def get_wrapped_sklearn_model(model_cls_str):
# model_cls_str は'^Sklearn.+$' にマッチする前提
model_cls_str = re.sub('Sklearn', '', model_cls_str)
model_cls_mod_notreg = _valid_not_regressor_models.get(model_cls_str)
model_cls_mod_reg = _valid_regressor_models.get(model_cls_str)
if model_cls_mod_notreg is None and model_cls_mod_reg is None:
raise Exception('{} is invalid.'.format(model_cls_str))
if model_cls_mod_notreg is not None and model_cls_mod_reg is not None:
raise Exception('unexpected state.')
model_cls_mod = model_cls_mod_notreg or model_cls_mod_reg
model_cls = load_object_by_str('{}@{}'.format(model_cls_str,
model_cls_mod))
cls_attrs = {
'fit':
_fit,
'reset':
_generate_reset_func(model_cls),
'dump':
dump_sklearn_model,
'load':
load_sklearn_model,
'base_init_finished':
_base_init_finished
if model_cls_mod_notreg else _base_init_finished_reg,
}
if hasattr(model_cls, 'predict'):
cls_attrs['predict'] = _predict
if hasattr(model_cls, 'predict_proba'):
cls_attrs['predict_proba'] = _predict_proba
logger.debug('model_cls_str: {} .. created.'.format(model_cls_str))
return type('WrappedSklearn{}'.format(model_cls_str), (WrappedModel, ),
cls_attrs)
def _get_preprocessor_strict(config):
name = config.get('name')
if name is None:
raise Exception('preprocessor_config.name must be set.')
cameledname = snake2camel(name)
ppcls = load_object_by_str(_preprocessor_name_alias.get(cameledname, name))
return ppcls(**config.get('kwargs', {}))
def test_model_regressor(pd_assert_equal):
init_kwargs_settings = [
{},
{
'normalize': True
},
]
fit_kwargs_settings = [{}]
evaluate_kwargs_settings = [{
'train_test_split_func': '[email protected]_selection',
'train_test_split_func_kwargs': {
'random_state': 0
},
'metrics': 'all',
}]
dataset_config = {
'loader_config': {
'name': '*****@*****.**',
'kwargs': {
'random_state': 0,
},
},
'target_column': 'target',
}
preprocessor = get_preprocessor({
'name': 'identify',
'kwargs': {},
})
ffunc_for_predictor = ffunc_for_target = load_object_by_str(
'*****@*****.**')
ds1 = get_dataset(dataset_config)
X1, y1 = ds1.get_predictor_target()
for init_kwargs in init_kwargs_settings:
for fit_kwargs in fit_kwargs_settings:
for evaluate_kwargs in evaluate_kwargs_settings:
mconfig = {
'name': 'SklearnLinearRegression',
'init_kwargs': init_kwargs,
'fit_kwargs': fit_kwargs,
'evaluate_kwargs': evaluate_kwargs,
'is_rebuild': False,
}
m = get_model(mconfig)
morigin = LinearRegression(**init_kwargs)
m.fit(X1, y1)
morigin.fit(X1, y1, **fit_kwargs)
# assertion
pd_assert_equal(
pd.Series(m.predict(X1)).astype('float64'),
pd.Series(morigin.predict(X1)).astype('float64'))
X_train, X_test, y_train, y_test = skl_train_test_split(
X1, y1, **evaluate_kwargs['train_test_split_func_kwargs'])
rev1 = m.evaluate(X1, y1, preprocessor, ffunc_for_predictor,
ffunc_for_target)
if not rev1['cv']:
met = rev1['metrics'][0]
mean_absolute_error = [
o['value'] for o in met
if o['name'] == 'mean_absolute_error'
][0]
mean_squared_error = [
o['value'] for o in met
if o['name'] == 'mean_squared_error'
][0]
median_absolute_error = [
o['value'] for o in met
if o['name'] == 'median_absolute_error'
][0]
r2_score = [
o['value'] for o in met if o['name'] == 'r2_score'
][0]
explained_variance = [
o['value'] for o in met
if o['name'] == 'explained_variance'
][0]
morigin.fit(ffunc_for_predictor(X_train),
ffunc_for_target(y_train))
assert math.fabs(
mean_absolute_error - skl_metrics.mean_absolute_error(
y_test, morigin.predict(X_test))) < 0.00001
assert math.fabs(
mean_squared_error - skl_metrics.mean_squared_error(
y_test, morigin.predict(X_test))) < 0.00001
assert math.fabs(
median_absolute_error -
skl_metrics.median_absolute_error(
y_test, morigin.predict(X_test))) < 0.00001
assert math.fabs(r2_score - skl_metrics.r2_score(
y_test, morigin.predict(X_test))) < 0.00001
assert math.fabs(
explained_variance -
skl_metrics.explained_variance_score(
y_test, morigin.predict(X_test))) < 0.00001
def train(train_id,
scenario_tag,
dataset_config=None,
model_config=None,
preprocessor_config=None,
formatter_config_for_predictor=None,
formatter_config_for_target=None,
evaluate_enabled=False,
fit_model_enabled=False,
dump_result_enabled=False):
"""
モデルの訓練を実行する関数
:param train_id: シナリオ中における訓練実行の識別子
:type train_id: str
:param scenario_tag: シナリオに付与されるタグ
:type scenario_tag: str
:param dataset_config: Datasetの設定。:class:`akebono.dataset.get_dataset` の引数。
:type dataset_config: dict
:param model_config: Modelの設定。:class:`akebono.model.get_model` の引数。
:type model_config: dict
:param preprocessor_config: Preprocessorの設定。:class:`akebono.preprocessor.get_preprocessor` の引数。
:type preprocessor_config: dict
:param formatter_config_for_predictor: 特徴用Formatterの設定。
:type formatter_config_for_predictor: dict
:param formatter_config_for_target: 目標用Formatterの設定。
:type formatter_config_for_target: dict
:param evaluate_enabled: モデルの評価を実行するかのフラグ
:type evaluate_enabled: bool
:param fit_model_enabled: モデルの訓練を実行するかのフラグ
:type fit_model_enabled: bool
:param dump_result_enabled: モデル、評価結果の永続化を実行するかのフラグ
:type dump_result_enabled: bool
"""
if model_config is None:
raise ValueError('model_config must be set.')
if dataset_config is None:
raise ValueError('dataset_config must be set.')
if preprocessor_config is None:
preprocessor_config = {
'name': 'identify',
'kwargs': {},
}
if formatter_config_for_predictor is None:
formatter_config_for_predictor = {
'name': '*****@*****.**',
}
if formatter_config_for_target is None:
formatter_config_for_target = {
'name': '*****@*****.**',
}
ret = {
'type': 'train',
'id': train_id,
'dataset_config': dataset_config,
'model_config': model_config,
'preprocessor_config': preprocessor_config,
'formatter_config_for_predictor': formatter_config_for_predictor,
'formatter_config_for_target': formatter_config_for_target,
'evaluate_enabled': evaluate_enabled,
'fit_model_enabled': fit_model_enabled,
'dump_result_enabled': dump_result_enabled
}
dataset = get_dataset(dataset_config)
preprocessor = get_preprocessor(preprocessor_config)
preprocessor.set_operation_mode('train')
logger.debug('load dataset start.')
X, y = dataset.get_predictor_target()
if X.index.size == 0:
raise EmptyDatasetError('empty record')
logger.debug('load dataset done.')
model_config['is_rebuild'] = False
model = get_model(model_config)
format_func_for_predictor = load_object_by_str(
formatter_config_for_predictor['name'])
format_func_for_target = load_object_by_str(
formatter_config_for_target['name'])
if evaluate_enabled:
logger.debug('evaluate start.')
rep = model.evaluate(X, y, preprocessor, format_func_for_predictor,
format_func_for_target)
gc.collect()
logger.debug('evaluate done.')
ret['evaluate'] = rep
if fit_model_enabled:
logger.debug('fit start.')
fX_p, _ = preprocessor.process(X, None)
fX = format_func_for_predictor(fX_p)
model.fit(fX, format_func_for_target(y))
gc.collect()
logger.debug('fit done.')
ret['model'] = model
if dump_result_enabled:
logger.debug('dump_train_result start.')
ret['preprocessor'] = preprocessor
dump_train_result(train_id, scenario_tag, ret)
logger.debug('dump_train_result done.')
