tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-07-19 17:17:40 +00:00
Code Issues Packages Projects Releases Wiki Activity

graph : update attn/kv_self names

Browse Source
This commit is contained in:
Georgi Gerganov
2025-02-14 17:22:55 +02:00
parent 828064564c
commit 1d801d27b9
4 changed files with 17 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
src/llama-graph.h
View File

@ -92,17 +92,17 @@ public:
int il,
bool worst_case) = 0;
virtual ggml_tensor * build_soft_max_ext(
virtual ggml_tensor * build_attn_soft_max(
ggml_context * ctx0,
ggml_tensor * kq,
float kq_scale) = 0;
virtual void build_k_shift(
virtual void build_kv_self_shift(
ggml_context * ctx0,
ggml_cgraph * graph) = 0;
// find holes from the beginning of the KV cache and fill them by moving data from the end of the cache
virtual void build_defrag(
virtual void build_kv_self_defrag(
ggml_context * ctx0,
ggml_cgraph * graph) = 0;