: 1. reader_class 2.train/evaluate 3.yaml_abs_path")
+if len(sys.argv) < 4:
+    raise ValueError(
+        "reader only accept 3 argument: 1. reader_class 2.train/evaluate/slotreader 3.yaml_abs_path"
+    )
 
 reader_package = sys.argv[1]
 
-if sys.argv[2] == "TRAIN":
+if sys.argv[2].upper() == "TRAIN":
     reader_name = "TrainReader"
-else:
+elif sys.argv[2].upper() == "EVALUATE":
     reader_name = "EvaluateReader"
+else:
+    reader_name = "SlotReader"
+    namespace = sys.argv[4]
+    sparse_slots = sys.argv[5].replace("#", " ")
+    dense_slots = sys.argv[6].replace("#", " ")
+    padding = int(sys.argv[7])
 
 yaml_abs_path = sys.argv[3]
-reader_class = lazy_instance_by_fliename(reader_package, reader_name)
-reader = reader_class(yaml_abs_path)
-reader.init()
-reader.run_from_stdin()
+
+if reader_name != "SlotReader":
+    reader_class = lazy_instance_by_fliename(reader_package, reader_name)
+    reader = reader_class(yaml_abs_path)
+    reader.init()
+    reader.run_from_stdin()
+else:
+    reader = SlotReader(yaml_abs_path)
+    reader.init(sparse_slots, dense_slots, padding)
+    reader.run_from_stdin()

core/utils/envs.py

@@ -95,7 +95,7 @@ def path_adapter(path):
         l_p = path.split("paddlerec.")[1].replace(".", "/")
         return os.path.join(package, l_p)
     else:
-        return path 
+        return path
 
 
 def windows_path_converter(path):
@@ -159,8 +159,8 @@ def pretty_print_envs(envs, header=None):
 
 def lazy_instance_by_package(package, class_name):
     models = get_global_env("train.model.models")
-    model_package = __import__(
-        package, globals(), locals(), package.split("."))
+    model_package = __import__(package,
+                               globals(), locals(), package.split("."))
     instance = getattr(model_package, class_name)
     return instance
 
@@ -170,8 +170,8 @@ def lazy_instance_by_fliename(abs, class_name):
     sys.path.append(dirname)
     package = os.path.splitext(os.path.basename(abs))[0]
 
-    model_package = __import__(
-        package, globals(), locals(), package.split("."))
+    model_package = __import__(package,
+                               globals(), locals(), package.split("."))
     instance = getattr(model_package, class_name)
     return instance
 
@@ -189,8 +189,7 @@ def get_platform():
 
 def find_free_port():
     def __free_port():
-        with closing(socket.socket(socket.AF_INET,
-                                   socket.SOCK_STREAM)) as s:
+        with closing(socket.socket(socket.AF_INET, socket.SOCK_STREAM)) as s:
             s.bind(('', 0))
             return s.getsockname()[1]
 

core/utils/fs.py

@@ -18,7 +18,7 @@ from paddle.fluid.incubate.fleet.utils.hdfs import HDFSClient
 
 
 def is_afs_path(path):
-    """R 
+    """is_afs_path
     """
     if path.startswith("afs") or path.startswith("hdfs"):
         return True
@@ -133,8 +133,9 @@ class FileHandler(object):
             if mode.find('a') >= 0:
                 org_content = self._hdfs_client.cat(dest_path)
             content = content + org_content
-            self._local_fs_client.write(content, temp_local_file,
-                                        mode)  # fleet hdfs_client only support upload, so write tmp file
+            self._local_fs_client.write(
+                content, temp_local_file, mode
+            )  # fleet hdfs_client only support upload, so write tmp file
             self._hdfs_client.delete(dest_path + ".tmp")
             self._hdfs_client.upload(dest_path + ".tmp", temp_local_file)
             self._hdfs_client.delete(dest_path + ".bak")
@@ -158,7 +159,8 @@ class FileHandler(object):
         files = []
         if is_afs_path(path):
             files = self._hdfs_client.ls(path)
-            files = [path + '/' + self.get_file_name(fi) for fi in files]  # absulte path
+            files = [path + '/' + self.get_file_name(fi)
+                     for fi in files]  # absulte path
         else:
             files = self._local_fs_client.ls(path)
             files = [path + '/' + fi for fi in files]  # absulte path

core/utils/util.py

@@ -22,6 +22,7 @@ from paddlerec.core.utils import fs as fs
 
 
 def save_program_proto(path, program=None):
+
     if program is None:
         _program = fluid.default_main_program()
     else:
@@ -175,7 +176,8 @@ class PathGenerator(object):
         """
         if template_name in self._templates:
             if 'time_format' in param:
-                str = param['time_format'].strftime(self._templates[template_name])
+                str = param['time_format'].strftime(self._templates[
+                    template_name])
                 return str.format(**param)
             return self._templates[template_name].format(**param)
         else:
@@ -198,31 +200,39 @@ class TimeTrainPass(object):
             self._begin_day = make_datetime(day_fields[0].strip())
             if len(day_fields) == 1 or len(day_fields[1]) == 0:
                 # 100 years, meaning to continuous running
-                self._end_day = self._begin_day + datetime.timedelta(days=36500)
+                self._end_day = self._begin_day + datetime.timedelta(
+                    days=36500)
             else:
                 # example: 2020212+10 
                 run_day = int(day_fields[1].strip())
-                self._end_day = self._begin_day + datetime.timedelta(days=run_day)
+                self._end_day = self._begin_day + datetime.timedelta(
+                    days=run_day)
         else:
             # example: {20191001..20191031}
-            days = os.popen("echo -n " + self._config['days']).read().split(" ")
+            days = os.popen("echo -n " + self._config['days']).read().split(
+                " ")
             self._begin_day = make_datetime(days[0])
             self._end_day = make_datetime(days[len(days) - 1])
         self._checkpoint_interval = self._config['checkpoint_interval']
         self._dump_inference_interval = self._config['dump_inference_interval']
-        self._interval_per_pass = self._config['train_time_interval']  # train N min data per pass
+        self._interval_per_pass = self._config[
+            'train_time_interval']  # train N min data per pass
 
         self._pass_id = 0
         self._inference_pass_id = 0
         self._pass_donefile_handler = None
         if 'pass_donefile_name' in self._config:
-            self._train_pass_donefile = global_config['output_path'] + '/' + self._config['pass_donefile_name']
+            self._train_pass_donefile = global_config[
+                'output_path'] + '/' + self._config['pass_donefile_name']
             if fs.is_afs_path(self._train_pass_donefile):
-                self._pass_donefile_handler = fs.FileHandler(global_config['io']['afs'])
+                self._pass_donefile_handler = fs.FileHandler(global_config[
+                    'io']['afs'])
             else:
-                self._pass_donefile_handler = fs.FileHandler(global_config['io']['local_fs'])
+                self._pass_donefile_handler = fs.FileHandler(global_config[
+                    'io']['local_fs'])
 
-            last_done = self._pass_donefile_handler.cat(self._train_pass_donefile).strip().split('\n')[-1]
+            last_done = self._pass_donefile_handler.cat(
+                self._train_pass_donefile).strip().split('\n')[-1]
             done_fileds = last_done.split('\t')
             if len(done_fileds) > 4:
                 self._base_key = done_fileds[1]
@@ -236,15 +246,18 @@ class TimeTrainPass(object):
         """
         return 24 * 60 / self._interval_per_pass
 
-    def save_train_progress(self, day, pass_id, base_key, model_path, is_checkpoint):
+    def save_train_progress(self, day, pass_id, base_key, model_path,
+                            is_checkpoint):
         """R
         """
         if is_checkpoint:
             self._checkpoint_pass_id = pass_id
             self._checkpoint_model_path = model_path
-        done_content = "%s\t%s\t%s\t%s\t%d\n" % (day, base_key,
-                                                 self._checkpoint_model_path, self._checkpoint_pass_id, pass_id)
-        self._pass_donefile_handler.write(done_content, self._train_pass_donefile, 'a')
+        done_content = "%s\t%s\t%s\t%s\t%d\n" % (
+            day, base_key, self._checkpoint_model_path,
+            self._checkpoint_pass_id, pass_id)
+        self._pass_donefile_handler.write(done_content,
+                                          self._train_pass_donefile, 'a')
         pass
 
     def init_pass_by_id(self, date_str, pass_id):
@@ -286,12 +299,14 @@ class TimeTrainPass(object):
         if self._pass_id < 1:
             self.init_pass_by_time(self._begin_day.strftime("%Y%m%d%H%M"))
         else:
-            next_time = self._current_train_time + datetime.timedelta(minutes=self._interval_per_pass)
+            next_time = self._current_train_time + datetime.timedelta(
+                minutes=self._interval_per_pass)
             if (next_time - self._end_day).total_seconds() > 0:
                 has_next = False
             else:
                 self.init_pass_by_time(next_time.strftime("%Y%m%d%H%M"))
-        if has_next and (self._inference_pass_id < self._pass_id or self._pass_id < old_pass_id):
+        if has_next and (self._inference_pass_id < self._pass_id or
+                         self._pass_id < old_pass_id):
             self._inference_pass_id = self._pass_id - 1
         return has_next
 
@@ -319,9 +334,11 @@ class TimeTrainPass(object):
         Return:
             date(current_train_time + delta_day)
         """
-        return (self._current_train_time + datetime.timedelta(days=delta_day)).strftime("%Y%m%d")
+        return (self._current_train_time + datetime.timedelta(days=delta_day)
+                ).strftime("%Y%m%d")
 
     def timestamp(self, delta_day=0):
         """R
         """
-        return (self._current_train_time + datetime.timedelta(days=delta_day)).timestamp()
+        return (self._current_train_time + datetime.timedelta(days=delta_day)
+                ).timestamp()

doc/__init__.py

+# 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.

doc/contribute.md

 # PaddleRec 贡献代码
-> 占位
+> 占位
\ No newline at end of file

doc/custom_dataset_reader.md

+# PaddleRec 推荐数据集格式
+
+当你的数据集格式为[slot:feasign]*这种模式，或者可以预处理为这种格式时，可以直接使用PaddleRec内置的Reader。
+好处是不用自己写Reader了，各个model之间的数据格式也都可以统一成一样的格式。
+
+## 数据格式说明
+
+假如你的原始数据格式为
+
+```bash
+<label> <integer feature 1> ... <integer feature 13> <categorical feature 1> ... <categorical feature 26>
+```
+
+其中```<label>```表示广告是否被点击，点击用1表示，未点击用0表示。```<integer feature>```代表数值特征（连续特征），共有13个连续特征。
+并且每个特征有一个特征值。
+```<categorical feature>```代表分类特征（离散特征），共有26个离散特征。相邻两个特征用```\t```分隔。
+
+假设这13个连续特征（dense slot）的name如下：
+
+```
+D1 D2 D3 D4 D4 D6 D7 D8 D9 D10 D11 D12 D13
+```
+
+这26个离散特征（sparse slot）的name如下：
+```
+S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S20 S21 S22 S23 S24 S25 S26
+```
+
+那么下面这条样本（1个label + 13个dense值 + 26个feasign）
+```
+1 0.1 0.4 0.2 0.3 0.5 0.8 0.3 0.2 0.1 0.5 0.6 0.3 0.9 60 16 91 50 52 52 28 69 63 33 87 69 48 59 27 12 95 36 37 41 17 3 86 19 88 60
+```
+
+可以转换成：
+```
+label:1 D1:0.1 D2:0.4 D3:0.2 D4:0.3 D5:0.5 D6:0.8 D7:0.3 D8:0.2 D9:0.1 D10:0.5 D11:0.6 D12:0.3 D13:0.9 S14:60 S15:16 S16:91 S17:50 S18:52 S19:52 S20:28 S21:69 S22:63 S23:33 S24:87 S25:69 S26:48 S27:59 S28:27 S29:12 S30:95 S31:36 S32:37 S33:41 S34:17 S35:3 S36:86 S37:19 S38:88 S39:60
+```
+
+注意：上面各个slot:feasign字段之间的顺序没有要求，比如```D1:0.1 D2:0.4```改成```D2:0.4 D1:0.1```也可以。
+
+
+## 配置
+
+reader中需要配置```sparse_slots```与```dense_slots```，例如
+
+```
+  workspace: xxxx
+
+  reader:
+    batch_size: 2
+    train_data_path: "{workspace}/data/train_data"
+    sparse_slots: "label S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S20 S21 S22 S23 S24 S25 S26"
+    dense_slots: "D1:1 D2:1 D3:1 D4:1 D4:1 D6:1 D7:1 D8:1 D9:1 D10:1 D11:1 D12:1 D13:1"
+
+  model:
+    xxxxx
+```
+
+sparse_slots表示稀疏特征的列表，以空格分开。
+
+dense_slots表示稠密特征的列表，以空格分开。每个字段的格式是```[dense_slot_name]:[dim1,dim2,dim3...]```，其中```dim1,dim2,dim3...```表示shape
+
+
+配置好了之后，这些slot对应的variable就可以在model中的如下变量啦：
+```
+self._sparse_data_var
+
+self._dense_data_var
+```
+
 # PaddleRec 自定义数据集及Reader
 
