Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

kv-cache : support attention rotation for heterogeneous iSWA (#21513)

Browse Source 
* kv-cache : support attention rotation for heterogeneous iSWA

* cont : remove assert
This commit is contained in:
Georgi Gerganov
2026-04-07 20:31:28 +03:00
committed by GitHub
parent 957d717ce5
commit 4eb19514dd
4 changed files with 58 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-graph.h
+4 -2
View File
@@ -308,7 +308,7 @@ public:
ggml_tensor * self_kq_mask = nullptr; // F32 [n_kv, n_batch/n_stream, 1, n_stream]
ggml_tensor * self_kq_mask_cnv = nullptr; // [n_kv, n_batch/n_stream, 1, n_stream]
// note: assumes v_rot^ == I
// note: assumes v_rot^2 == I
ggml_tensor * self_k_rot = nullptr;
ggml_tensor * self_v_rot = nullptr;
@@ -388,10 +388,12 @@ public:
ggml_tensor * self_kq_mask_swa = nullptr; // F32 [n_kv, n_batch/n_stream, 1, n_stream]
ggml_tensor * self_kq_mask_swa_cnv = nullptr; // [n_kv, n_batch/n_stream, 1, n_stream]
// note: using same rotation matrices for both base and swa cache
ggml_tensor * self_k_rot = nullptr;
ggml_tensor * self_v_rot = nullptr;
ggml_tensor * self_k_rot_swa = nullptr;
ggml_tensor * self_v_rot_swa = nullptr;
const llama_hparams hparams;
const llama_cparams cparams;