uniport.data_loader.load_data

uniport.data_loader.load_data(adatas, mode='h', use_rep=['X', 'X'], num_cell=None, max_gene=None, adata_cm=None, use_specific=False, domain_name='domain_id', batch_size=256, drop_last=True, shuffle=True, num_workers=4)[source]

Load data for training.

Parameters:

  • adatas – A list of AnnData matrice.

  • mode – training mode. Choose between [‘h’, ‘d’, ‘v’].

  • use_rep – use ‘.X’ or ‘.obsm’.

  • num_cell – numbers of cells of each adata in adatas.

  • max_gene – maximum number of genes of each adata in adatas.

  • adata_cm – adata with common genes of adatas.

  • use_specific – use dataset-specific genes.

  • domain_name – domain name of each adata in adatas.

  • batch_size – size of each mini batch for training.

  • drop_last – drop the last samples that not up to one batch.

  • shuffle – shuffle the data

  • num_workers – number parallel load processes according to cpu cores.

Returns:

  • trainloader – data loader for training

  • testloader – data loader for testing