cellpin.models.CellPin.train_model

CellPin.train_model(dataset, custom_callbacks=None, train_size=0.8, freeze_pretrained=False, require_pretrained=True, **trainer_kwargs)

Stage-2 main training (both views, full ELBO + invariance + SNN).

Args:

dataset: Training dataset (scAnnDataset). custom_callbacks: Extra PyTorch-Lightning callbacks. train_size: Fraction of data used for training. freeze_pretrained: If True, freeze the full-gene encoder and

decoder (Stage 1 weights) during Stage 2.

require_pretrained: If True (default), raise an error when

freeze_pretrained=True but pretrain_model was never called, preventing silent training against a random frozen decoder.

**trainer_kwargs: Forwarded to CellPinTrainer.

Returns:

:

Fitted CellPinTrainer.

Raises:

RuntimeError: If require_pretrained=True, freeze_pretrained=True,

and pretraining has not been completed.