[KongchangAI]
Concept

LLM-Surgeon

An optimization method for the curvature-aware pruning objective. It addresses the same problem as CACTUS but is less efficient to solve, and CACTUS substantially surpasses it in speed.

Timeline (last 90 days)

Oct 14

Under clearly weaker hardware conditions, CACTUS is still more than 28 times faster than LLM-Surgeon.

Unverified50%

All Facts (1)

Source Articles