 用户自定义数据集及配置异步Reader，需要关注以下几个步骤：

doc/design.md

@@ -279,4 +279,4 @@ class Metric(object):
         pass
 ```
 
-全局指标的计算及输出，需要分别继承并实现以上四个成员函数。具体实现的例子，可以参考[auc_metric.py](../core/metrics/auc_metrics.py)
+全局指标的计算及输出，需要分别继承并实现以上四个成员函数。具体实现的例子，可以参考[auc_metric.py](../core/metrics/auc_metrics.py)
\ No newline at end of file

doc/distributed_train.md

@@ -7,5 +7,3 @@
 
 ### K8S集群运行分布式
 > 占位
-
-

doc/faq.md

 # 常见问题FAQ
-> 占位
+> 占位
\ No newline at end of file

doc/local_train.md

 # PaddleRec 单机训练
-> 占位
+> 占位
\ No newline at end of file

doc/model_list.md

@@ -12,4 +12,3 @@
 |  多任务  |        [ESMM]()        |      ✓      |      x      |       ✓       |     x      |       ✓       |      ✓       |
 |   匹配   |        [DSSM]()        |      ✓      |      x      |       ✓       |     x      |       ✓       |      ✓       |
 |   匹配   |  [Multiview-Simnet]()  |      ✓      |      x      |       ✓       |     x      |       ✓       |      ✓       |
-

doc/optimization_model.md

 # PaddleRec 模型调参
-> 占位
+> 占位
\ No newline at end of file

doc/predict.md

-# PaddleRec 离线预测
+# PaddleRec 离线预测
\ No newline at end of file

doc/ps_background.md

@@ -5,4 +5,3 @@
 
 
 ## [参数服务器训练](https://www.paddlepaddle.org.cn/tutorials/projectdetail/464839)
-

models/contentunderstanding/__init__.py

+# 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/contentunderstanding/classification/config.yaml

@@ -37,4 +37,3 @@ train:
       dirname: "inference"
       epoch_interval: 100
       save_last: True
-

models/contentunderstanding/classification/model.py

@@ -31,7 +31,8 @@ class Model(ModelBase):
     def train_net(self):
         """ network definition """
 
-        data = fluid.data(name="input", shape=[None, self.max_len], dtype='int64')
+        data = fluid.data(
+            name="input", shape=[None, self.max_len], dtype='int64')
         label = fluid.data(name="label", shape=[None, 1], dtype='int64')
         seq_len = fluid.data(name="seq_len", shape=[None], dtype='int64')
 
@@ -51,7 +52,9 @@ class Model(ModelBase):
         # full connect layer
         fc_1 = fluid.layers.fc(input=[conv], size=self.hid_dim)
         # softmax layer
-        prediction = fluid.layers.fc(input=[fc_1], size=self.class_dim, act="softmax")
+        prediction = fluid.layers.fc(input=[fc_1],
+                                     size=self.class_dim,
+                                     act="softmax")
         cost = fluid.layers.cross_entropy(input=prediction, label=label)
         avg_cost = fluid.layers.mean(x=cost)
         acc = fluid.layers.accuracy(input=prediction, label=label)

models/contentunderstanding/classification/reader.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-
 import sys
 
 from paddlerec.core.reader import Reader
@@ -38,7 +37,8 @@ class TrainReader(Reader):
             data = [int(i) for i in data]
             label = [int(i) for i in label]
             seq_len = [int(i) for i in seq_len]
-            print >> sys.stderr, str([('data', data), ('label', label), ('seq_len', seq_len)])
+            print >> sys.stderr, str(
+                [('data', data), ('label', label), ('seq_len', seq_len)])
             yield [('data', data), ('label', label), ('seq_len', seq_len)]
 
         return data_iter

models/contentunderstanding/readme.md

@@ -71,13 +71,13 @@ python text2paddle.py raw_big_train_data/ raw_big_test_data/ train_big_data test
 ### 训练
 
 ```
-python -m paddlerec.run -m paddlerec.models.contentunderstanding.classification -d cpu -e single
+python -m paddlerec.run -m paddlerec.models.contentunderstanding.classification
 ```
 
 ### 预测
 
