kvcache-ai/ktransformers

Suggestion to add DeepSeek v2.5 support

arisau opened this issue · 4 comments

Recently DeepSeek 2.5 MoE model was released. Please consider adding a support for it. (Q5_K_M or Q4_K_M)

It appears to be supported automatically, as some users have already tested it successfully.

Perfect. Thank you.
just to confirm, only Q4_K_M works?

We have supported all Qx_km format~

Understood. Thank you once again.