Merge pull request #10 from RUCAIBox/master

update

main.py

@@ -1,23 +1,24 @@
-from config import Config
-from data import Dataset, data_preparation
-from model.general_recommender.bprmf import BPRMF
-from trainer import Trainer
-from utils import init_logger
 from logging import getLogger
+from recbox.config import Config
+from recbox.data import Dataset, data_preparation
+from recbox.model.general_recommender.bprmf import BPRMF
+from recbox.trainer import Trainer
+from recbox.utils import init_logger, get_model
 
 config = Config('properties/overall.config')
 config.init()
 init_logger(config)
 logger = getLogger()
+
 dataset = Dataset(config)
 logger.info(dataset)
 
-model = BPRMF(config, dataset).to(config['device'])
-logger.info(model)
-
 # If you want to customize the evaluation setting,
 # please refer to `data_preparation()` in `data/utils.py`.
-train_data, test_data, valid_data = data_preparation(config, model, dataset)
+train_data, test_data, valid_data = data_preparation(config, dataset)
+
+model = get_model(config['model'])(config, train_data).to(config['device'])
+logger.info(model)
 
 trainer = Trainer(config, model)
 

properties/model/DMF.config

@@ -0,0 +1,5 @@
+[model]
+
+layers = [64,64]
+reg_weight = 0.0
+min_y_hat = 1e-6

properties/model/GRU4Rec.config

@@ -1,5 +1,11 @@
 [model]
 
+MAX_ITEM_LIST_LENGTH=50
+TARGET_PREFIX='target_'
+LIST_SUFFIX='_list'
+POSITION_FIELD='position_id'
+ITEM_LIST_LENGTH_FIELD='item_length'
+
 embedding_size=64
-num_layers=3
-dropout=0.1
+num_layers=1
+dropout=0

properties/model/ItemKNN.config

@@ -0,0 +1,3 @@
+[model]
+
+k=100

properties/model/NGCF.config

@@ -2,6 +2,6 @@
 
 embedding_size=64
 layers=[64,64,64]
-node_dropout=0.1
+node_dropout=0.0
 message_dropout=0.1
 delay=1e-5

properties/model/SASRec.config

@@ -0,0 +1,13 @@
+[model]
+
+MAX_ITEM_LIST_LENGTH=50
+TARGET_PREFIX='target_'
+LIST_SUFFIX='_list'
+POSITION_FIELD='position_id'
+ITEM_LIST_LENGTH_FIELD='item_length'
+
+embedding_size=64
+n_head=1
+d_ff=128
+dropout=0
+num_blocks=1

recbox/config/__init__.py

@@ -0,0 +1,2 @@
+from .configurator import Config
+from .eval_setting import EvalSetting

config/abstract_configurator.py → recbox/config/abstract_configurator.py

@@ -1,6 +1,6 @@
 import os
-from utils.enum_type import *
 from configparser import ConfigParser
+from ..utils.enum_type import *
 
 
 class AbstractConfig(object):

config/cmd_configurator.py → recbox/config/cmd_configurator.py

@@ -1,4 +1,4 @@
-from config.abstract_configurator import AbstractConfig
+from .abstract_configurator import AbstractConfig
 
 
 class CmdConfig(AbstractConfig):

config/configurator.py → recbox/config/configurator.py

@@ -3,10 +3,10 @@
 import random
 import sys
 import numpy as np
-from config.running_configurator import RunningConfig
-from config.model_configurator import ModelConfig
-from config.data_configurator import DataConfig
-from config.cmd_configurator import CmdConfig
+from .running_configurator import RunningConfig
+from .model_configurator import ModelConfig
+from .data_configurator import DataConfig
+from .cmd_configurator import CmdConfig
 
 
 class Config(object):
@@ -61,24 +61,27 @@ def __init__(self, config_file_name, config_dict=None):
         self.dataset_args = DataConfig(dataset_arg_file_name, self.cmd_args_dict)
 
         self.device = None
+        self._init_device()
 
-    def init(self):
+    def _init_device(self):
         """
         This function is a global initialization function that fix random seed and gpu device.
         """
-        init_seed = self.run_args['seed']
         use_gpu = self.run_args['use_gpu']
         if use_gpu:
             gpu_id = self.run_args['gpu_id']
             os.environ["CUDA_VISIBLE_DEVICES"] = str(gpu_id)
         # Get the device that run on.
         self.device = torch.device("cuda" if torch.cuda.is_available() and use_gpu else "cpu")
 
+    def init(self):
+        init_seed = self.run_args['seed']
         random.seed(init_seed)
         np.random.seed(init_seed)
         torch.manual_seed(init_seed)
         torch.cuda.manual_seed(init_seed)
         torch.cuda.manual_seed_all(init_seed)
+        torch.backends.cudnn.benchmark = False
         torch.backends.cudnn.deterministic = True
 
     def _read_cmd_line(self):

config/data_configurator.py → recbox/config/data_configurator.py

@@ -1,4 +1,4 @@
-from config.abstract_configurator import AbstractConfig
+from .abstract_configurator import AbstractConfig
 
 
 class DataConfig(AbstractConfig):

config/eval_setting.py → recbox/config/eval_setting.py

@@ -3,10 +3,11 @@
 # @Email  : houyupeng@ruc.edu.cn
 
 # UPDATE:
-# @Time   : 2020/8/13 11:11
+# @Time   : 2020/8/19 18:56
 # @Author : Yupeng Hou
 # @Email  : houyupeng@ruc.edu.cn
 
+
 class EvalSetting(object):
     def __init__(self, config):
         self.config = config
@@ -76,14 +77,16 @@ def random_ordering(self):
         self.set_ordering('shuffle')
 
     def sort_by(self, field, ascending=None):
+        if not isinstance(field, list):
+            field = [field]
         if ascending is None:
             ascending = [True] * len(field)
             if len(ascending) == 1:
                 ascending = True
         self.set_ordering('by', field=field, ascending=ascending)
 
     def temporal_ordering(self):
-        self.sort_by(field=self.config['TIMESTAMP_FIELD'])
+        self.sort_by(field=self.config['TIME_FIELD'])
 
     r"""Setting about split method
 

config/model_configurator.py → recbox/config/model_configurator.py

@@ -1,4 +1,4 @@
-from config.abstract_configurator import AbstractConfig
+from .abstract_configurator import AbstractConfig
 
 
 class ModelConfig(AbstractConfig):

config/running_configurator.py → recbox/config/running_configurator.py

@@ -1,7 +1,7 @@
-from config.abstract_configurator import AbstractConfig
 import os
-from evaluator import loss_metrics, topk_metrics
-from utils import EvaluatorType
+from ..evaluator import loss_metrics, topk_metrics
+from ..utils import EvaluatorType, get_model
+from .abstract_configurator import AbstractConfig
 
 
 class RunningConfig(AbstractConfig):
@@ -39,7 +39,7 @@ def __init__(self, config_file_name, cmd_args):
                     eval_type = EvaluatorType.RANKING
         self['eval_type'] = eval_type
 
-        smaller_metric = ['rmse','mae', 'logloss']
+        smaller_metric = ['rmse', 'mae', 'logloss']
 
         if 'valid_metric' not in self:
             valid_metric = self['metric'][0]
@@ -54,6 +54,6 @@ def __init__(self, config_file_name, cmd_args):
             else:
                 self['valid_metric_bigger'] = True
 
-
-
-
+        model = get_model(self['model'])
+        self['MODEL_TYPE'] = model.type
+        self['MODEL_INPUT_TYPE'] = model.input_type