deepfold.losses.confidence.compute_predicted_aligned_error

deepfold.losses.confidence.compute_predicted_aligned_error(logits: Tensor, max_bin: int = 31, num_bins: int = 64) → Dict[str, Tensor]

Computes aligned confidence metrics from logits.

Parameters:

• logits – [*, num_res, num_res, num_bins] the logits output from PredictedAlignedErrorHead.

• max_bin – Maximum bin value

• num_bins – Number of bins

Returns:

[*, num_res, num_res, num_bins] the predicted

aligned error probabilities over bins for each residue pair.

predicted_aligned_error: [*, num_res, num_res] the expected aligned distance

error for each pair of residues.

max_predicted_aligned_error: [*] the maximum predicted error possible.

Return type:

aligned_confidence_probs