рефакторинг блокнота research в части логирования в mlflow

iis_project/sklearn_utils/__init__.py

@@ -0,0 +1,60 @@
+from collections.abc import Container, Sequence, Mapping
+from typing import TypeAlias, TypeVar
+
+
+ParamsFilterSpec: TypeAlias = (
+    bool
+    | Container[str]
+    | tuple[bool, Container[str]]
+    | Mapping[str, 'ParamsFilterSpec']
+    | tuple[bool, 'ParamsFilterSpec']
+)
+
+
+V = TypeVar('V')
+
+
+def _split_param_key(key: str) -> tuple[str, ...]:
+    return tuple(key.split('__'))
+
+
+def _match_key_to_filter_spec(
+    key: Sequence[str], spec: ParamsFilterSpec, empty_default: bool,
+) -> bool:
+    if isinstance(spec, Sequence) and (len(spec) == 2) and isinstance(spec[0], bool):
+        if (len(key) == 0) and (not spec[0]):
+            return empty_default
+        spec = spec[1]
+    if isinstance(spec, Mapping):
+        if len(key) == 0:
+            return empty_default
+        spec_nested = spec.get(key[0])
+        if spec_nested is None:
+            return False
+        return _whether_to_include_param(key[1:], spec_nested)
+    elif isinstance(spec, Container):
+        if len(key) == 0:
+            return True
+        return (key[0] in spec)
+    return bool(spec)
+
+
+def _whether_to_include_param(
+    key: Sequence[str], include: ParamsFilterSpec = True, exclude: ParamsFilterSpec = False,
+) -> bool:
+    return (
+        (not _match_key_to_filter_spec(key, exclude, empty_default=False))
+        and _match_key_to_filter_spec(key, include, empty_default=True)
+    )
+
+
+def filter_params(
+    params: Mapping[str, V],
+    include: ParamsFilterSpec = True,
+    exclude: ParamsFilterSpec = False,
+) -> Mapping[str, V]:
+    return {
+        k: v
+        for k, v in params.items()
+        if _whether_to_include_param(_split_param_key(k), include, exclude)
+    }

iis_project/sklearn_utils/compose.py

@@ -0,0 +1,3 @@
+COLUMN_TRANSFORMER_PARAMS_COMMON_INCLUDE = [
+    'remainder', 'sparse_threshold', 'transformer_weights',
+]

iis_project/sklearn_utils/ensemble.py

@@ -0,0 +1 @@
+RANDOM_FOREST_REGRESSOR_PARAMS_COMMON_EXCLUDE = ['n_jobs', 'verbose', 'warm_start']

iis_project/sklearn_utils/pipeline.py

@@ -0,0 +1 @@
+PIPELINE_PARAMS_COMMON_INCLUDE = ['transform_input']

iis_project/sklearn_utils/preprocessing.py

@@ -0,0 +1 @@
+STANDARD_SCALER_PARAMS_COMMON_EXCLUDE = ['copy']