AI Inference & Serving
Unstructured Pruning
Removing or zeroing selected individual connections without necessarily removing whole channels or changing tensor dimensions. The result can be sparse while retaining the original dense shape. Exploiting that sparsity requires an appropriate representation and execution path; merely writing zeros into a dense tensor does not guarantee smaller storage or faster inference.
Reviewed
Sources
Member lesson
The definition and sources are public. The complete practical lesson is for members.