SPACEL.Spoint.base_model.SpointModel.train

SpointModel.train(max_steps=5000, save_mode='all', save_path=None, prefix=None, sm_step=10, st_step=10, test_step_gap=1, convergence=0.001, early_stop=False, early_stop_max=2000, sm_lr=None, st_lr=None, batch_size=1024, rec_w=1, infer_w=1, m_w=1, scvi_max_epochs=100, scvi_early_stopping=True, scvi_batch_size=4096)

Training Spoint model.

Obtain latent feature from scVI then feed in Spoint model for training.

Parameters:

• max_steps – The max step of training. The training process will be stop when achive max step.

• save_mode – A string determinates how the model is saved. It must be one of ‘best’ and ‘all’.

• save_path – A string representing the path directory where the model is saved.

• prefix – A string added to the prefix of file name of saved model.

• convergence – The threshold of early stop.

• early_stop – If True, turn on early stop.

• early_stop_max – The max steps of loss difference less than convergence.

• sm_lr – Learning rate for simulated data.

• st_lr – Learning rate for spatial transcriptomic data.

• batch_size – Batch size of the data be feeded in model once.

• rec_w – The weight of reconstruction loss.

• infer_w – The weig ht of inference loss.

• m_w – The weight of MMD loss.

• scvi_max_epochs – The max epoch of scVI.

• scvi_batch_size – The batch size of scVI.

Returns:

None