Merge pull request #56 from dayyass/develop

release v0.1.2

setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = text-classification-baseline
-version = 0.1.1
+version = 0.1.2
 author = Dani El-Ayyass
 author_email = [email protected]
 description = TF-IDF + LogReg baseline for text classification

tests/test_pipeline.py

@@ -2,8 +2,8 @@
 import unittest
 
 from data.load_20newsgroups import load_20newsgroups
+from text_clf.__main__ import train
 from text_clf.config import load_default_config
-from text_clf.train import train
 
 
 class TestUsage(unittest.TestCase):

text_clf/__init__.py

@@ -1,4 +1,4 @@
-from .train import train
+from .__main__ import train
 
-__version__ = "0.1.1"
+__version__ = "0.1.2"
 __all__ = ["train"]

text_clf/__main__.py

@@ -1,5 +1,32 @@
-from .train import train
-from .utils import get_argparse
+import traceback
+
+from .config import get_config
+from .train import _train
+from .utils import close_logger, get_argparse, get_logger
+
+
+def train(path_to_config: str = "config.yaml") -> None:
+    """
+    Function to train baseline model with exception handler.
+
+    :param str path_to_config: path to config.
+    """
+
+    # load config
+    config = get_config(path_to_config=path_to_config)
+
+    # get logger
+    logger = get_logger(path_to_logfile=config["path_to_save_logfile"])
+
+    try:
+        _train(
+            config=config,
+            logger=logger,
+        )
+    except:  # noqa
+        close_logger(logger)
+
+        print(traceback.format_exc())
 
 
 def main() -> int:

text_clf/config.py

@@ -29,6 +29,9 @@ def get_config(path_to_config: str) -> Dict[str, Any]:
         / f"model_{datetime.datetime.now().strftime('%Y-%m-%d %H-%M-%S')}"
     )
 
+    # mkdir if not exists
+    config["path_to_save_folder"].absolute().mkdir(parents=True, exist_ok=True)
+
     config["path_to_config"] = path_to_config
     config["path_to_save_model"] = config["path_to_save_folder"] / "model.joblib"
     config["path_to_save_logfile"] = config["path_to_save_folder"] / "logging.txt"

text_clf/train.py

@@ -1,35 +1,31 @@
+import logging
 import time
+from typing import Any, Dict
 
 from sklearn.feature_extraction.text import TfidfVectorizer
 from sklearn.linear_model import LogisticRegression
 from sklearn.metrics import classification_report
 from sklearn.pipeline import Pipeline
 from sklearn.preprocessing import LabelEncoder
 
-from .config import get_config
 from .data import load_data
 from .save import save_model
-from .utils import get_logger, set_seed
+from .utils import close_logger, set_seed
 
 
-def train(path_to_config: str = "config.yaml") -> None:
+def _train(
+    config: Dict[str, Any],
+    logger: logging.Logger,
+) -> None:
     """
     Function to train baseline model.
 
-    :param str path_to_config: path to config.
+    :param Dict[str, Any] config: config.
+    :param logging.Logger logger: logger.
     """
 
-    # load config
-    config = get_config(path_to_config)
-
-    # mkdir if not exists
-    config["path_to_save_folder"].absolute().mkdir(parents=True, exist_ok=True)
-
-    # get logger
-    logger = get_logger(config["path_to_save_logfile"])
-
     # log config
-    with open(path_to_config, mode="r") as fp:
+    with open(config["path_to_config"], mode="r") as fp:
         logger.info(f"Config:\n\n{fp.read()}")
 
     # reproducibility
@@ -109,3 +105,5 @@ def train(path_to_config: str = "config.yaml") -> None:
     )
 
     logger.info("Done!")
+
+    close_logger(logger)

text_clf/utils.py

@@ -59,6 +59,19 @@ def get_logger(path_to_logfile: str) -> logging.Logger:
     return logger
 
 
+def close_logger(logger: logging.Logger) -> None:
+    """
+    Close logger.
+    Source: https://stackoverflow.com/questions/15435652/python-does-not-release-filehandles-to-logfile
+
+    :param logging.Logger logger: logger.
+    """
+
+    for handler in logger.handlers[:]:
+        handler.close()
+        logger.removeHandler(handler)
+
+
 def set_seed(seed: int) -> None:
     """
     Set seed for reproducibility.