DarthAffe/stable-diffusion.cpp
1
0 Fork 0
mirror of https://github.com/leejet/stable-diffusion.cpp.git synced 2026-11-11 11:26:34 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat: add configurable Qwen cache types and early cache scheduling (#2045)

Browse Source
This commit is contained in:
leejet authored and GitHub committed 2026-09-25 01:51:01 +08:00
1 parent 4dfe8f5d45
commit 740c7ae193
11 files changed
+185 -61
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

src/core/ggml_extend.cpp
+8 -1
View File
@@ -623,7 +623,11 @@ ggml_tensor* ggml_ext_attention_ext(ggml_context* ctx,
bool skip_reshape,
bool flash_attn,
float kv_scale,
bool sage_attn) { // avoid overflow
bool sage_attn,
bool* used_flash_attn) { // avoid overflow
if (used_flash_attn != nullptr) {
*used_flash_attn = false;
}
int64_t L_q;
int64_t L_k;
int64_t C;
@@ -755,6 +759,9 @@ ggml_tensor* ggml_ext_attention_ext(ggml_context* ctx,
if (can_use_flash_attn) {
kqv = build_kqv(q, k, v, mask);
if (kqv != nullptr) {
if (used_flash_attn != nullptr) {
*used_flash_attn = true;
}
kqv = ggml_view_4d(ctx,
kqv,
d_head,