def main(_): params = ModelParams() for key, value in params.__dict__.items(): print(key, "=", value) if params.alg_name == "dnn": model = dnn.model_estimator(params) elif params.alg_name == "dnn_pool": model = dnn_pool.model_estimator(params) elif params.alg_name == "dnn_cate": model = dnn_cate.model_estimator(params) elif params.alg_name == "deepfm": model = deepfm.model_estimator(params) elif params.alg_name == "deepfm_pool": model = deepfm_pool.model_estimator(params) elif params.alg_name == "deepfm_cate": model = deepfm_cate.model_estimator(params) elif params.alg_name == "din": model = din.model_estimator(params) elif params.alg_name == "dinfm": model = dinfm.model_estimator(params) elif params.alg_name == "dien": model = dien.model_estimator(params) elif params.alg_name == "dnn_emb": model = dnn_emb.model_estimator(params) elif params.alg_name == "dnn_autoint": model = dnn_autoint.model_estimator(params) else: model = dnn.model_estimator(params) print("alg_name = %s is error" % params.alg_name) exit(-1) if params.action_type == "train": start_time = time.time() train_files = data_load.get_file_list(params.train_path) predict_files = data_load.get_file_list(params.predict_path) print("--------------train------------") trained_model_path = op.model_fit(model, params, train_files, predict_files) end_time = time.time() print("model_save training time: %.2f s" % (end_time - start_time)) # save model_pb path to a file f = tf.gfile.GFile(params.model_pb + "/test", 'w') f.write(str(trained_model_path, encoding="utf-8")) print("--------------predict------------") op.model_predict(trained_model_path, predict_files, params) elif params.action_type == "pred": print("--------------predict------------") predict_files = data_load.get_file_list(params.predict_path) op.model_predict( '/Users/R.Stalker/PycharmProjects/deep_learing_estimator/files/model_save_pb/deepfm', predict_files, params) else: print("action_type = %s is error !!!" % params.action_type)
def main(_): params = ModelParams() for key, value in params.__dict__.items(): print(key, "=", value) print("---delete old data...") delete_dt = my_utils.shift_date_time(params.dt, -1) print("---delete_dt:", delete_dt) print(params.train_path[:-9] + delete_dt) print(params.predict_path[:-9] + delete_dt) shutil.rmtree(params.train_path[:-9] + delete_dt, ignore_errors=True) shutil.rmtree(params.predict_path[:-9] + delete_dt, ignore_errors=True) if params.alg_name == "dnn": model = dnn.model_estimator(params) elif params.alg_name == "deepfm": model = deepfm.model_estimator(params) elif params.alg_name == "deepfm_pool": model = deepfm_pool.model_estimator(params) elif params.alg_name == "din": model = din.model_estimator(params) elif params.alg_name == "dnn_pool": model = dnn_pool.model_estimator(params) elif params.alg_name == "dinfm": model = dinfm.model_estimator(params) elif params.alg_name == "dien": model = dien.model_estimator(params) elif params.alg_name == "dnn_autoint": model = dnn_autoint.model_estimator(params) else: model = dnn.model_estimator(params) print("alg_name = %s is error" % params.alg_name) exit(-1) if params.mode == "train": start_time = time.time() train_files = data_load.get_file_list(params.train_path) predict_files = data_load.get_file_list(params.predict_path) print("--------------train------------") trained_model_path = op.model_fit(model, params, train_files, predict_files) end_time = time.time() # save model_pb path to a file f = tf.gfile.GFile(params.model_pb[:-9] + "latest_model_path", 'w') f.write(str(trained_model_path, encoding="utf-8")) print("model_save training time: %.2f s" % (end_time - start_time)) print("--------------predict------------") op.model_predict(trained_model_path, predict_files, params) elif params.mode == "eval": print("--------------predict------------") predict_files = data_load.get_file_list(params.predict_path) op.model_predict(params.model_pb, predict_files, params) else: print("action_type = %s is error !!!" % params.mode)
def main(_): params = ModelParams() for key, value in params.__dict__.items(): print(key, "=", value) if params.alg_name == "esmm": model = esmm.model_estimator(params) elif params.alg_name == "mmoe": model = mmoe.model_estimator(params) else: model = esmm.model_estimator(params) print("alg_name = %s is error" % params.alg_name) exit(-1) if params.action_type == "train": start_time = time.time() train_files = data_load.get_file_list(params.train_path) predict_files = data_load.get_file_list(params.predict_path) print("--------------train------------") trained_model_path = op.model_fit(model, params, train_files, predict_files) end_time = time.time() print("model_save training time: %.2f s" % (end_time - start_time)) # save model_pb path to a file f = tf.gfile.GFile(params.model_pb + "/test", 'w') f.write(str(trained_model_path, encoding="utf-8")) print("--------------predict------------") op.model_predict(trained_model_path, predict_files, params) elif params.action_type == "pred": print("--------------predict------------") predict_files = data_load.get_file_list(params.predict_path) op.model_predict('files/model_save_pb/esmm/1582094784', predict_files, params) else: print("action_type = %s is error !!!" % params.action_type)