Skip to content

G
gm_strategy_cvr
Commit 31aa4e0e authored by 赵威's avatar 赵威
Browse files
Options

update tractate model

parent e1ead83a
Hide whitespace changes
Inline Side-by-side
Showing with 60 additions and 62 deletions
src/models/esmm/diary_model.py
View file @ 31aa4e0e
......@@ -92,7 +92,6 @@ _int_columns = [
"business_second_skip_num",
"service_price",
"service_sold_num",
"recommend_service_price",
]
_float_columns = [
"one_ctr",
......@@ -179,6 +178,7 @@ _categorical_columns = [
"service_city",
"recommend_service_id",
"recommend_service_city",
"recommend_service_price",
"device_fd2",
"device_sd2",
"device_fs2",
......
src/train_tractate.py
View file @ 31aa4e0e
......@@ -22,71 +22,69 @@ def main():
tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.INFO)
# tractate_train_columns = set(tractate_fe.INT_COLUMNS + tractate_fe.FLOAT_COLUMNS + tractate_fe.CATEGORICAL_COLUMNS)
# print("features: " + str(len(tractate_train_columns)))
# tractate_predict_columns = set(PREDICTION_ALL_COLUMNS)
# print(tractate_predict_columns.difference(tractate_train_columns))
# print(tractate_train_columns.difference(tractate_predict_columns))
# assert tractate_predict_columns == tractate_train_columns
# # dataset_path = Path("~/data/cvr_data").expanduser() # local
# dataset_path = Path("/srv/apps/node2vec_git/cvr_data/") # server
# tractate_df, tractate_click_df, tractate_conversion_df = tractate_fe.read_csv_data(dataset_path)
# tractate_df = tractate_fe.tractate_feature_engineering(tractate_df)
# device_df = device_fe.read_csv_data(dataset_path)
# device_df = device_fe.device_feature_engineering(device_df, "tractate")
# # print(device_df.columns)
# # print(device_df.dtypes, "\n")
# cc_df = click_fe.click_feature_engineering(tractate_click_df, tractate_conversion_df)
# df = tractate_fe.join_features(device_df, tractate_df, cc_df)
# # for i in df.columns:
# # print(i)
# # print(df.dtypes)
# train_df, test_df = train_test_split(df, test_size=0.2)
# train_df, val_df = train_test_split(train_df, test_size=0.2)
# all_features = fe.build_features(df, tractate_fe.INT_COLUMNS, tractate_fe.FLOAT_COLUMNS, tractate_fe.CATEGORICAL_COLUMNS,
# tractate_fe.CROSS_COLUMNS)
# params = {"feature_columns": all_features, "hidden_units": [360, 200, 80, 2], "learning_rate": 0.2}
# model_path = str(Path("/data/files/model_tmp/tractate/").expanduser())
# if os.path.exists(model_path):
# shutil.rmtree(model_path)
# session_config = tf.compat.v1.ConfigProto()
# session_config.gpu_options.allow_growth = True
# session_config.gpu_options.per_process_gpu_memory_fraction = 0.7
# # session_config.inter_op_parallelism_threads = 1
# # session_config.intra_op_parallelism_threads = 1
# estimator_config = tf.estimator.RunConfig(session_config=session_config)
# model = tf.estimator.Estimator(model_fn=esmm_model_fn, params=params, model_dir=model_path, config=estimator_config)
# # TODO 50000
# train_spec = tf.estimator.TrainSpec(input_fn=lambda: esmm_input_fn(train_df, shuffle=True), max_steps=12000)
# eval_spec = tf.estimator.EvalSpec(input_fn=lambda: esmm_input_fn(val_df, shuffle=False))
# res = tf.estimator.train_and_evaluate(model, train_spec, eval_spec)
# print("@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@")
# print(res[0])
# ctr_auc = str(res[0]["ctr_auc"])
# ctcvr_auc = str(res[0]["ctcvr_auc"])
# print("ctr_auc: " + ctr_auc)
# print("ctcvr_auc: " + ctcvr_auc)
# print("@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@")
tractate_train_columns = set(tractate_fe.INT_COLUMNS + tractate_fe.FLOAT_COLUMNS + tractate_fe.CATEGORICAL_COLUMNS)
print("features: " + str(len(tractate_train_columns)))
tractate_predict_columns = set(PREDICTION_ALL_COLUMNS)