 ```
-python -m paddlerec.run -m paddlerec.models.contentunderstanding.classification -d cpu -e single
+python -m paddlerec.run -m paddlerec.models.contentunderstanding.classification
 ```
 
 ## 效果对比
@@ -87,19 +87,3 @@ python -m paddlerec.run -m paddlerec.models.contentunderstanding.classification
 | :------------------: | :--------------------: | :---------: |:---------: | :---------: |:---------: |
 |       ag news dataset        |       TagSpace       |       --        |       --          |       --          |       --          |
 |       --        |       Classification       |       --        |       --          |       --          |       --          |
-
-
-## 分布式
-### 模型训练性能 (样本/s)
-|       数据集        |       模型       |       单机        |       同步 (4节点)          |       同步 (8节点)          |  同步 (16节点)          |  同步 (32节点)          |
-| :------------------: | :--------------------: | :---------: |:---------: |:---------: |:---------: |:---------: |
-|       --        |       TagSpace       |       --        |       --          |       --          |  --          |  --          |
-|       --        |       Classification       |       --        |       --          |       --          |   --          |   --          |
-
-
-----
-
-|       数据集        |       模型       |       单机        |       异步 (4节点)          |       异步 (8节点)          |  异步 (16节点)          |  异步 (32节点)          |
-| :------------------: | :--------------------: | :---------: |:---------: |:---------: |:---------: |:---------: |
-|       --        |       TagSpace       |       --        |       --          |       --          |  --          |  --          |
-|       --        |       Classification       |       --        |       --          |       --          |   --          |   --          |

models/contentunderstanding/tagspace/config.yaml

@@ -47,4 +47,3 @@ train:
       dirname: "inference"
       epoch_interval: 100
       save_last: True
-

models/contentunderstanding/tagspace/model.py

@@ -26,8 +26,10 @@ class Model(ModelBase):
         ModelBase.__init__(self, config)
         self.cost = None
         self.metrics = {}
-        self.vocab_text_size = envs.get_global_env("vocab_text_size", None, self._namespace)
-        self.vocab_tag_size = envs.get_global_env("vocab_tag_size", None, self._namespace)
+        self.vocab_text_size = envs.get_global_env("vocab_text_size", None,
+                                                   self._namespace)
+        self.vocab_tag_size = envs.get_global_env("vocab_tag_size", None,
+                                                  self._namespace)
         self.emb_dim = envs.get_global_env("emb_dim", None, self._namespace)
         self.hid_dim = envs.get_global_env("hid_dim", None, self._namespace)
         self.win_size = envs.get_global_env("win_size", None, self._namespace)
@@ -35,8 +37,9 @@ class Model(ModelBase):
         self.neg_size = envs.get_global_env("neg_size", None, self._namespace)
 
     def train_net(self):
-        """ network definition """
-        text = fluid.data(name="text", shape=[None, 1], lod_level=1, dtype='int64')
+        """ network"""
+        text = fluid.data(
+            name="text", shape=[None, 1], lod_level=1, dtype='int64')
         pos_tag = fluid.data(
             name="pos_tag", shape=[None, 1], lod_level=1, dtype='int64')
         neg_tag = fluid.data(
@@ -45,13 +48,19 @@ class Model(ModelBase):
         self._data_var = [text, pos_tag, neg_tag]
 
         text_emb = fluid.embedding(
-            input=text, size=[self.vocab_text_size, self.emb_dim], param_attr="text_emb")
+            input=text,
+            size=[self.vocab_text_size, self.emb_dim],
+            param_attr="text_emb")
         text_emb = fluid.layers.squeeze(input=text_emb, axes=[1])
         pos_tag_emb = fluid.embedding(
-            input=pos_tag, size=[self.vocab_tag_size, self.emb_dim], param_attr="tag_emb")
+            input=pos_tag,
+            size=[self.vocab_tag_size, self.emb_dim],
+            param_attr="tag_emb")
         pos_tag_emb = fluid.layers.squeeze(input=pos_tag_emb, axes=[1])
         neg_tag_emb = fluid.embedding(
-            input=neg_tag, size=[self.vocab_tag_size, self.emb_dim], param_attr="tag_emb")
+            input=neg_tag,
+            size=[self.vocab_tag_size, self.emb_dim],
+            param_attr="tag_emb")
         neg_tag_emb = fluid.layers.squeeze(input=neg_tag_emb, axes=[1])
 
         conv_1d = fluid.nets.sequence_conv_pool(
@@ -65,7 +74,8 @@ class Model(ModelBase):
                                    size=self.emb_dim,
                                    param_attr="text_hid")
         cos_pos = nn.cos_sim(pos_tag_emb, text_hid)
-        mul_text_hid = fluid.layers.sequence_expand_as(x=text_hid, y=neg_tag_emb)
+        mul_text_hid = fluid.layers.sequence_expand_as(
+            x=text_hid, y=neg_tag_emb)
         mul_cos_neg = nn.cos_sim(neg_tag_emb, mul_text_hid)
         cos_neg_all = fluid.layers.sequence_reshape(
             input=mul_cos_neg, new_dim=self.neg_size)
@@ -74,7 +84,10 @@ class Model(ModelBase):
         #calculate hinge loss
         loss_part1 = nn.elementwise_sub(
             tensor.fill_constant_batch_size_like(
-                input=cos_pos, shape=[-1, 1], value=self.margin, dtype='float32'),
+                input=cos_pos,
+                shape=[-1, 1],
+                value=self.margin,
+                dtype='float32'),
             cos_pos)
         loss_part2 = nn.elementwise_add(loss_part1, cos_neg)
         loss_part3 = nn.elementwise_max(
@@ -85,7 +98,7 @@ class Model(ModelBase):
         less = tensor.cast(cf.less_than(cos_neg, cos_pos), dtype='float32')
         correct = nn.reduce_sum(less)
         self.cost = avg_cost
-        
+
         self.metrics["correct"] = correct
         self.metrics["cos_pos"] = cos_pos
 
@@ -96,7 +109,8 @@ class Model(ModelBase):
         return self.metrics
 
     def optimizer(self):
-        learning_rate = envs.get_global_env("hyper_parameters.base_lr", None, self._namespace)
+        learning_rate = envs.get_global_env("hyper_parameters.base_lr", None,
+                                            self._namespace)
         sgd_optimizer = fluid.optimizer.Adagrad(learning_rate=learning_rate)
         return sgd_optimizer
 

models/contentunderstanding/tagspace/reader.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-
 import sys
 
 import numpy as np

models/match/__init__.py

+# 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/match/dssm/model.py

@@ -23,13 +23,26 @@ class Model(ModelBase):
         ModelBase.__init__(self, config)
 
     def input(self):
-        TRIGRAM_D = envs.get_global_env("hyper_parameters.TRIGRAM_D", None, self._namespace)
-        Neg = envs.get_global_env("hyper_parameters.NEG", None, self._namespace)
-
-        self.query = fluid.data(name="query", shape=[-1, TRIGRAM_D], dtype='float32', lod_level=0)
-        self.doc_pos = fluid.data(name="doc_pos", shape=[-1, TRIGRAM_D], dtype='float32', lod_level=0)
-        self.doc_negs = [fluid.data(name="doc_neg_" + str(i), shape=[-1, TRIGRAM_D], dtype="float32", lod_level=0) for i
-                         in range(Neg)]
+        TRIGRAM_D = envs.get_global_env("hyper_parameters.TRIGRAM_D", None,
+                                        self._namespace)
+
+        Neg = envs.get_global_env("hyper_parameters.NEG", None,
+                                  self._namespace)
+
+        self.query = fluid.data(
+            name="query", shape=[-1, TRIGRAM_D], dtype='float32', lod_level=0)
+        self.doc_pos = fluid.data(
+            name="doc_pos",
+            shape=[-1, TRIGRAM_D],
+            dtype='float32',
+            lod_level=0)
+        self.doc_negs = [
+            fluid.data(
+                name="doc_neg_" + str(i),
+                shape=[-1, TRIGRAM_D],
+                dtype="float32",
+                lod_level=0) for i in range(Neg)
+        ]
         self._data_var.append(self.query)
         self._data_var.append(self.doc_pos)
         for input in self.doc_negs:
@@ -37,16 +50,24 @@ class Model(ModelBase):
 
         if self._platform != "LINUX":
             self._data_loader = fluid.io.DataLoader.from_generator(
-                feed_list=self._data_var, capacity=64, use_double_buffer=False, iterable=False)
+                feed_list=self._data_var,
+                capacity=64,
+                use_double_buffer=False,
+                iterable=False)
 
     def net(self, is_infer=False):
-        hidden_layers = envs.get_global_env("hyper_parameters.fc_sizes", None, self._namespace)
-        hidden_acts = envs.get_global_env("hyper_parameters.fc_acts", None, self._namespace)
+        hidden_layers = envs.get_global_env("hyper_parameters.fc_sizes", None,
+                                            self._namespace)
+        hidden_acts = envs.get_global_env("hyper_parameters.fc_acts", None,
+                                          self._namespace)
 
         def fc(data, hidden_layers, hidden_acts, names):
             fc_inputs = [data]
             for i in range(len(hidden_layers)):
-                xavier = fluid.initializer.Xavier(uniform=True, fan_in=fc_inputs[-1].shape[1], fan_out=hidden_layers[i])
+                xavier = fluid.initializer.Xavier(
+                    uniform=True,
+                    fan_in=fc_inputs[-1].shape[1],
+                    fan_out=hidden_layers[i])
                 out = fluid.layers.fc(input=fc_inputs[-1],
                                       size=hidden_layers[i],
                                       act=hidden_acts[i],
@@ -56,8 +77,10 @@ class Model(ModelBase):
                 fc_inputs.append(out)
             return fc_inputs[-1]
 
-        query_fc = fc(self.query, hidden_layers, hidden_acts, ['query_l1', 'query_l2', 'query_l3'])
-        doc_pos_fc = fc(self.doc_pos, hidden_layers, hidden_acts, ['doc_pos_l1', 'doc_pos_l2', 'doc_pos_l3'])
+        query_fc = fc(self.query, hidden_layers, hidden_acts,
+                      ['query_l1', 'query_l2', 'query_l3'])
+        doc_pos_fc = fc(self.doc_pos, hidden_layers, hidden_acts,
+                        ['doc_pos_l1', 'doc_pos_l2', 'doc_pos_l3'])
         self.R_Q_D_p = fluid.layers.cos_sim(query_fc, doc_pos_fc)
 
         if is_infer:
@@ -65,13 +88,17 @@ class Model(ModelBase):
 
         R_Q_D_ns = []
         for i, doc_neg in enumerate(self.doc_negs):
-            doc_neg_fc_i = fc(doc_neg, hidden_layers, hidden_acts,
-                              ['doc_neg_l1_' + str(i), 'doc_neg_l2_' + str(i), 'doc_neg_l3_' + str(i)])
+            doc_neg_fc_i = fc(doc_neg, hidden_layers, hidden_acts, [
+                'doc_neg_l1_' + str(i), 'doc_neg_l2_' + str(i),
+                'doc_neg_l3_' + str(i)
+            ])
             R_Q_D_ns.append(fluid.layers.cos_sim(query_fc, doc_neg_fc_i))
-        concat_Rs = fluid.layers.concat(input=[self.R_Q_D_p] + R_Q_D_ns, axis=-1)
+        concat_Rs = fluid.layers.concat(
+            input=[self.R_Q_D_p] + R_Q_D_ns, axis=-1)
         prob = fluid.layers.softmax(concat_Rs, axis=1)
 
-        hit_prob = fluid.layers.slice(prob, axes=[0, 1], starts=[0, 0], ends=[4, 1])
+        hit_prob = fluid.layers.slice(
+            prob, axes=[0, 1], starts=[0, 0], ends=[4, 1])
         loss = -fluid.layers.reduce_sum(fluid.layers.log(hit_prob))
         self.avg_cost = fluid.layers.mean(x=loss)
 
@@ -91,18 +118,28 @@ class Model(ModelBase):
         self.metrics()
 
     def optimizer(self):
-        learning_rate = envs.get_global_env("hyper_parameters.learning_rate", None, self._namespace)
+        learning_rate = envs.get_global_env("hyper_parameters.learning_rate",
+                                            None, self._namespace)
         optimizer = fluid.optimizer.SGD(learning_rate)
         return optimizer
 
     def infer_input(self):
-        TRIGRAM_D = envs.get_global_env("hyper_parameters.TRIGRAM_D", None, self._namespace)
-        self.query = fluid.data(name="query", shape=[-1, TRIGRAM_D], dtype='float32', lod_level=0)
-        self.doc_pos = fluid.data(name="doc_pos", shape=[-1, TRIGRAM_D], dtype='float32', lod_level=0)
+        TRIGRAM_D = envs.get_global_env("hyper_parameters.TRIGRAM_D", None,
+                                        self._namespace)
+        self.query = fluid.data(
+            name="query", shape=[-1, TRIGRAM_D], dtype='float32', lod_level=0)
+        self.doc_pos = fluid.data(
+            name="doc_pos",
+            shape=[-1, TRIGRAM_D],
+            dtype='float32',
+            lod_level=0)
         self._infer_data_var = [self.query, self.doc_pos]
 
         self._infer_data_loader = fluid.io.DataLoader.from_generator(
-            feed_list=self._infer_data_var, capacity=64, use_double_buffer=False, iterable=False)
+            feed_list=self._infer_data_var,
+            capacity=64,
+            use_double_buffer=False,
+            iterable=False)
 
     def infer_net(self):
         self.infer_input()

models/match/multiview-simnet/data_process.sh

@@ -22,4 +22,3 @@ mkdir -p data/train
 mkdir -p data/test
 
 python generate_synthetic_data.py 
-

models/match/multiview-simnet/evaluate_reader.py

@@ -18,8 +18,10 @@ from paddlerec.core.utils import envs
 
 class EvaluateReader(Reader):
     def init(self):
-        self.query_slots = envs.get_global_env("hyper_parameters.query_slots", None, "train.model")
-        self.title_slots = envs.get_global_env("hyper_parameters.title_slots", None, "train.model")
+        self.query_slots = envs.get_global_env("hyper_parameters.query_slots",
+                                               None, "train.model")
+        self.title_slots = envs.get_global_env("hyper_parameters.title_slots",
+                                               None, "train.model")
 
         self.all_slots = []
         for i in range(self.query_slots):

models/match/multiview-simnet/generate_synthetic_data.py

@@ -21,7 +21,11 @@ class Dataset:
 
 
 class SyntheticDataset(Dataset):
-    def __init__(self, sparse_feature_dim, query_slot_num, title_slot_num, dataset_size=10000):
+    def __init__(self,
+                 sparse_feature_dim,
+                 query_slot_num,
+                 title_slot_num,
+                 dataset_size=10000):
         # ids are randomly generated
         self.ids_per_slot = 10
         self.sparse_feature_dim = sparse_feature_dim
@@ -46,14 +50,20 @@ class SyntheticDataset(Dataset):
                 for i in range(self.title_slot_num):
                     pt_slot = generate_ids(self.ids_per_slot,
                                            self.sparse_feature_dim)
-                    pt_slot = [str(fea) + ':' + str(i + self.query_slot_num) for fea in pt_slot]
+                    pt_slot = [
+                        str(fea) + ':' + str(i + self.query_slot_num)
+                        for fea in pt_slot
+                    ]
                     pos_title_slots += pt_slot
                 if is_train:
                     for i in range(self.title_slot_num):
                         nt_slot = generate_ids(self.ids_per_slot,
                                                self.sparse_feature_dim)
-                        nt_slot = [str(fea) + ':' + str(i + self.query_slot_num + self.title_slot_num) for fea in
-                                   nt_slot]
+                        nt_slot = [
+                            str(fea) + ':' +
+                            str(i + self.query_slot_num + self.title_slot_num)
+                            for fea in nt_slot
+                        ]
                         neg_title_slots += nt_slot
                     yield query_slots + pos_title_slots + neg_title_slots
                 else:
@@ -76,7 +86,8 @@ if __name__ == '__main__':
     query_slots = 1
     title_slots = 1
     dataset_size = 10
-    dataset = SyntheticDataset(sparse_feature_dim, query_slots, title_slots, dataset_size)
+    dataset = SyntheticDataset(sparse_feature_dim, query_slots, title_slots,
+                               dataset_size)
     train_reader = dataset.train()
     test_reader = dataset.test()
 

models/match/multiview-simnet/model.py

@@ -103,12 +103,18 @@ class Model(ModelBase):
 
     def init_config(self):
         self._fetch_interval = 1
-        query_encoder = envs.get_global_env("hyper_parameters.query_encoder", None, self._namespace)
-        title_encoder = envs.get_global_env("hyper_parameters.title_encoder", None, self._namespace)
-        query_encode_dim = envs.get_global_env("hyper_parameters.query_encode_dim", None, self._namespace)
-        title_encode_dim = envs.get_global_env("hyper_parameters.title_encode_dim", None, self._namespace)
-        query_slots = envs.get_global_env("hyper_parameters.query_slots", None, self._namespace)
-        title_slots = envs.get_global_env("hyper_parameters.title_slots", None, self._namespace)
+        query_encoder = envs.get_global_env("hyper_parameters.query_encoder",
+                                            None, self._namespace)
+        title_encoder = envs.get_global_env("hyper_parameters.title_encoder",
+                                            None, self._namespace)
+        query_encode_dim = envs.get_global_env(
+            "hyper_parameters.query_encode_dim", None, self._namespace)
+        title_encode_dim = envs.get_global_env(
+            "hyper_parameters.title_encode_dim", None, self._namespace)
+        query_slots = envs.get_global_env("hyper_parameters.query_slots", None,
+                                          self._namespace)
+        title_slots = envs.get_global_env("hyper_parameters.title_slots", None,
+                                          self._namespace)
         factory = SimpleEncoderFactory()
         self.query_encoders = [
             factory.create(query_encoder, query_encode_dim)
@@ -119,10 +125,13 @@ class Model(ModelBase):
             for i in range(title_slots)
         ]
 
-        self.emb_size = envs.get_global_env("hyper_parameters.sparse_feature_dim", None, self._namespace)
-        self.emb_dim = envs.get_global_env("hyper_parameters.embedding_dim", None, self._namespace)
+        self.emb_size = envs.get_global_env(
+            "hyper_parameters.sparse_feature_dim", None, self._namespace)
+        self.emb_dim = envs.get_global_env("hyper_parameters.embedding_dim",
+                                           None, self._namespace)
         self.emb_shape = [self.emb_size, self.emb_dim]
-        self.hidden_size = envs.get_global_env("hyper_parameters.hidden_size", None, self._namespace)
+        self.hidden_size = envs.get_global_env("hyper_parameters.hidden_size",
+                                               None, self._namespace)
         self.margin = 0.1
 
     def input(self, is_train=True):
@@ -133,8 +142,10 @@ class Model(ModelBase):
         ]
         self.pt_slots = [
             fluid.data(
-                name="%d" % (i + len(self.query_encoders)), shape=[None, 1], lod_level=1, dtype='int64')
-            for i in range(len(self.title_encoders))
+                name="%d" % (i + len(self.query_encoders)),
+                shape=[None, 1],
+                lod_level=1,
+                dtype='int64') for i in range(len(self.title_encoders))
         ]
 
         if is_train == False:
@@ -142,9 +153,11 @@ class Model(ModelBase):
 
         self.nt_slots = [
             fluid.data(
-                name="%d" % (i + len(self.query_encoders) + len(self.title_encoders)), shape=[None, 1], lod_level=1,
-                dtype='int64')
-            for i in range(len(self.title_encoders))
+                name="%d" %
+                (i + len(self.query_encoders) + len(self.title_encoders)),
+                shape=[None, 1],
+                lod_level=1,
+                dtype='int64') for i in range(len(self.title_encoders))
         ]
 
         return self.q_slots + self.pt_slots + self.nt_slots
@@ -153,11 +166,15 @@ class Model(ModelBase):
         res = self.input()
         self._data_var = res
 
-        use_dataloader = envs.get_global_env("hyper_parameters.use_DataLoader", False, self._namespace)
+        use_dataloader = envs.get_global_env("hyper_parameters.use_DataLoader",
+                                             False, self._namespace)
 
         if self._platform != "LINUX" or use_dataloader:
             self._data_loader = fluid.io.DataLoader.from_generator(
-                feed_list=self._data_var, capacity=256, use_double_buffer=False, iterable=False)
+                feed_list=self._data_var,
+                capacity=256,
+                use_double_buffer=False,
+                iterable=False)
 
     def get_acc(self, x, y):
         less = tensor.cast(cf.less_than(x, y), dtype='float32')
@@ -190,10 +207,12 @@ class Model(ModelBase):
             self.query_encoders[i].forward(emb) for i, emb in enumerate(q_embs)
         ]
         pt_encodes = [
-            self.title_encoders[i].forward(emb) for i, emb in enumerate(pt_embs)
+            self.title_encoders[i].forward(emb)
+            for i, emb in enumerate(pt_embs)
         ]
         nt_encodes = [
-            self.title_encoders[i].forward(emb) for i, emb in enumerate(nt_embs)
+            self.title_encoders[i].forward(emb)
+            for i, emb in enumerate(nt_embs)
         ]
 
         # concat multi view for query, pos_title, neg_title
@@ -252,7 +271,8 @@ class Model(ModelBase):
         self.metrics()
 
     def optimizer(self):
-        learning_rate = envs.get_global_env("hyper_parameters.learning_rate", None, self._namespace)
+        learning_rate = envs.get_global_env("hyper_parameters.learning_rate",
+                                            None, self._namespace)
         optimizer = fluid.optimizer.Adam(learning_rate=learning_rate)
         return optimizer
 
@@ -261,7 +281,10 @@ class Model(ModelBase):
         self._infer_data_var = res
 
         self._infer_data_loader = fluid.io.DataLoader.from_generator(
-            feed_list=self._infer_data_var, capacity=64, use_double_buffer=False, iterable=False)
+            feed_list=self._infer_data_var,
+            capacity=64,
+            use_double_buffer=False,
+            iterable=False)
 
     def infer_net(self):
         self.infer_input()
@@ -281,7 +304,8 @@ class Model(ModelBase):
             self.query_encoders[i].forward(emb) for i, emb in enumerate(q_embs)
         ]
         pt_encodes = [
-            self.title_encoders[i].forward(emb) for i, emb in enumerate(pt_embs)
+            self.title_encoders[i].forward(emb)
+            for i, emb in enumerate(pt_embs)
         ]
         # concat multi view for query, pos_title, neg_title
         q_concat = fluid.layers.concat(q_encodes)

models/match/multiview-simnet/reader.py

@@ -18,8 +18,10 @@ from paddlerec.core.utils import envs
 
 class TrainReader(Reader):
     def init(self):
-        self.query_slots = envs.get_global_env("hyper_parameters.query_slots", None, "train.model")
-        self.title_slots = envs.get_global_env("hyper_parameters.title_slots", None, "train.model")
+        self.query_slots = envs.get_global_env("hyper_parameters.query_slots",
+                                               None, "train.model")
+        self.title_slots = envs.get_global_env("hyper_parameters.title_slots",
+                                               None, "train.model")
 
         self.all_slots = []
         for i in range(self.query_slots):

models/match/readme.md

@@ -37,4 +37,3 @@
 python -m paddlerec.run -m paddlerec.models.match.dssm # dssm
 python -m paddlerec.run -m paddlerec.models.match.multiview-simnet # multiview-simnet
 ```
-

models/multitask/__init__.py

+# 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/multitask/esmm/esmm_infer_reader.py

@@ -20,9 +20,11 @@ from paddlerec.core.reader import Reader
 
 class EvaluateReader(Reader):
     def init(self):
-        all_field_id = ['101', '109_14', '110_14', '127_14', '150_14', '121', '122', '124', '125', '126', '127', '128',
-                        '129',
-                        '205', '206', '207', '210', '216', '508', '509', '702', '853', '301']
+        all_field_id = [
+            '101', '109_14', '110_14', '127_14', '150_14', '121', '122', '124',
+            '125', '126', '127', '128', '129', '205', '206', '207', '210',
+            '216', '508', '509', '702', '853', '301'
+        ]
         self.all_field_id_dict = defaultdict(int)
         for i, field_id in enumerate(all_field_id):
             self.all_field_id_dict[field_id] = [False, i]

models/multitask/esmm/esmm_reader.py

@@ -21,9 +21,11 @@ from paddlerec.core.reader import Reader
 
 class TrainReader(Reader):
     def init(self):
-        all_field_id = ['101', '109_14', '110_14', '127_14', '150_14', '121', '122', '124', '125', '126', '127', '128',
-                        '129',
-                        '205', '206', '207', '210', '216', '508', '509', '702', '853', '301']
+        all_field_id = [
+            '101', '109_14', '110_14', '127_14', '150_14', '121', '122', '124',
+            '125', '126', '127', '128', '129', '205', '206', '207', '210',
+            '216', '508', '509', '702', '853', '301'
+        ]
         self.all_field_id_dict = defaultdict(int)
         for i, field_id in enumerate(all_field_id):
             self.all_field_id_dict[field_id] = [False, i]

models/multitask/esmm/model.py

@@ -28,11 +28,13 @@ class Model(ModelBase):
         init_stddev = 1.0
         scales = 1.0 / np.sqrt(data.shape[1])
 
-        p_attr = fluid.param_attr.ParamAttr(name='%s_weight' % tag,
-                                            initializer=fluid.initializer.NormalInitializer(loc=0.0,
-                                                                                            scale=init_stddev * scales))
+        p_attr = fluid.param_attr.ParamAttr(
+            name='%s_weight' % tag,
+            initializer=fluid.initializer.NormalInitializer(
+                loc=0.0, scale=init_stddev * scales))
 
-        b_attr = fluid.ParamAttr(name='%s_bias' % tag, initializer=fluid.initializer.Constant(0.1))
+        b_attr = fluid.ParamAttr(
+            name='%s_bias' % tag, initializer=fluid.initializer.Constant(0.1))
 
         out = fluid.layers.fc(input=data,
                               size=out_dim,
@@ -44,7 +46,11 @@ class Model(ModelBase):
 
     def input_data(self):
         sparse_input_ids = [
-            fluid.data(name="field_" + str(i), shape=[-1, 1], dtype="int64", lod_level=1) for i in range(0, 23)
+            fluid.data(
+                name="field_" + str(i),
+                shape=[-1, 1],
+                dtype="int64",
+                lod_level=1) for i in range(0, 23)
         ]
         label_ctr = fluid.data(name="ctr", shape=[-1, 1], dtype="int64")
         label_cvr = fluid.data(name="cvr", shape=[-1, 1], dtype="int64")
@@ -55,19 +61,23 @@ class Model(ModelBase):
 
     def net(self, inputs, is_infer=False):
 
-        vocab_size = envs.get_global_env("hyper_parameters.vocab_size", None, self._namespace)
-        embed_size = envs.get_global_env("hyper_parameters.embed_size", None, self._namespace)
+        vocab_size = envs.get_global_env("hyper_parameters.vocab_size", None,
+                                         self._namespace)
+        embed_size = envs.get_global_env("hyper_parameters.embed_size", None,
+                                         self._namespace)
         emb = []
         for data in inputs[0:-2]:
-            feat_emb = fluid.embedding(input=data,
-                                       size=[vocab_size, embed_size],
-                                       param_attr=fluid.ParamAttr(name='dis_emb',
-                                                                  learning_rate=5,
-                                                                  initializer=fluid.initializer.Xavier(
-                                                                      fan_in=embed_size, fan_out=embed_size)
-                                                                  ),
-                                       is_sparse=True)
-            field_emb = fluid.layers.sequence_pool(input=feat_emb, pool_type='sum')
+            feat_emb = fluid.embedding(
+                input=data,
+                size=[vocab_size, embed_size],
+                param_attr=fluid.ParamAttr(
+                    name='dis_emb',
+                    learning_rate=5,
+                    initializer=fluid.initializer.Xavier(
+                        fan_in=embed_size, fan_out=embed_size)),
+                is_sparse=True)
+            field_emb = fluid.layers.sequence_pool(
+                input=feat_emb, pool_type='sum')
             emb.append(field_emb)
         concat_emb = fluid.layers.concat(emb, axis=1)
 
@@ -85,14 +95,20 @@ class Model(ModelBase):
         ctr_clk = inputs[-2]
         ctcvr_buy = inputs[-1]
 
-        ctr_prop_one = fluid.layers.slice(ctr_out, axes=[1], starts=[1], ends=[2])
-        cvr_prop_one = fluid.layers.slice(cvr_out, axes=[1], starts=[1], ends=[2])
+        ctr_prop_one = fluid.layers.slice(
+            ctr_out, axes=[1], starts=[1], ends=[2])
+        cvr_prop_one = fluid.layers.slice(
+            cvr_out, axes=[1], starts=[1], ends=[2])
 
-        ctcvr_prop_one = fluid.layers.elementwise_mul(ctr_prop_one, cvr_prop_one)
-        ctcvr_prop = fluid.layers.concat(input=[1 - ctcvr_prop_one, ctcvr_prop_one], axis=1)
+        ctcvr_prop_one = fluid.layers.elementwise_mul(ctr_prop_one,
+                                                      cvr_prop_one)
+        ctcvr_prop = fluid.layers.concat(
+            input=[1 - ctcvr_prop_one, ctcvr_prop_one], axis=1)
 
-        auc_ctr, batch_auc_ctr, auc_states_ctr = fluid.layers.auc(input=ctr_out, label=ctr_clk)
-        auc_ctcvr, batch_auc_ctcvr, auc_states_ctcvr = fluid.layers.auc(input=ctcvr_prop, label=ctcvr_buy)
+        auc_ctr, batch_auc_ctr, auc_states_ctr = fluid.layers.auc(
+            input=ctr_out, label=ctr_clk)
+        auc_ctcvr, batch_auc_ctcvr, auc_states_ctcvr = fluid.layers.auc(
+            input=ctcvr_prop, label=ctcvr_buy)
 
         if is_infer:
             self._infer_results["AUC_ctr"] = auc_ctr
@@ -100,7 +116,8 @@ class Model(ModelBase):
             return
 
         loss_ctr = fluid.layers.cross_entropy(input=ctr_out, label=ctr_clk)
-        loss_ctcvr = fluid.layers.cross_entropy(input=ctcvr_prop, label=ctcvr_buy)
+        loss_ctcvr = fluid.layers.cross_entropy(
+            input=ctcvr_prop, label=ctcvr_buy)
         cost = loss_ctr + loss_ctcvr
         avg_cost = fluid.layers.mean(cost)
 
@@ -117,5 +134,8 @@ class Model(ModelBase):
     def infer_net(self):
         self._infer_data_var = self.input_data()
         self._infer_data_loader = fluid.io.DataLoader.from_generator(
-            feed_list=self._infer_data_var, capacity=64, use_double_buffer=False, iterable=False)
+            feed_list=self._infer_data_var,
+            capacity=64,
+            use_double_buffer=False,
+            iterable=False)
         self.net(self._infer_data_var, is_infer=True)

models/multitask/mmoe/census_infer_reader.py

@@ -19,6 +19,7 @@ from paddlerec.core.reader import Reader
 
 class EvaluateReader(Reader):
     def init(self):
+
         pass
 
     def generate_sample(self, line):

models/multitask/mmoe/census_reader.py

@@ -24,6 +24,7 @@ class TrainReader(Reader):
     def generate_sample(self, line):
         """
         Read the data line by line and process it as a dictionary
+
         """
 
         def reader():

models/multitask/mmoe/model.py

@@ -23,44 +23,58 @@ class Model(ModelBase):
         ModelBase.__init__(self, config)
 
     def MMOE(self, is_infer=False):
-
-        feature_size = envs.get_global_env("hyper_parameters.feature_size", None, self._namespace)
-        expert_num = envs.get_global_env("hyper_parameters.expert_num", None, self._namespace)
-        gate_num = envs.get_global_env("hyper_parameters.gate_num", None, self._namespace)
-        expert_size = envs.get_global_env("hyper_parameters.expert_size", None, self._namespace)
-        tower_size = envs.get_global_env("hyper_parameters.tower_size", None, self._namespace)
-
-        input_data = fluid.data(name="input", shape=[-1, feature_size], dtype="float32")
-        label_income = fluid.data(name="label_income", shape=[-1, 2], dtype="float32", lod_level=0)
-        label_marital = fluid.data(name="label_marital", shape=[-1, 2], dtype="float32", lod_level=0)
+        feature_size = envs.get_global_env("hyper_parameters.feature_size",
+                                           None, self._namespace)
+        expert_num = envs.get_global_env("hyper_parameters.expert_num", None,
+                                         self._namespace)
+        gate_num = envs.get_global_env("hyper_parameters.gate_num", None,
+                                       self._namespace)
+        expert_size = envs.get_global_env("hyper_parameters.expert_size", None,
+                                          self._namespace)
+        tower_size = envs.get_global_env("hyper_parameters.tower_size", None,
+                                         self._namespace)
+
+        input_data = fluid.data(
+            name="input", shape=[-1, feature_size], dtype="float32")
+        label_income = fluid.data(
+            name="label_income", shape=[-1, 2], dtype="float32", lod_level=0)
+        label_marital = fluid.data(
+            name="label_marital", shape=[-1, 2], dtype="float32", lod_level=0)
         if is_infer:
             self._infer_data_var = [input_data, label_income, label_marital]
             self._infer_data_loader = fluid.io.DataLoader.from_generator(
-                feed_list=self._infer_data_var, capacity=64, use_double_buffer=False, iterable=False)
+                feed_list=self._infer_data_var,
+                capacity=64,
+                use_double_buffer=False,
+                iterable=False)
 
         self._data_var.extend([input_data, label_income, label_marital])
         # f_{i}(x) = activation(W_{i} * x + b), where activation is ReLU according to the paper
         expert_outputs = []
         for i in range(0, expert_num):
-            expert_output = fluid.layers.fc(input=input_data,
-                                            size=expert_size,
-                                            act='relu',
-                                            bias_attr=fluid.ParamAttr(learning_rate=1.0),
-                                            name='expert_' + str(i))
+            expert_output = fluid.layers.fc(
+                input=input_data,
+                size=expert_size,
+                act='relu',
+                bias_attr=fluid.ParamAttr(learning_rate=1.0),
+                name='expert_' + str(i))
             expert_outputs.append(expert_output)
         expert_concat = fluid.layers.concat(expert_outputs, axis=1)
-        expert_concat = fluid.layers.reshape(expert_concat, [-1, expert_num, expert_size])
+        expert_concat = fluid.layers.reshape(expert_concat,
+                                             [-1, expert_num, expert_size])
 
         # g^{k}(x) = activation(W_{gk} * x + b), where activation is softmax according to the paper
         output_layers = []
         for i in range(0, gate_num):
-            cur_gate = fluid.layers.fc(input=input_data,
-                                       size=expert_num,
-                                       act='softmax',
-                                       bias_attr=fluid.ParamAttr(learning_rate=1.0),
-                                       name='gate_' + str(i))
+            cur_gate = fluid.layers.fc(
+                input=input_data,
+                size=expert_num,
+                act='softmax',
+                bias_attr=fluid.ParamAttr(learning_rate=1.0),
+                name='gate_' + str(i))
             # f^{k}(x) = sum_{i=1}^{n}(g^{k}(x)_{i} * f_{i}(x))
-            cur_gate_expert = fluid.layers.elementwise_mul(expert_concat, cur_gate, axis=0)
+            cur_gate_expert = fluid.layers.elementwise_mul(
+                expert_concat, cur_gate, axis=0)
             cur_gate_expert = fluid.layers.reduce_sum(cur_gate_expert, dim=1)
             # Build tower layer
             cur_tower = fluid.layers.fc(input=cur_gate_expert,
@@ -74,25 +88,33 @@ class Model(ModelBase):
 
             output_layers.append(out)
 
