Files
ik_llama.cpp/github-data/pull_requests/431 - Forgotten MMQ ref and typo.md
2025-07-23 13:31:53 +02:00

1.3 KiB

🔀 #431 - Forgotten MMQ ref and typo

Author Nexesenex
State Closed
Created 2025-05-18
Updated 2025-05-22

Description


💬 Conversation

👤 ikawrakow submitted a review the 2025-05-18 at 14:36:30: APPROVED

Hey, you are back!


👤 Nexesenex commented the 2025-05-18 at 14:48:44:

Hey! Yeah, you sounded the horn with those MMQ Kernels for the IQ_K quants, I waited for them for a long time. I merge your IQ quants (included the KS ones with success last year, before the rev 14 of the GGUF format broke compatibility with them, possibly due to the template change introduced in https://github.com/ikawrakow/ik_llama.cpp/pull/45 ) Meanwhile, I was amusing myself merging models, among other nerdy delights. Congrats for all the amazing developments you made, even if it's hard for me to swing between mainline and IK_Llama to feed my Croco. Also, Turboderp switched on QTIP based quants for Exllamav3. Things are getting exciting!