Adds a batch of data to be considered for the running top k. The zeroth dimension enumerates the observations. All other dim

True if there is an active tqdm progress loop on the stack.

Applies an ablation recipe of the form [(layer, unit, alpha)...].

Unhooks all hooked layers in the model.

3x3 convolution with padding

3x3 convolution with padding

Makes some images.

Returns two lists: first, a list of tuples [(label, category), ...] where the label and category are human-readable strings indicatin

Lists label and category names.

Given input data as taken from the training distirbution, normalizes every channel to reflect quantile values, uniformly dist

When within a progress loop, post_progress(k=str) will display the given k=str status on the right-hand-side of the progress status bar.

Retains a list of a layers at once.

Retrieve retained data that was previously hooked by retain_layer. Call this after the model is run. If clear is set, then the

In a progressive GAN, the number of channels can change after each upsampling. This function reads the state dict to figure the number o