-        pred_income = fluid.layers.clip(output_layers[0], min=1e-15, max=1.0 - 1e-15)
-        pred_marital = fluid.layers.clip(output_layers[1], min=1e-15, max=1.0 - 1e-15)
-
-        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=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'))
+        pred_income = fluid.layers.clip(
+            output_layers[0], min=1e-15, max=1.0 - 1e-15)
+        pred_marital = fluid.layers.clip(
+            output_layers[1], min=1e-15, max=1.0 - 1e-15)
+
+        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=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'))
         if is_infer:
             self._infer_results["AUC_income"] = auc_income
             self._infer_results["AUC_marital"] = auc_marital
             return
 
-        cost_income = fluid.layers.cross_entropy(input=pred_income, label=label_income, soft_label=True)
-        cost_marital = fluid.layers.cross_entropy(input=pred_marital, label=label_marital, soft_label=True)
+        cost_income = fluid.layers.cross_entropy(
+            input=pred_income, label=label_income, soft_label=True)
+        cost_marital = fluid.layers.cross_entropy(
+            input=pred_marital, label=label_marital, soft_label=True)
 
         avg_cost_income = fluid.layers.mean(x=cost_income)
         avg_cost_marital = fluid.layers.mean(x=cost_marital)

models/multitask/readme.md

@@ -56,4 +56,3 @@ python -m paddlerec.run -m paddlerec.models.multitask.esmm # esmm
 |       Census-income Data     |       Share-Bottom       |       --        |     0.93120/0.99256         |
 |       Census-income Data        |       MMoE       |       --        |       0.94465/0.99324         |
 |          Ali-CCP     |    ESMM       |       --        |      0.97181/0.49967          |
-

models/multitask/share-bottom/model.py

@@ -24,27 +24,38 @@ class Model(ModelBase):
 
     def model(self, is_infer=False):
 
-        feature_size = envs.get_global_env("hyper_parameters.feature_size", None, self._namespace)
-        bottom_size = envs.get_global_env("hyper_parameters.bottom_size", None, self._namespace)
-        tower_size = envs.get_global_env("hyper_parameters.tower_size", None, self._namespace)
-        tower_nums = envs.get_global_env("hyper_parameters.tower_nums", None, self._namespace)
-
-        input_data = fluid.data(name="input", shape=[-1, feature_size], dtype="float32")
-        label_income = fluid.data(name="label_income", shape=[-1, 2], dtype="float32", lod_level=0)
-        label_marital = fluid.data(name="label_marital", shape=[-1, 2], dtype="float32", lod_level=0)
+        feature_size = envs.get_global_env("hyper_parameters.feature_size",
+                                           None, self._namespace)
+        bottom_size = envs.get_global_env("hyper_parameters.bottom_size", None,
+                                          self._namespace)
+        tower_size = envs.get_global_env("hyper_parameters.tower_size", None,
+                                         self._namespace)
+        tower_nums = envs.get_global_env("hyper_parameters.tower_nums", None,
+                                         self._namespace)
+
+        input_data = fluid.data(
+            name="input", shape=[-1, feature_size], dtype="float32")
+        label_income = fluid.data(
+            name="label_income", shape=[-1, 2], dtype="float32", lod_level=0)
+        label_marital = fluid.data(
+            name="label_marital", shape=[-1, 2], dtype="float32", lod_level=0)
 
         if is_infer:
             self._infer_data_var = [input_data, label_income, label_marital]
             self._infer_data_loader = fluid.io.DataLoader.from_generator(
-                feed_list=self._infer_data_var, capacity=64, use_double_buffer=False, iterable=False)
+                feed_list=self._infer_data_var,
+                capacity=64,
+                use_double_buffer=False,
+                iterable=False)
 
         self._data_var.extend([input_data, label_income, label_marital])
 
-        bottom_output = fluid.layers.fc(input=input_data,
-                                        size=bottom_size,
-                                        act='relu',
-                                        bias_attr=fluid.ParamAttr(learning_rate=1.0),
-                                        name='bottom_output')
+        bottom_output = fluid.layers.fc(
+            input=input_data,
+            size=bottom_size,
+            act='relu',
+            bias_attr=fluid.ParamAttr(learning_rate=1.0),
+            name='bottom_output')
 
         # Build tower layer from bottom layer
         output_layers = []
@@ -59,26 +70,34 @@ class Model(ModelBase):
                                            name='output_layer_' + str(index))
             output_layers.append(output_layer)
 
-        pred_income = fluid.layers.clip(output_layers[0], min=1e-15, max=1.0 - 1e-15)
-        pred_marital = fluid.layers.clip(output_layers[1], min=1e-15, max=1.0 - 1e-15)
-
-        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=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'))
+        pred_income = fluid.layers.clip(
+            output_layers[0], min=1e-15, max=1.0 - 1e-15)
+        pred_marital = fluid.layers.clip(
+            output_layers[1], min=1e-15, max=1.0 - 1e-15)
+
+        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=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'))
 
         if is_infer:
             self._infer_results["AUC_income"] = auc_income
             self._infer_results["AUC_marital"] = auc_marital
             return
 
-        cost_income = fluid.layers.cross_entropy(input=pred_income, label=label_income, soft_label=True)
-        cost_marital = fluid.layers.cross_entropy(input=pred_marital, label=label_marital, soft_label=True)
+        cost_income = fluid.layers.cross_entropy(
+            input=pred_income, label=label_income, soft_label=True)
+        cost_marital = fluid.layers.cross_entropy(
+            input=pred_marital, label=label_marital, soft_label=True)
         cost = fluid.layers.elementwise_add(cost_income, cost_marital, axis=1)
 
         avg_cost = fluid.layers.mean(x=cost)

models/rank/__init__.py

+# 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/dcn/config.yaml

@@ -22,9 +22,10 @@ train:
 
   reader:
     batch_size: 2
-    class: "{workspace}/criteo_reader.py"
-    train_data_path: "{workspace}/data/train"
+    train_data_path: "{workspace}/data/slot_train"
     feat_dict_name: "{workspace}/data/vocab"
+    sparse_slots: "label C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C14 C15 C16 C17 C18 C19 C20 C21 C22 C23 C24 C25 C26"
+    dense_slots: "I1:1 I2:1 I3:1 I4:1 I5:1 I6:1 I7:1 I8:1 I9:1 I10:1 I11:1 I12:1 I13:1"
 
   model:
     models: "{workspace}/model.py"

models/rank/dcn/data/download.py

+# 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 os
 import sys
 import io

models/rank/dcn/criteo_reader.py → models/rank/dcn/data/get_slot_data.py

@@ -11,21 +11,32 @@
 # 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.
-
-from __future__ import print_function
+import math
+import sys
+import yaml
+from paddlerec.core.reader import Reader
+from paddlerec.core.utils import envs
 import math
 import os
-
 try:
     import cPickle as pickle
 except ImportError:
     import pickle
+from collections import Counter
+import os
+import paddle.fluid.incubate.data_generator as dg
 
-from paddlerec.core.reader import Reader
-from paddlerec.core.utils import envs
 
+class TrainReader(dg.MultiSlotDataGenerator):
+    def __init__(self, config):
+        dg.MultiSlotDataGenerator.__init__(self)
+
+        if os.path.isfile(config):
+            with open(config, 'r') as rb:
+                _config = yaml.load(rb.read(), Loader=yaml.FullLoader)
+        else:
+            raise ValueError("reader config only support yaml")
 
-class TrainReader(Reader):
     def init(self):
         self.cont_min_ = [0, -3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
         self.cont_max_ = [
@@ -48,7 +59,7 @@ class TrainReader(Reader):
         self.cat_feat_idx_dict_list = [{} for _ in range(26)]
 
         # TODO: set vocabulary dictionary
-        vocab_dir = envs.get_global_env("feat_dict_name", None, "train.reader")
+        vocab_dir = "./vocab/"
         for i in range(26):
             lookup_idx = 1  # remain 0 for default value
             for line in open(
@@ -72,11 +83,11 @@ class TrainReader(Reader):
                     if idx == 2 else math.log(1 + float(features[idx])))
         for idx in self.cat_idx_:
             if features[idx] == '' or features[
-                idx] not in self.cat_feat_idx_dict_list[idx - 14]:
+                    idx] not in self.cat_feat_idx_dict_list[idx - 14]:
                 label_feat_list[idx].append(0)
             else:
                 label_feat_list[idx].append(self.cat_feat_idx_dict_list[
-                                                idx - 14][features[idx]])
+                    idx - 14][features[idx]])
         label_feat_list[0].append(int(features[0]))
         return label_feat_list
 
@@ -87,6 +98,18 @@ class TrainReader(Reader):
 
         def data_iter():
             label_feat_list = self._process_line(line)
-            yield list(zip(self.label_feat_names, label_feat_list))
+            s = ""
+            for i in list(zip(self.label_feat_names, label_feat_list)):
+                k = i[0]
+                v = i[1]
+                for j in v:
+                    s += " " + k + ":" + str(j)
+            print s.strip()
+            yield None
 
         return data_iter
+
+
+reader = TrainReader("../config.yaml")
+reader.init()
+reader.run_from_stdin()

models/rank/dcn/data/preprocess.py

+# 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.
+
 from __future__ import print_function, absolute_import, division
 
 import os

models/rank/dcn/data/run.sh

+python download.py
+python preprocess.py
+
+mkdir slot_train
+for i in `ls ./train`
+do
+    cat train/$i | python get_slot_data.py > slot_train/$i
+done
+
+mkdir slot_test_valid
+for i in `ls ./test_valid`
+do
+    cat test_valid/$i | python get_slot_data.py > slot_test_valid/$i
+done

models/rank/dcn/model.py

@@ -25,12 +25,23 @@ class Model(ModelBase):
         ModelBase.__init__(self, config)
 
     def init_network(self):
-        self.cross_num = envs.get_global_env("hyper_parameters.cross_num", None, self._namespace)
-        self.dnn_hidden_units = envs.get_global_env("hyper_parameters.dnn_hidden_units", None, self._namespace)
-        self.l2_reg_cross = envs.get_global_env("hyper_parameters.l2_reg_cross", None, self._namespace)
-        self.dnn_use_bn = envs.get_global_env("hyper_parameters.dnn_use_bn", None, self._namespace)
-        self.clip_by_norm = envs.get_global_env("hyper_parameters.clip_by_norm", None, self._namespace)
-        cat_feat_num = envs.get_global_env("hyper_parameters.cat_feat_num", None, self._namespace)
+        self.cross_num = envs.get_global_env("hyper_parameters.cross_num",
+                                             None, self._namespace)
+        self.dnn_hidden_units = envs.get_global_env(
+            "hyper_parameters.dnn_hidden_units", None, self._namespace)
+        self.l2_reg_cross = envs.get_global_env(
+            "hyper_parameters.l2_reg_cross", None, self._namespace)
+        self.dnn_use_bn = envs.get_global_env("hyper_parameters.dnn_use_bn",
+                                              None, self._namespace)
+        self.clip_by_norm = envs.get_global_env(
+            "hyper_parameters.clip_by_norm", None, self._namespace)
+        cat_feat_num = envs.get_global_env("hyper_parameters.cat_feat_num",
+                                           None, self._namespace)
+
+        self.sparse_inputs = self._sparse_data_var[1:]
+        self.dense_inputs = self._dense_data_var
+        self.target_input = self._sparse_data_var[0]
+
         cat_feat_dims_dict = OrderedDict()
         for line in open(cat_feat_num):
             spls = line.strip().split()
@@ -38,10 +49,11 @@ class Model(ModelBase):
             cat_feat_dims_dict[spls[0]] = int(spls[1])
         self.cat_feat_dims_dict = cat_feat_dims_dict if cat_feat_dims_dict else OrderedDict(
         )
-        self.is_sparse = envs.get_global_env("hyper_parameters.is_sparse", None, self._namespace)
+        self.is_sparse = envs.get_global_env("hyper_parameters.is_sparse",
+                                             None, self._namespace)
 
-        self.dense_feat_names = ['I' + str(i) for i in range(1, 14)]
-        self.sparse_feat_names = ['C' + str(i) for i in range(1, 27)]
+        self.dense_feat_names = [i.name for i in self.dense_inputs]
+        self.sparse_feat_names = [i.name for i in self.sparse_inputs]
 
         # {feat_name: dims}
         self.feat_dims_dict = OrderedDict(
@@ -51,21 +63,20 @@ class Model(ModelBase):
         self.net_input = None
         self.loss = None
 
-    def _create_embedding_input(self, data_dict):
+    def _create_embedding_input(self):
         # sparse embedding
-        sparse_emb_dict = OrderedDict((name, fluid.embedding(
-            input=fluid.layers.cast(
-                data_dict[name], dtype='int64'),
-            size=[
-                self.feat_dims_dict[name] + 1,
-                6 * int(pow(self.feat_dims_dict[name], 0.25))
-            ],
-            is_sparse=self.is_sparse)) for name in self.sparse_feat_names)
+        sparse_emb_dict = OrderedDict()
+        for var in self.sparse_inputs:
+            sparse_emb_dict[var.name] = fluid.embedding(
+                input=var,
+                size=[
+                    self.feat_dims_dict[var.name] + 1,
+                    6 * int(pow(self.feat_dims_dict[var.name], 0.25))
+                ],
+                is_sparse=self.is_sparse)
 