print(tractate_predict_columns.difference(tractate_train_columns))
print(tractate_train_columns.difference(tractate_predict_columns))
assert tractate_predict_columns == tractate_train_columns
# dataset_path = Path("~/data/cvr_data").expanduser() # local
dataset_path = Path("/srv/apps/node2vec_git/cvr_data/") # server
tractate_df, tractate_click_df, tractate_conversion_df = tractate_fe.read_csv_data(dataset_path)
tractate_df = tractate_fe.tractate_feature_engineering(tractate_df)
device_df = device_fe.read_csv_data(dataset_path)
device_df = device_fe.device_feature_engineering(device_df, "tractate")
# print(device_df.columns)
# print(device_df.dtypes, "\n")
cc_df = click_fe.click_feature_engineering(tractate_click_df, tractate_conversion_df)
df = tractate_fe.join_features(device_df, tractate_df, cc_df)
# for i in df.columns:
# print(i)
# print(df.dtypes)
train_df, test_df = train_test_split(df, test_size=0.2)
train_df, val_df = train_test_split(train_df, test_size=0.2)
all_features = fe.build_features(df, tractate_fe.INT_COLUMNS, tractate_fe.FLOAT_COLUMNS, tractate_fe.CATEGORICAL_COLUMNS,
tractate_fe.CROSS_COLUMNS)
params = {"feature_columns": all_features, "hidden_units": [360, 200, 80, 2], "learning_rate": 0.2}
model_path = str(Path("/data/files/model_tmp/tractate/").expanduser())
if os.path.exists(model_path):
shutil.rmtree(model_path)
session_config = tf.compat.v1.ConfigProto()
session_config.gpu_options.allow_growth = True
session_config.gpu_options.per_process_gpu_memory_fraction = 0.7
# session_config.inter_op_parallelism_threads = 1
# session_config.intra_op_parallelism_threads = 1
estimator_config = tf.estimator.RunConfig(session_config=session_config)
model = tf.estimator.Estimator(model_fn=esmm_model_fn, params=params, model_dir=model_path, config=estimator_config)
train_spec = tf.estimator.TrainSpec(input_fn=lambda: esmm_input_fn(train_df, shuffle=True), max_steps=50000)
eval_spec = tf.estimator.EvalSpec(input_fn=lambda: esmm_input_fn(val_df, shuffle=False))
res = tf.estimator.train_and_evaluate(model, train_spec, eval_spec)
print("@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@")
print(res[0])
ctr_auc = str(res[0]["ctr_auc"])
ctcvr_auc = str(res[0]["ctcvr_auc"])
print("ctr_auc: " + ctr_auc)
print("ctcvr_auc: " + ctcvr_auc)
print("@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@")
total_time = "{:.2f}".format((time.time() - time_begin) / 60)
# model_export_path = str(Path("/data/files/models/tractate/").expanduser())
# save_path = model_export(model, all_features, model_export_path)
# print("save to: " + save_path)
# # TODO save model
# # set_essm_model_save_path("tractate", save_path)
# # record_esmm_auc_to_db("tractate", ctr_auc, ctcvr_auc, total_time, save_path)
# print("============================================================")
model_export_path = str(Path("/data/files/models/tractate/").expanduser())
save_path = model_export(model, all_features, model_export_path)
print("save to: " + save_path)
set_essm_model_save_path("tractate", save_path)
record_esmm_auc_to_db("tractate", ctr_auc, ctcvr_auc, total_time, save_path)
print("============================================================")
# save_path = get_essm_model_save_path("tractate")
# print("load path: " + save_path)
save_path = get_essm_model_save_path("tractate")
print("load path: " + save_path)
# save_path = str(Path("~/data/models/tractate/1598236893").expanduser()) # local
save_path = str(Path("~/data/models/tractate/1598236893").expanduser()) # local
save_path = "/data/files/models/tractate/1599128140" # server
predict_fn = tf.contrib.predictor.from_saved_model(save_path)
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment