coco_pipe.decoding.foundation_models.estimators#

Clone-safe sklearn estimators backed by lazily loaded foundation models.

Classes#

`FrozenBackboneTransformer`	Target-independent frozen feature extractor suitable for sklearn pipelines.
`FoundationClassifier`	Lazy trainable foundation-model classifier with grouped validation.

Functions#

clear_frozen_embedding_cache()

Empty the shared frozen-backbone embedding cache.

Module Contents#

coco_pipe.decoding.foundation_models.estimators.clear_frozen_embedding_cache()#

Empty the shared frozen-backbone embedding cache.

Return type:: None

class coco_pipe.decoding.foundation_models.estimators.FrozenBackboneTransformer(model_key, backend='auto', device='auto', pooling='mean', sfreq=None, ch_names=None, cache_embeddings=False, backend_kwargs=None)#

Bases: sklearn.base.BaseEstimator, sklearn.base.TransformerMixin

Target-independent frozen feature extractor suitable for sklearn pipelines.

Parameters:

model_key (str)
backend (str)
device (str)
pooling (str)
sfreq (float | None)
ch_names (list[str] | None)
cache_embeddings (bool)
backend_kwargs (dict[str, Any] | None)

backend_ = None#

prepared_ = None#

backend = 'auto'#

device = 'auto'#

pooling = 'mean'#

sfreq = None#

ch_names = None#

cache_embeddings = False#

backend_kwargs = None#

fit(X, y=None)#

Parameters:

X (numpy.ndarray)
y (numpy.ndarray | None)

transform(X)#

Parameters:: X (numpy.ndarray)
Return type:: numpy.ndarray

class coco_pipe.decoding.foundation_models.estimators.FoundationClassifier(model_key, backend='auto', train_mode='full', device='auto', n_outputs=None, sfreq=None, ch_names=None, trainer=None, lora=None, backend_kwargs=None, checkpoints=None, class_weight='balanced', random_state=42)#

Bases: sklearn.base.BaseEstimator, sklearn.base.ClassifierMixin

Lazy trainable foundation-model classifier with grouped validation.

Parameters:

model_key (str)
backend (str)
train_mode (str)
device (str)
n_outputs (int | None)
sfreq (float | None)
ch_names (list[str] | None)
trainer (dict[str, Any] | None)
lora (dict[str, Any] | None)
backend_kwargs (dict[str, Any] | None)
checkpoints (dict[str, Any] | None)
class_weight (str | dict[Any, float] | None)
random_state (int)

backend_ = None#

prepared_ = None#

checkpoint_path_ = None#

backend = 'auto'#

train_mode = 'full'#

device = 'auto'#

n_outputs = None#

sfreq = None#

ch_names = None#

trainer = None#

lora = None#

backend_kwargs = None#

checkpoints = None#

class_weight = 'balanced'#

random_state = 42#

fit(X, y, groups=None)#

Parameters:

X (numpy.ndarray)
y (numpy.ndarray)
groups (numpy.ndarray | None)

restore_checkpoint(path=None)#

Restore every saved backend component into this fitted estimator.

Parameters:: path (str | pathlib.Path | None)

predict(X)#

Parameters:: X (numpy.ndarray)
Return type:: numpy.ndarray

predict_proba(X)#

Parameters:: X (numpy.ndarray)
Return type:: numpy.ndarray

get_training_history()#

Return type:: list[dict[str, Any]]

get_checkpoint_manifest()#

Return type:: dict[str, Any]

get_model_card_info()#

Return type:: dict[str, Any]

get_failure_diagnostics()#

Return type:: dict[str, Any]

get_artifact_metadata()#

Return type:: dict[str, Any]