CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-08 10:07:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

graph : simplify attention api

Browse Source 
ggml-ci
This commit is contained in:
Georgi Gerganov
2025-02-19 18:43:49 +02:00
parent e17e4b72d1
commit 2eacb4c1bf
4 changed files with 47 additions and 75 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
src/llama-model.cpp
View File

@@ -4258,13 +4258,7 @@ struct llm_build_context {
ggml_build_forward_expand(gf, k_cur);
ggml_build_forward_expand(gf, v_cur);
//build_kv_store(gf, k_cur, v_cur, il);
lgf->build_attn_kv_store(ctx0, gf, k_cur, v_cur, n_tokens, il, worst_case);
struct ggml_tensor * cur;
//cur = build_kqv(gf, wo, wo_b, q_cur, kq_mask, kq_scale, il);
cur = lgf->build_attn_qkv(ctx0, gf, wo, wo_b, q_cur, n_tokens, kq_scale, il, worst_case);
ggml_tensor * cur = lgf->build_attn(ctx0, gf, wo, wo_b, k_cur, v_cur, q_cur, n_tokens, kq_scale, il, worst_case);
cb(cur, "kqv_out", il);
return cur;