         # combine dense and sparse_emb
-        dense_input_list = [
-            data_dict[name] for name in data_dict if name.startswith('I')
-        ]
+        dense_input_list = self.dense_inputs
         sparse_emb_list = list(sparse_emb_dict.values())
 
         sparse_input = fluid.layers.concat(sparse_emb_list, axis=-1)
@@ -111,17 +122,13 @@ class Model(ModelBase):
         return fluid.layers.reduce_sum(fluid.layers.square(w))
 
     def train_net(self):
+        self.model._init_slots()
         self.init_network()
-        self.target_input = fluid.data(
-            name='label', shape=[None, 1], dtype='float32')
-        data_dict = OrderedDict()
-        for feat_name in self.feat_dims_dict:
-            data_dict[feat_name] = fluid.data(
-                name=feat_name, shape=[None, 1], dtype='float32')
 
-        self.net_input = self._create_embedding_input(data_dict)
+        self.net_input = self._create_embedding_input()
 
-        deep_out = self._deep_net(self.net_input, self.dnn_hidden_units, self.dnn_use_bn, False)
+        deep_out = self._deep_net(self.net_input, self.dnn_hidden_units,
+                                  self.dnn_use_bn, False)
 
         cross_out, l2_reg_cross_loss = self._cross_net(self.net_input,
                                                        self.cross_num)
@@ -130,9 +137,6 @@ class Model(ModelBase):
         logit = fluid.layers.fc(last_out, 1)
 
         self.prob = fluid.layers.sigmoid(logit)
-        self._data_var = [self.target_input] + [
-            data_dict[dense_name] for dense_name in self.dense_feat_names
-        ] + [data_dict[sparse_name] for sparse_name in self.sparse_feat_names]
 
         # auc
         prob_2d = fluid.layers.concat([1 - self.prob, self.prob], 1)
@@ -143,7 +147,9 @@ class Model(ModelBase):
         self._metrics["BATCH_AUC"] = batch_auc_var
 
         # logloss
-        logloss = fluid.layers.log_loss(self.prob, self.target_input)
+        logloss = fluid.layers.log_loss(
+            self.prob, fluid.layers.cast(
+                self.target_input, dtype='float32'))
         self.avg_logloss = fluid.layers.reduce_mean(logloss)
 
         # reg_coeff * l2_reg_cross
@@ -152,9 +158,11 @@ class Model(ModelBase):
         self._cost = self.loss
 
     def optimizer(self):
-        learning_rate = envs.get_global_env("hyper_parameters.learning_rate", None, self._namespace)
+        learning_rate = envs.get_global_env("hyper_parameters.learning_rate",
+                                            None, self._namespace)
         optimizer = fluid.optimizer.Adam(learning_rate, lazy_mode=True)
         return optimizer
 
     def infer_net(self, parameter_list):
+        self.model._init_slots()
         self.deepfm_net()

models/rank/deepfm/config.yaml

@@ -22,9 +22,10 @@ train:
 
   reader:
     batch_size: 2
-    class: "{workspace}/criteo_reader.py"
-    train_data_path: "{workspace}/data/train_data"
-    feat_dict_name: "{workspace}/data/aid_data/feat_dict_10.pkl2"
+    train_data_path: "{workspace}/data/slot_train_data"
+    feat_dict_name: "{workspace}/data/feat_dict_10.pkl2"
+    sparse_slots: "label feat_idx"
+    dense_slots: "feat_value:39"
 
   model:
     models: "{workspace}/model.py"

models/rank/deepfm/data/download_preprocess.py

+# 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 os
 import shutil
 import sys

models/rank/deepfm/criteo_reader.py → models/rank/deepfm/data/get_slot_data.py

@@ -12,18 +12,25 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from __future__ import print_function
-
+import yaml
+from paddlerec.core.reader import Reader
+from paddlerec.core.utils import envs
 try:
     import cPickle as pickle
 except ImportError:
     import pickle
 
-from paddlerec.core.reader import Reader
-from paddlerec.core.utils import envs
 
+class TrainReader(dg.MultiSlotDataGenerator):
+    def __init__(self, config):
+        dg.MultiSlotDataGenerator.__init__(self)
+
+        if os.path.isfile(config):
+            with open(config, 'r') as rb:
+                _config = yaml.load(rb.read(), Loader=yaml.FullLoader)
+        else:
+            raise ValueError("reader config only support yaml")
 
-class TrainReader(Reader):
     def init(self):
         self.cont_min_ = [0, -3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
         self.cont_max_ = [
@@ -37,7 +44,7 @@ class TrainReader(Reader):
         self.continuous_range_ = range(1, 14)
         self.categorical_range_ = range(14, 40)
         # load preprocessed feature dict 
-        self.feat_dict_name = envs.get_global_env("feat_dict_name", None, "train.reader")
+        self.feat_dict_name = "aid_data/feat_dict_10.pkl2"
         self.feat_dict_ = pickle.load(open(self.feat_dict_name, 'rb'))
 
     def _process_line(self, line):
@@ -70,6 +77,19 @@ class TrainReader(Reader):
 
         def data_iter():
             feat_idx, feat_value, label = self._process_line(line)
-            yield [('feat_idx', feat_idx), ('feat_value', feat_value), ('label', label)]
+            s = ""
+            for i in [('feat_idx', feat_idx), ('feat_value', feat_value),
+                      ('label', label)]:
+                k = i[0]
+                v = i[1]
+                for j in v:
+                    s += " " + k + ":" + str(j)
+            print s.strip()
+            yield None
 
         return data_iter
+
+
+reader = TrainReader("../config.yaml")
+reader.init()
+reader.run_from_stdin()

models/rank/deepfm/data/preprocess.py

+# 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 os
 import numpy
 from collections import Counter

models/rank/deepfm/data/run.sh

+python download_preprocess.py 
+
+mkdir slot_train_data
+for i in `ls ./train_data`
+do
+    cat train_data/$i | python get_slot_data.py > slot_train_data/$i
+done
+
+mkdir slot_test_data
+for i in `ls ./test_data`
+do
+    cat test_data/$i | python get_slot_data.py > slot_test_data/$i
+done

models/rank/deepfm/model.py

@@ -27,31 +27,26 @@ class Model(ModelBase):
     def deepfm_net(self):
         init_value_ = 0.1
         is_distributed = True if envs.get_trainer() == "CtrTrainer" else False
-        sparse_feature_number = envs.get_global_env("hyper_parameters.sparse_feature_number", None, self._namespace)
-        sparse_feature_dim = envs.get_global_env("hyper_parameters.sparse_feature_dim", None, self._namespace)
+        sparse_feature_number = envs.get_global_env(
+            "hyper_parameters.sparse_feature_number", None, self._namespace)
+        sparse_feature_dim = envs.get_global_env(
+            "hyper_parameters.sparse_feature_dim", None, self._namespace)
 
         # ------------------------- network input --------------------------
 
-        num_field = envs.get_global_env("hyper_parameters.num_field", None, self._namespace)
-        raw_feat_idx = fluid.data(name='feat_idx', shape=[None, num_field],
-                                  dtype='int64')  # None * num_field(defalut:39)
-        raw_feat_value = fluid.data(name='feat_value', shape=[None, num_field], dtype='float32')  # None * num_field
-        self.label = fluid.data(name='label', shape=[None, 1], dtype='float32')  # None * 1
-        feat_idx = fluid.layers.reshape(raw_feat_idx, [-1, 1])  # (None * num_field) * 1
-        feat_value = fluid.layers.reshape(raw_feat_value, [-1, num_field, 1])  # None * num_field * 1
+        num_field = envs.get_global_env("hyper_parameters.num_field", None,
+                                        self._namespace)
 
-        # ------------------------- set _data_var --------------------------
+        raw_feat_idx = self._sparse_data_var[1]
+        raw_feat_value = self._dense_data_var[0]
+        self.label = self._sparse_data_var[0]
 
-        self._data_var.append(raw_feat_idx)
-        self._data_var.append(raw_feat_value)
-        self._data_var.append(self.label)
-        if self._platform != "LINUX":
-            self._data_loader = fluid.io.DataLoader.from_generator(
-                feed_list=self._data_var, capacity=64, use_double_buffer=False, iterable=False)
+        feat_idx = raw_feat_idx
+        feat_value = fluid.layers.reshape(
+            raw_feat_value, [-1, num_field, 1])  # None * num_field * 1
 
-        # ------------------------- first order term --------------------------
-
-        reg = envs.get_global_env("hyper_parameters.reg", 1e-4, self._namespace)
+        reg = envs.get_global_env("hyper_parameters.reg", 1e-4,
+                                  self._namespace)
         first_weights_re = fluid.embedding(
             input=feat_idx,
             is_sparse=True,
@@ -65,7 +60,8 @@ class Model(ModelBase):
                 regularizer=fluid.regularizer.L1DecayRegularizer(reg)))
         first_weights = fluid.layers.reshape(
             first_weights_re, shape=[-1, num_field, 1])  # None * num_field * 1
-        y_first_order = fluid.layers.reduce_sum((first_weights * feat_value), 1)
+        y_first_order = fluid.layers.reduce_sum((first_weights * feat_value),
+                                                1)
 
         # ------------------------- second order term --------------------------
 
@@ -78,7 +74,8 @@ class Model(ModelBase):
             padding_idx=0,
             param_attr=fluid.ParamAttr(
                 initializer=fluid.initializer.TruncatedNormalInitializer(
-                    loc=0.0, scale=init_value_ / math.sqrt(float(sparse_feature_dim)))))
+                    loc=0.0,
+                    scale=init_value_ / math.sqrt(float(sparse_feature_dim)))))
         feat_embeddings = fluid.layers.reshape(
             feat_embeddings_re,
             shape=[-1, num_field,
@@ -86,8 +83,8 @@ class Model(ModelBase):
         feat_embeddings = feat_embeddings * feat_value  # None * num_field * embedding_size
 
         # sum_square part
-        summed_features_emb = fluid.layers.reduce_sum(feat_embeddings,
-                                                      1)  # None * embedding_size
+        summed_features_emb = fluid.layers.reduce_sum(
+            feat_embeddings, 1)  # None * embedding_size
         summed_features_emb_square = fluid.layers.square(
             summed_features_emb)  # None * embedding_size
 
@@ -98,13 +95,16 @@ class Model(ModelBase):
             squared_features_emb, 1)  # None * embedding_size
 
         y_second_order = 0.5 * fluid.layers.reduce_sum(
-            summed_features_emb_square - squared_sum_features_emb, 1,
+            summed_features_emb_square - squared_sum_features_emb,
+            1,
             keep_dim=True)  # None * 1
 
         # ------------------------- DNN --------------------------
 
-        layer_sizes = envs.get_global_env("hyper_parameters.fc_sizes", None, self._namespace)
-        act = envs.get_global_env("hyper_parameters.act", None, self._namespace)
+        layer_sizes = envs.get_global_env("hyper_parameters.fc_sizes", None,
+                                          self._namespace)
+        act = envs.get_global_env("hyper_parameters.act", None,
+                                  self._namespace)
         y_dnn = fluid.layers.reshape(feat_embeddings,
                                      [-1, num_field * sparse_feature_dim])
         for s in layer_sizes:
@@ -131,14 +131,17 @@ class Model(ModelBase):
 
         # ------------------------- DeepFM --------------------------
 
-        self.predict = fluid.layers.sigmoid(y_first_order + y_second_order + y_dnn)
+        self.predict = fluid.layers.sigmoid(y_first_order + y_second_order +
+                                            y_dnn)
 
     def train_net(self):
+        self.model._init_slots()
         self.deepfm_net()
 
         # ------------------------- Cost(logloss) --------------------------
 
-        cost = fluid.layers.log_loss(input=self.predict, label=self.label)
+        cost = fluid.layers.log_loss(
+            input=self.predict, label=fluid.layers.cast(self.label, "float32"))
         avg_cost = fluid.layers.reduce_sum(cost)
 
         self._cost = avg_cost
@@ -154,9 +157,11 @@ class Model(ModelBase):
         self._metrics["BATCH_AUC"] = batch_auc_var
 
     def optimizer(self):
-        learning_rate = envs.get_global_env("hyper_parameters.learning_rate", None, self._namespace)
+        learning_rate = envs.get_global_env("hyper_parameters.learning_rate",
+                                            None, self._namespace)
         optimizer = fluid.optimizer.Adam(learning_rate, lazy_mode=True)
         return optimizer
 
     def infer_net(self, parameter_list):
+        self.model._init_slots()
         self.deepfm_net()

models/rank/din/data/build_dataset.py

+# 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.
+
 from __future__ import print_function
 import random
 import pickle

models/rank/din/data/convert_pd.py

+# 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.
+
 from __future__ import print_function
 import pickle
 import pandas as pd

models/rank/din/data/remap_id.py

+# 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.
+
 from __future__ import print_function
 import random
 import pickle

models/rank/din/model.py

@@ -21,14 +21,14 @@ from paddlerec.core.model import Model as ModelBase
 class Model(ModelBase):
     def __init__(self, config):
         ModelBase.__init__(self, config)
-    
+
     def config_read(self, config_path):
         with open(config_path, "r") as fin:
             user_count = int(fin.readline().strip())
             item_count = int(fin.readline().strip())
             cat_count = int(fin.readline().strip())
         return user_count, item_count, cat_count
-    
+
     def din_attention(self, hist, target_expand, mask):
         """activation weight"""
 
