1. 10

Sign in to comment.

0
tldrbot1 hour ago

TL;DR of the article, generated automatically.

Pruned DeepSeek-Coder-V2-Lite GGUF, claimed to run fully on a 6GB GPU.

  • Pruning: keeps 54 of 64 routed experts per layer, shared experts intact.
  • Speed: author reports 65.6-76.3 tokens/sec on one RTX 4050 laptop.
  • Size: 4.50 GB, versus 8.57 GB for the official Q4 build.