return ret
def evaluate(model,
X,
y,
preprocessor,
format_func_for_predictor,
format_func_for_target,
train_test_split_func='[email protected]_selection',
train_test_split_func_kwargs={},
cross_val_iterator=None,
cross_val_iterator_kwargs={},
metrics='all'):
if model.model_type is None:
model.set_model_type(y=y)
model_type = model.model_type
# cross_val_iteratorはNone or iterable
if cross_val_iterator is not None:
if isinstance(cross_val_iterator, str):
cross_val_iterator = load_object_by_str(cross_val_iterator)
else:
# train_test_split_func はNone or string or function
if isinstance(train_test_split_func, str):
train_test_split_func = load_object_by_str(train_test_split_func)
result = {'metrics': []}
train_test_iterator = None
# not CV mode
if cross_val_iterator is None:
train_test_iterator = [
train_test_split_func(X, y, **train_test_split_func_kwargs)
]
result['cv'] = False
# CV mode
else:
train_test_iterator = _sklearn_cross_val_iter2train_test_iter(
X, y, cross_val_iterator, cross_val_iterator_kwargs)
result['cv'] = True
for X_train_raw, X_test_raw, y_train_raw, y_test_raw in train_test_iterator:
y_train, y_test = format_func_for_target(
y_train_raw), format_func_for_target(y_test_raw)
preprocessor.reset()
X_train_p, X_test_p = preprocessor.process(X_train_raw, X_test_raw)
X_train, X_test = format_func_for_predictor(
X_train_p), format_func_for_predictor(X_test_p)
one_result = []
model, y_pred, y_pred_proba = _fit_and_predict(X_train, X_test,
y_train, model)
if metrics == 'all':
_preload_metrics = None
if model_type == 'binary_classifier':
_preload_metrics = _binary_classifier_metrics
elif model_type == 'regressor':
_preload_metrics = _regressor_metrics
elif model_type == 'multiple_classifier':
_preload_metrics = _multiple_classifier_metrics
else:
raise Exception('not supported.')
for m in _preload_metrics:
one_result.append(
_get_evaluated_result(y_pred, y_pred_proba, y_test, m))
else:
raise Exception('not supported.')
result['metrics'].append(one_result)
return result
def get_dataset(dataset_config):
"""
Datasetを生成するための関数
:param dataset_config: Datasetについての設定
:type dataset_config: dict
:return: :class:`Dataset` object
Usage:
>>> from akebono.dataset import get_dataset
>>> dataset_config = {
'loader_config': {
'name': '*****@*****.**',
'kwargs': {
'n_features': 1,
'noise': 30.0,
'random_state': 0,
},
},
'target_column': 'target',
'cache_enabled': False,
}
>>> ds = get_dataset(dataset_config)
>>> ds
<akebono.dataset.model.Dataset object at 0x11291acc0>
"""
dataset_name = dataset_config.get('name')
target_column = dataset_config.get('target_column', 'target')
cache_enabled = dataset_config.get('cache_enabled', False)
evacuated_columns = dataset_config.get('evacuated_columns', [])
if not isinstance(evacuated_columns, list):
raise TypeError('evacuated_columns must be list.')
loader_config = dataset_config.get('loader_config')
if not isinstance(loader_config, dict):
raise Exception(
'loader_config must be specified and this type is dict.')
load_func = loader_config.get('name')
load_func = _loader_name_alias.get(load_func, load_func) # エイリアスがあったらそれを使う
if load_func is None:
raise Exception('loader_config.name must be specified.')
load_func = load_object_by_str(load_func)
load_func_kwargs = Param(loader_config.get('kwargs', {}))
loader_param = loader_config.get('param', {})
_reserved_params = (
'dataset_name',
'target_column',
)
for rp in _reserved_params:
if rp in loader_param:
raise KeyError('{} is reserved param.'.format(rp))
loader_param['dataset_name'] = dataset_name
loader_param['target_column'] = target_column
preprocess_func_str = dataset_config.get(
'preprocess_func', '*****@*****.**')
preprocess_func_hash = get_hash(preprocess_func_str)
preprocess_func = load_object_by_str(preprocess_func_str)
preprocess_func_kwargs = Param(
dataset_config.get('preprocess_func_kwargs', {}))
def _core_func():
return preprocess_func(
load_func(copy.deepcopy(load_func_kwargs.value), loader_param),
**copy.copy(preprocess_func_kwargs.value))
fname = '{}_{}_{}_{}'.format(
dataset_name, load_func_kwargs.get_hashed_id(length=24),
preprocess_func_hash[:24],
preprocess_func_kwargs.get_hashed_id(length=24))
dataset_loading_cache_enabled = dataset_config.get(
'dataset_loading_cache_enabled', True)
if dataset_loading_cache_enabled:
ds = datasetholder.get(fname)
if ds is not None:
logger.debug(
'dataset_loading_cache enabled .. {} get done.'.format(
ds.name))
return ds
pkl_fname = fname + '.pkl'
if cache_enabled:
if dataset_name is not None:
logger.info('dataset cache enabled')
_core_func = cache_located_at(
pathjoin(settings.cache_dir, pkl_fname))(_core_func)
else:
raise Exception(
'dataset_config.cache_enabled is True, but dataset_config.name is None'
)
ds = Dataset(fname, _core_func(), target_column, evacuated_columns)
if dataset_loading_cache_enabled:
datasetholder.set(ds)
logger.debug('dataset_loading_cache enabled .. {} set done.'.format(
ds.name))
return ds