torchdms.evaluation

Evaluating models.

Functions

build_evaluation_dict	Evaluate & organize all testing data paired with metadata.
complete_error_summary
error_df_of_evaluation_dict	Build a dataframe that describes the error per test point.
error_summary_of_data
error_summary_of_error_df	Build a dataframe summarizing error.

torchdms.evaluation.build_evaluation_dict(model, test_data, device='cpu')

Evaluate & organize all testing data paired with metadata.

A function which takes a trained model, matching test dataset (BinaryMapDataset w/ the same input dimensions.) and return a dictionary containing the

• samples: binary encodings numpy array shape (num samples, num possible mutations)

• predictions and targets: both numpy arrays of shape (num samples, num targets)

This should have everything mostly needed to do plotting about testing data (not things like loss or latent space prediction).

torchdms.evaluation.error_df_of_evaluation_dict(evaluation_dict)

Build a dataframe that describes the error per test point.

torchdms.evaluation.error_summary_of_error_df(error_df, model)

Build a dataframe summarizing error.