tech.compute.tensor.dimensions.shape — techascent/tech.compute 3.0

A shape vector entry can be a number of things. We want to be precise with handling them and abstract that handling so new things have a clear path.

->2d clj

(->2d shape)

Given dimensions, return new dimensions with the lowest (fastest-changing) dimension unchanged and the rest of the dimensions multiplied into the higher dimension.

->batch clj

(->batch shape)

Given dimensions, return new dimensions with the lowest (fastest-changing) dimension unchanged and the rest of the dimensions multiplied into the higher dimension.

classified-sequence->count clj

(classified-sequence->count {:keys [type min-item max-item sequence]})

classified-sequence->elem-idx clj

(classified-sequence->elem-idx {:keys [type min-item max-item sequence] :as dim}
                               shape-idx)

classified-sequence->sequence clj

(classified-sequence->sequence {:keys [min-item max-item type sequence]})

classified-sequence-keys clj

classified-sequence? clj

(classified-sequence? item)

classify-sequence clj

(classify-sequence item-seq)

combine-classified-sequences clj

(combine-classified-sequences source-sequence select-sequence)

Room for optimization here. But simplest way is easiest to get correct.

direct-shape? clj

(direct-shape? shape)

ecount clj

(ecount shape)

Return the element count indicated by the dimension map

indirect-shape? clj

(indirect-shape? shape)

monotonic-operators clj

reverse-classified-sequence clj

(reverse-classified-sequence {:keys [type sequence] :as item})

reverse-shape clj

(reverse-shape shape-vec)

shape->count-vec clj

(shape->count-vec shape-vec)

shape-entry->count clj

(shape-entry->count shape-entry)

Return a vector of counts of each shape.