[Join request] vivekkalyanarangan

#34
by vivekkalyanarangan - opened

Hi, I'd like to join blog-explorers to publish a technical article on quantizing Gemma 4 E4B's per-layer embeddings: KL divergence by bit width and language, decode speed, and how llama.cpp keeps the 5.6 GB table on the host and pages rows on demand. I'm writing a book on model quantization. Thanks!

Sign up or log in to comment