@@ -58,56 +58,66 @@ class Model(ModelBase):
         out = fluid.layers.matmul(weight, hist)
         out = fluid.layers.reshape(x=out, shape=[0, hidden_size])
         return out
-    
+
     def train_net(self):
         seq_len = -1
-        self.item_emb_size = envs.get_global_env("hyper_parameters.item_emb_size", 64, self._namespace)
-        self.cat_emb_size = envs.get_global_env("hyper_parameters.cat_emb_size", 64, self._namespace)
-        self.act = envs.get_global_env("hyper_parameters.act", "sigmoid", self._namespace)
+        self.item_emb_size = envs.get_global_env(
+            "hyper_parameters.item_emb_size", 64, self._namespace)
+        self.cat_emb_size = envs.get_global_env(
+            "hyper_parameters.cat_emb_size", 64, self._namespace)
+        self.act = envs.get_global_env("hyper_parameters.act", "sigmoid",
+                                       self._namespace)
         #item_emb_size = 64
         #cat_emb_size = 64
-        self.is_sparse = envs.get_global_env("hyper_parameters.is_sparse", False, self._namespace)
+        self.is_sparse = envs.get_global_env("hyper_parameters.is_sparse",
+                                             False, self._namespace)
         #significant for speeding up the training process
-        self.config_path = envs.get_global_env("hyper_parameters.config_path", "data/config.txt", self._namespace)
-        self.use_DataLoader = envs.get_global_env("hyper_parameters.use_DataLoader", False, self._namespace)
+        self.config_path = envs.get_global_env(
+            "hyper_parameters.config_path", "data/config.txt", self._namespace)
+        self.use_DataLoader = envs.get_global_env(
+            "hyper_parameters.use_DataLoader", False, self._namespace)
         user_count, item_count, cat_count = self.config_read(self.config_path)
 
-
         item_emb_attr = fluid.ParamAttr(name="item_emb")
         cat_emb_attr = fluid.ParamAttr(name="cat_emb")
 
         hist_item_seq = fluid.data(
             name="hist_item_seq", shape=[None, seq_len], dtype="int64")
         self._data_var.append(hist_item_seq)
-        
+
         hist_cat_seq = fluid.data(
             name="hist_cat_seq", shape=[None, seq_len], dtype="int64")
         self._data_var.append(hist_cat_seq)
-        
-        target_item = fluid.data(name="target_item", shape=[None], dtype="int64")
+
+        target_item = fluid.data(
+            name="target_item", shape=[None], dtype="int64")
         self._data_var.append(target_item)
-        
+
         target_cat = fluid.data(name="target_cat", shape=[None], dtype="int64")
         self._data_var.append(target_cat)
-        
+
         label = fluid.data(name="label", shape=[None, 1], dtype="float32")
         self._data_var.append(label)
-        
-        mask = fluid.data(name="mask", shape=[None, seq_len, 1], dtype="float32")
+
+        mask = fluid.data(
+            name="mask", shape=[None, seq_len, 1], dtype="float32")
         self._data_var.append(mask)
-        
+
         target_item_seq = fluid.data(
             name="target_item_seq", shape=[None, seq_len], dtype="int64")
         self._data_var.append(target_item_seq)
-        
+
         target_cat_seq = fluid.data(
             name="target_cat_seq", shape=[None, seq_len], dtype="int64")
         self._data_var.append(target_cat_seq)
 
         if self.use_DataLoader:
             self._data_loader = fluid.io.DataLoader.from_generator(
-                feed_list=self._data_var, capacity=10000, use_double_buffer=False, iterable=False)
-        
+                feed_list=self._data_var,
+                capacity=10000,
+                use_double_buffer=False,
+                iterable=False)
+
         hist_item_emb = fluid.embedding(
             input=hist_item_seq,
             size=[item_count, self.item_emb_size],
@@ -149,7 +159,8 @@ class Model(ModelBase):
             size=[item_count, 1],
             param_attr=fluid.initializer.Constant(value=0.0))
 
-        hist_seq_concat = fluid.layers.concat([hist_item_emb, hist_cat_emb], axis=2)
+        hist_seq_concat = fluid.layers.concat(
+            [hist_item_emb, hist_cat_emb], axis=2)
         target_seq_concat = fluid.layers.concat(
             [target_item_seq_emb, target_cat_seq_emb], axis=2)
         target_concat = fluid.layers.concat(
@@ -157,21 +168,22 @@ class Model(ModelBase):
 
         out = self.din_attention(hist_seq_concat, target_seq_concat, mask)
         out_fc = fluid.layers.fc(name="out_fc",
-                                input=out,
-                                size=self.item_emb_size + self.cat_emb_size,
-                                num_flatten_dims=1)
+                                 input=out,
+                                 size=self.item_emb_size + self.cat_emb_size,
+                                 num_flatten_dims=1)
         embedding_concat = fluid.layers.concat([out_fc, target_concat], axis=1)
 
         fc1 = fluid.layers.fc(name="fc1",
-                            input=embedding_concat,
-                            size=80,
-                            act=self.act)
+                              input=embedding_concat,
+                              size=80,
+                              act=self.act)
         fc2 = fluid.layers.fc(name="fc2", input=fc1, size=40, act=self.act)
         fc3 = fluid.layers.fc(name="fc3", input=fc2, size=1)
         logit = fc3 + item_b
 
-        loss = fluid.layers.sigmoid_cross_entropy_with_logits(x=logit, label=label)
-        
+        loss = fluid.layers.sigmoid_cross_entropy_with_logits(
+            x=logit, label=label)
+
         avg_loss = fluid.layers.mean(loss)
         self._cost = avg_loss
 
@@ -179,14 +191,14 @@ class Model(ModelBase):
         predict_2d = fluid.layers.concat([1 - self.predict, self.predict], 1)
         label_int = fluid.layers.cast(label, 'int64')
         auc_var, batch_auc_var, _ = fluid.layers.auc(input=predict_2d,
-                                                            label=label_int,
-                                                            slide_steps=0)
+                                                     label=label_int,
+                                                     slide_steps=0)
         self._metrics["AUC"] = auc_var
         self._metrics["BATCH_AUC"] = batch_auc_var
 
-
     def optimizer(self):
-        learning_rate = envs.get_global_env("hyper_parameters.learning_rate", None, self._namespace)
+        learning_rate = envs.get_global_env("hyper_parameters.learning_rate",
+                                            None, self._namespace)
         optimizer = fluid.optimizer.Adam(learning_rate, lazy_mode=True)
         return optimizer
 

models/rank/din/reader.py

@@ -29,13 +29,15 @@ from paddlerec.core.utils import envs
 
 class TrainReader(Reader):
     def init(self):
-        self.train_data_path = envs.get_global_env("train_data_path", None, "train.reader")
+        self.train_data_path = envs.get_global_env("train_data_path", None,
+                                                   "train.reader")
         self.res = []
         self.max_len = 0
 
         data_file_list = os.listdir(self.train_data_path)
         for i in range(0, len(data_file_list)):
-            train_data_file = os.path.join(self.train_data_path, data_file_list[i])
+            train_data_file = os.path.join(self.train_data_path,
+                                           data_file_list[i])
             with open(train_data_file, "r") as fin:
                 for line in fin:
                     line = line.strip().split(';')
@@ -78,11 +80,13 @@ class TrainReader(Reader):
         len_array = [len(x[0]) for x in b]
         mask = np.array(
             [[0] * x + [-1e9] * (max_len - x) for x in len_array]).reshape(
-            [-1, max_len, 1])
+                [-1, max_len, 1])
         target_item_seq = np.array(
-            [[x[2]] * max_len for x in b]).astype("int64").reshape([-1, max_len])
+            [[x[2]] * max_len for x in b]).astype("int64").reshape(
+                [-1, max_len])
         target_cat_seq = np.array(
-            [[x[3]] * max_len for x in b]).astype("int64").reshape([-1, max_len])
+            [[x[3]] * max_len for x in b]).astype("int64").reshape(
+                [-1, max_len])
         res = []
         for i in range(len(b)):
             res.append([
@@ -127,4 +131,5 @@ class TrainReader(Reader):
     def generate_batch_from_trainfiles(self, files):
         data_set = self.base_read(files)
         random.shuffle(data_set)
-        return self.batch_reader(data_set, self.batch_size, self.batch_size * 20)
+        return self.batch_reader(data_set, self.batch_size,
+                                 self.batch_size * 20)

models/rank/dnn/config.yaml

@@ -23,9 +23,10 @@ train:
 
   reader:
     batch_size: 2
-    class: "{workspace}/../criteo_reader.py"
-    train_data_path: "{workspace}/data/train"
+    train_data_path: "{workspace}/data/slot_train_data"
     reader_debug_mode: False
+    sparse_slots: "click 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26"
+    dense_slots: "dense_var:13"
 
   model:
     models: "{workspace}/model.py"

models/rank/dnn/data/download.sh

+wget --no-check-certificate https://fleet.bj.bcebos.com/ctr_data.tar.gz
+tar -zxvf ctr_data.tar.gz
+mv ./raw_data ./train_data_full
+mkdir train_data && cd train_data
+cp ../train_data_full/part-0 ../train_data_full/part-1 ./ && cd ..
+mv ./test_data ./test_data_full
+mkdir test_data && cd test_data
+cp ../test_data_full/part-220 ./  && cd ..
+echo "Complete data download."
+echo "Full Train data stored in ./train_data_full "
+echo "Full Test data stored in ./test_data_full "
+echo "Rapid Verification train data stored in ./train_data "
+echo "Rapid Verification test data stored in ./test_data "

models/rank/criteo_reader.py → models/rank/dnn/data/get_slot_data.py

-#   Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#   Copyright (c) 2019 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.
@@ -12,20 +12,21 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from __future__ import print_function
+import paddle.fluid.incubate.data_generator as dg
 
-from paddlerec.core.reader import Reader
-from paddlerec.core.utils import envs
+cont_min_ = [0, -3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+cont_max_ = [20, 600, 100, 50, 64000, 500, 100, 50, 500, 10, 10, 10, 50]
+cont_diff_ = [20, 603, 100, 50, 64000, 500, 100, 50, 500, 10, 10, 10, 50]
+hash_dim_ = 1000001
+continuous_range_ = range(1, 14)
+categorical_range_ = range(14, 40)
 
 
-class TrainReader(Reader):
-    def init(self):
-        self.cont_min_ = [0, -3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
-        self.cont_max_ = [20, 600, 100, 50, 64000, 500, 100, 50, 500, 10, 10, 10, 50]
-        self.cont_diff_ = [20, 603, 100, 50, 64000, 500, 100, 50, 500, 10, 10, 10, 50]
-        self.hash_dim_ = envs.get_global_env("hyper_parameters.sparse_feature_number", None, "train.model")
-        self.continuous_range_ = range(1, 14)
-        self.categorical_range_ = range(14, 40)
+class CriteoDataset(dg.MultiSlotDataGenerator):
+    """
+    DacDataset: inheritance MultiSlotDataGeneratior, Implement data reading
+    Help document: http://wiki.baidu.com/pages/viewpage.action?pageId=728820675
+    """
 
     def generate_sample(self, line):
         """
@@ -37,25 +38,34 @@ class TrainReader(Reader):
             This function needs to be implemented by the user, based on data format
             """
             features = line.rstrip('\n').split('\t')
-
             dense_feature = []
             sparse_feature = []
-            for idx in self.continuous_range_:
+            for idx in continuous_range_:
                 if features[idx] == "":
                     dense_feature.append(0.0)
                 else:
                     dense_feature.append(
-                        (float(features[idx]) - self.cont_min_[idx - 1]) /
-                        self.cont_diff_[idx - 1])
-
-            for idx in self.categorical_range_:
+                        (float(features[idx]) - cont_min_[idx - 1]) /
+                        cont_diff_[idx - 1])
+            for idx in categorical_range_:
                 sparse_feature.append(
-                    [hash(str(idx) + features[idx]) % self.hash_dim_])
+                    [hash(str(idx) + features[idx]) % hash_dim_])
             label = [int(features[0])]
-            feature_name = ["D"]
-            for idx in self.categorical_range_:
-                feature_name.append("S" + str(idx - 13))
+            process_line = dense_feature, sparse_feature, label
+            feature_name = ["dense_feature"]
+            for idx in categorical_range_:
+                feature_name.append("C" + str(idx - 13))
             feature_name.append("label")
-            yield zip(feature_name, [dense_feature] + sparse_feature + [label])
+            s = "click:" + str(label[0])
+            for i in dense_feature:
+                s += " dense_feature:" + str(i)
+            for i in range(1, 1 + len(categorical_range_)):
+                s += " " + str(i) + ":" + str(sparse_feature[i - 1][0])
+            print s.strip()
+            yield None
 
         return reader
+
+
+d = CriteoDataset()
+d.run_from_stdin()

