chore: improve logging

chaine/crf.py

@@ -15,7 +15,7 @@
 
 from chaine._core.crf import Model as _Model
 from chaine._core.crf import Trainer as _Trainer
-from chaine.logging import Logger
+from chaine.logging import Logger, set_level
 from chaine.optimization.spaces import (
     APSearchSpace,
     AROWSearchSpace,
@@ -182,7 +182,7 @@ def train(
         model_filepath : Filepath, optional (default=model.chaine)
             Path to model location.
         """
-        LOGGER.info("Loading training data")
+        LOGGER.info("Start training")
         for i, (sequence, labels_) in enumerate(zip(dataset, labels)):
             if not is_valid_sequence(sequence):
                 raise ValueError(f"Invalid format: {sequence}")
@@ -277,6 +277,9 @@ def optimize_hyperparameters(
         list[dict[str, dict]]
             Sorted list of hyperparameters and evaluation scores.
         """
+        # make logging less verbose
+        set_level("chaine._core.crf", "ERROR")
+
         # set random seed
         random.seed(self.seed)
 
@@ -309,6 +312,9 @@ def optimize_hyperparameters(
         LOGGER.info(f"Best baseline: {baselines[0][f'mean_{self.metric}']}")
         LOGGER.info(f"Best optimized model: {results[0]['stats'][f'mean_{self.metric}']}")
 
+        # make logging verbose again
+        set_level("chaine._core.crf", "INFO")
+
         return results
 
     def _metric(self, trial: dict[str, dict]) -> float:

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "chaine"
-version = "2.0.0a3"
+version = "2.0.0a4"
 description = "Linear-chain conditional random fields for natural language processing"
 authors = ["Severin Simmler <[email protected]>"]
 readme = "README.md"