MC Dropout API¶

deepuq.methods.mc_dropout ¶

Monte-Carlo dropout inference wrappers.

Bases: Module

Wrap any model with dropout to perform MC Dropout at inference.

Args: model: torch.nn.Module with Dropout layers n_mc: number of stochastic forward passes apply_softmax: whether to convert logits to probabilities

predict(x: Tensor)

Run stochastic dropout passes and return predictive mean/variance.

predict_uq(x: Tensor) -> UQResult

Return standardized uncertainty fields.