models/rank/dnn/data/run.sh

+sh download.sh
+
+mkdir slot_train_data_full
+for i in `ls ./train_data_full`
+do
+    cat train_data_full/$i | python get_slot_data.py > slot_train_data_full/$i
+done
+
+mkdir slot_test_data_full
+for i in `ls ./test_data_full`
+do
+    cat test_data_full/$i | python get_slot_data.py > slot_test_data_full/$i
+done
+
+mkdir slot_train_data
+for i in `ls ./train_data`
+do
+    cat train_data/$i | python get_slot_data.py > slot_train_data/$i
+done
+
+mkdir slot_test_data
+for i in `ls ./test_data`
+do
+    cat test_data/$i | python get_slot_data.py > slot_test_data/$i
+done

models/rank/dnn/model.py

@@ -25,48 +25,16 @@ class Model(ModelBase):
         ModelBase.__init__(self, config)
 
     def input(self):
-        def sparse_inputs():
-            ids = envs.get_global_env("hyper_parameters.sparse_inputs_slots", None, self._namespace)
-
-            sparse_input_ids = [
-                fluid.layers.data(name="S" + str(i),
-                                  shape=[1],
-                                  lod_level=1,
-                                  dtype="int64") for i in range(1, ids)
-            ]
-            return sparse_input_ids
-
-        def dense_input():
-            dim = envs.get_global_env("hyper_parameters.dense_input_dim", None, self._namespace)
-
-            dense_input_var = fluid.layers.data(name="D",
-                                                shape=[dim],
-                                                dtype="float32")
-            return dense_input_var
-
-        def label_input():
-            label = fluid.layers.data(name="click", shape=[1], dtype="int64")
-            return label
-
-        self.sparse_inputs = sparse_inputs()
-        self.dense_input = dense_input()
-        self.label_input = label_input()
-
-        self._data_var.append(self.dense_input)
-
-        for input in self.sparse_inputs:
-            self._data_var.append(input)
-
-        self._data_var.append(self.label_input)
-
-        if self._platform != "LINUX":
-            self._data_loader = fluid.io.DataLoader.from_generator(
-                feed_list=self._data_var, capacity=64, use_double_buffer=False, iterable=False)
+        self.sparse_inputs = self._sparse_data_var[1:]
+        self.dense_input = self._dense_data_var[0]
+        self.label_input = self._sparse_data_var[0]
 
     def net(self):
         is_distributed = True if envs.get_trainer() == "CtrTrainer" else False
-        sparse_feature_number = envs.get_global_env("hyper_parameters.sparse_feature_number", None, self._namespace)
-        sparse_feature_dim = envs.get_global_env("hyper_parameters.sparse_feature_dim", None, self._namespace)
+        sparse_feature_number = envs.get_global_env(
+            "hyper_parameters.sparse_feature_number", None, self._namespace)
+        sparse_feature_dim = envs.get_global_env(
+            "hyper_parameters.sparse_feature_dim", None, self._namespace)
 
         def embedding_layer(input):
             emb = fluid.layers.embedding(
@@ -76,25 +44,27 @@ class Model(ModelBase):
                 size=[sparse_feature_number, sparse_feature_dim],
                 param_attr=fluid.ParamAttr(
                     name="SparseFeatFactors",
-                    initializer=fluid.initializer.Uniform()),
-            )
-            emb_sum = fluid.layers.sequence_pool(
-                input=emb, pool_type='sum')
+                    initializer=fluid.initializer.Uniform()), )
+            emb_sum = fluid.layers.sequence_pool(input=emb, pool_type='sum')
             return emb_sum
 
         def fc(input, output_size):
             output = fluid.layers.fc(
-                input=input, size=output_size,
-                act='relu', param_attr=fluid.ParamAttr(
+                input=input,
+                size=output_size,
+                act='relu',
+                param_attr=fluid.ParamAttr(
                     initializer=fluid.initializer.Normal(
                         scale=1.0 / math.sqrt(input.shape[1]))))
             return output
 
         sparse_embed_seq = list(map(embedding_layer, self.sparse_inputs))
-        concated = fluid.layers.concat(sparse_embed_seq + [self.dense_input], axis=1)
+        concated = fluid.layers.concat(
+            sparse_embed_seq + [self.dense_input], axis=1)
 
         fcs = [concated]
-        hidden_layers = envs.get_global_env("hyper_parameters.fc_sizes", None, self._namespace)
+        hidden_layers = envs.get_global_env("hyper_parameters.fc_sizes", None,
+                                            self._namespace)
 
         for size in hidden_layers:
             fcs.append(fc(fcs[-1], size))
@@ -109,29 +79,33 @@ class Model(ModelBase):
         self.predict = predict
 
     def avg_loss(self):
-        cost = fluid.layers.cross_entropy(input=self.predict, label=self.label_input)
+        cost = fluid.layers.cross_entropy(
+            input=self.predict, label=self.label_input)
         avg_cost = fluid.layers.reduce_mean(cost)
         self._cost = avg_cost
 
     def metrics(self):
         auc, batch_auc, _ = fluid.layers.auc(input=self.predict,
                                              label=self.label_input,
-                                             num_thresholds=2 ** 12,
+                                             num_thresholds=2**12,
                                              slide_steps=20)
         self._metrics["AUC"] = auc
         self._metrics["BATCH_AUC"] = batch_auc
 
     def train_net(self):
+        self.model._init_slots()
         self.input()
         self.net()
         self.avg_loss()
         self.metrics()
 
     def optimizer(self):
-        learning_rate = envs.get_global_env("hyper_parameters.learning_rate", None, self._namespace)
+        learning_rate = envs.get_global_env("hyper_parameters.learning_rate",
+                                            None, self._namespace)
         optimizer = fluid.optimizer.Adam(learning_rate, lazy_mode=True)
         return optimizer
 
     def infer_net(self):
+        self.model._init_slots()
         self.input()
         self.net()

models/rank/readme.md

@@ -59,6 +59,13 @@
 ## 使用教程
 ### 数据处理
 参考每个模型目录数据下载&预处理脚本
+
+```
+sh run.sh
+```
+
+数据读取默认使用core/reader.py
+
 ### 训练
 ```
 python -m paddlerec.run -m paddlerec.models.rank.dnn # 以DNN为例

models/rank/wide_deep/config.yaml

@@ -22,8 +22,9 @@ train:
 
   reader:
     batch_size: 2
-    class: "{workspace}/reader.py"
-    train_data_path: "{workspace}/data/train_data"
+    train_data_path: "{workspace}/data/slot_train_data"
+    sparse_slots: "label"
+    dense_slots: "wide_input:8 deep_input:58"
 
   model:
     models: "{workspace}/model.py"

models/rank/wide_deep/create_data.sh → models/rank/wide_deep/data/create_data.sh

 mkdir train_data
 mkdir test_data
-mkdir data
-train_path="/home/yaoxuefeng/repos/models/models/PaddleRec/ctr/wide_deep/data/adult.data"
-test_path="/home/yaoxuefeng/repos/models/models/PaddleRec/ctr/wide_deep/data/adult.test"
-train_data_path="/home/yaoxuefeng/repos/models/models/PaddleRec/ctr/wide_deep/train_data/train_data.csv"
-test_data_path="/home/yaoxuefeng/repos/models/models/PaddleRec/ctr/wide_deep/test_data/test_data.csv"
+train_path="adult.data"
+test_path="adult.test"
+train_data_path="./train_data/train_data.csv"
+test_data_path="./test_data/test_data.csv"
 
-#pip install -r requirements.txt
+pip install -r requirements.txt
 
-#wget -P data/ https://archive.ics.uci.edu/ml/machine-learning-databases/adult/adult.data
-#wget -P data/ https://archive.ics.uci.edu/ml/machine-learning-databases/adult/adult.test
+wget -P data/ https://archive.ics.uci.edu/ml/machine-learning-databases/adult/adult.data
+wget -P data/ https://archive.ics.uci.edu/ml/machine-learning-databases/adult/adult.test
 
 python data_preparation.py --train_path ${train_path} \
                            --test_path ${test_path} \

models/rank/wide_deep/data/data_preparation.py

+# 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 os
+import io
+import args
+import pandas as pd
+from sklearn import preprocessing
+
+
+def _clean_file(source_path, target_path):
+    """makes changes to match the CSV format."""
+    with io.open(source_path, 'r') as temp_eval_file:
+        with io.open(target_path, 'w') as eval_file:
+            for line in temp_eval_file:
+                line = line.strip()
+                line = line.replace(', ', ',')
+                if not line or ',' not in line:
+                    continue
+                if line[-1] == '.':
+                    line = line[:-1]
+                line += '\n'
+                eval_file.write(line)
+
+
+def build_model_columns(train_data_path, test_data_path):
+    # The column names are from
+    # https://www2.1010data.com/documentationcenter/prod/Tutorials/MachineLearningExamples/CensusIncomeDataSet.html
+    column_names = [
+        'age', 'workclass', 'fnlwgt', 'education', 'education_num',
+        'marital_status', 'occupation', 'relationship', 'race', 'gender',
+        'capital_gain', 'capital_loss', 'hours_per_week', 'native_country',
+        'income_bracket'
+    ]
+
+    # Load the dataset in Pandas
+    train_df = pd.read_csv(
+        train_data_path,
+        delimiter=',',
+        header=None,
+        index_col=None,
+        names=column_names)
+    test_df = pd.read_csv(
+        test_data_path,
+        delimiter=',',
+        header=None,
+        index_col=None,
+        names=column_names)
+
+    # First group of tasks according to the paper
+    #label_columns = ['income_50k', 'marital_stat']
+    categorical_columns = [
+        'education', 'marital_status', 'relationship', 'workclass',
+        'occupation'
+    ]
+    for col in categorical_columns:
+        label_train = preprocessing.LabelEncoder()
+        train_df[col] = label_train.fit_transform(train_df[col])
+        label_test = preprocessing.LabelEncoder()
+        test_df[col] = label_test.fit_transform(test_df[col])
+
+    bins = [18, 25, 30, 35, 40, 45, 50, 55, 60, 65]
+    train_df['age_buckets'] = pd.cut(train_df['age'].values.tolist(),
+                                     bins,
+                                     labels=False)
+    test_df['age_buckets'] = pd.cut(test_df['age'].values.tolist(),
+                                    bins,
+                                    labels=False)
+
+    base_columns = [
+        'education', 'marital_status', 'relationship', 'workclass',
+        'occupation', 'age_buckets'
+    ]
+
+    train_df['education_occupation'] = train_df['education'].astype(
+        str) + '_' + train_df['occupation'].astype(str)
+    test_df['education_occupation'] = test_df['education'].astype(
+        str) + '_' + test_df['occupation'].astype(str)
+    train_df['age_buckets_education_occupation'] = train_df[
+        'age_buckets'].astype(str) + '_' + train_df['education'].astype(
+            str) + '_' + train_df['occupation'].astype(str)
+    test_df['age_buckets_education_occupation'] = test_df[
+        'age_buckets'].astype(str) + '_' + test_df['education'].astype(
+            str) + '_' + test_df['occupation'].astype(str)
+    crossed_columns = [
+        'education_occupation', 'age_buckets_education_occupation'
+    ]
+
+    for col in crossed_columns:
+        label_train = preprocessing.LabelEncoder()
+        train_df[col] = label_train.fit_transform(train_df[col])
+        label_test = preprocessing.LabelEncoder()
+        test_df[col] = label_test.fit_transform(test_df[col])
+
+    wide_columns = base_columns + crossed_columns
+
+    train_df_temp = pd.get_dummies(
+        train_df[categorical_columns], columns=categorical_columns)
+    test_df_temp = pd.get_dummies(
+        test_df[categorical_columns], columns=categorical_columns)
+    train_df = train_df.join(train_df_temp)
+    test_df = test_df.join(test_df_temp)
+
+    deep_columns = list(train_df_temp.columns) + [
+        'age', 'education_num', 'capital_gain', 'capital_loss',
+        'hours_per_week'
+    ]
+
+    train_df['label'] = train_df['income_bracket'].apply(
+        lambda x: 1 if x == '>50K' else 0)
+    test_df['label'] = test_df['income_bracket'].apply(
+        lambda x: 1 if x == '>50K' else 0)
+
+    with io.open('train_data/columns.txt', 'w') as f:
+        write_str = str(len(wide_columns)) + '\n' + str(len(
+            deep_columns)) + '\n'
+        f.write(write_str)
+        f.close()
+    with io.open('test_data/columns.txt', 'w') as f:
+        write_str = str(len(wide_columns)) + '\n' + str(len(
+            deep_columns)) + '\n'
+        f.write(write_str)
+        f.close()
+
+    train_df[wide_columns + deep_columns + ['label']].fillna(0).to_csv(
+        train_data_path, index=False)
+    test_df[wide_columns + deep_columns + ['label']].fillna(0).to_csv(
+        test_data_path, index=False)
+
+
+def clean_file(train_path, test_path, train_data_path, test_data_path):
+    _clean_file(train_path, train_data_path)
+    _clean_file(test_path, test_data_path)
+
+
+if __name__ == '__main__':
+    args = args.parse_args()
+    clean_file(args.train_path, args.test_path, args.train_data_path,
+               args.test_data_path)
+    build_model_columns(args.train_data_path, args.test_data_path)