thek0tyara/llama-cpp-turboquant
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 4

llama: dynamic head_dim and n_rot for SWA (#20301)

Browse source 
* llama: dynamic head_dim and n_rot for SWA

* also add gguf_writer wrappers

* fix build

* build_rope_shift arg reorder
This commit is contained in:
Xuan-Son Nguyen 2026-03-09 22:22:39 +01:00 • committed by GitHub
parent 23fbfcb1ad
commit 59db9a357d
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
112 changed files with 419 additions and 346 deletions
Download patch file Download diff file Expand all files Collapse all files

2
src/models/deepseek2.cpp
View file

@ -8,7 +8,7 @@ llm_build_deepseek2::llm_build_deepseek2(const llama_model & model, const llm_gr
const int64_t n_embd_head_k = hparams.n_embd_head_k_mla();
const int64_t n_embd_head_v = hparams.n_embd_head_v_mla();
const int64_t n_embd_head_qk_rope = hparams.n_rot;
const int64_t n_embd_head_qk_rope = hparams.n_rot();
const int64_t n_embd_head_qk_nope = n_embd_head_k - n_embd_head_qk_rope;
const uint32_t kv_lora_rank = hparams.n_lora_kv;