ladder.scripts.metrics.gen_profile_reproduction

ladder.scripts.metrics.gen_profile_reproduction(point_dataset, model, source=None, target=None, n_trials=10, lib_size=10000.0, verbose=False, use_cuda=False)

Used to generate new points by performing a full pass through the models.

The points and their normalized profiles are logged for each generation, up to the number of n_trials. If source and target are provided, performs a transfer. Otherwise runs for reconstruction.

Parameters:

• point_dataset – The Dataset object to be provided.

• model (Predictive) – The model generator function wrapped in Predictive.

• source (Tensor, optional) – Optional source condition encoding tensor to subset the provided TensorDataset input.

• target (Tensor, optional) – Optional target conditon encoding tensor to subset the provided TensorDataset input.

• n_trials (int, default: 10) – Number of times to repeat the generative process. Must be positive.

• lib_size (float, default: 1e4) – Library size for normalized profiles.

• verbose (bool, default: False) – If True, prints how many iterations are left for n_trials.

• use_cuda (bool, default: False) – If True, attempts to use CUDA device for the generative process.

Returns:

profilesTensor

Normalized pseudo-bulk profiles for the points generated.

predsTensor

Points generated by the model.