nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 9 Packages Projects Releases Wiki Activity

spec : add DFlash support (#22105)

Browse Source 
* spec: add DFlash v2 support

* dflash: support sliding window attention per layer_types

* docs: add dflash section

---------

Co-authored-by: Kashif Rasul <kashif.rasul@gmail.com>
This commit is contained in:
Ruixiang Wang
2026-06-28 16:01:34 +03:00
committed by GitHub
co-authored by Kashif Rasul
parent c1a1c8ee94
commit d1b34251bc
14 changed files with 712 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-llama-archs.cpp
+2 -2
View File
@@ -451,7 +451,7 @@ static int save_models(const llm_arch target_arch, const size_t seed, const ggml
if (arch == LLM_ARCH_GEMMA4 || arch == LLM_ARCH_GEMMA4_ASSISTANT) {
continue; // FIXME: ISWA KV cache initialization needs more fixture params
}
if (arch == LLM_ARCH_EAGLE3) {
if (arch == LLM_ARCH_EAGLE3 || arch == LLM_ARCH_DFLASH) {
continue;
}
for (bool moe : {false, true}) {
@@ -557,7 +557,7 @@ static int test_backends(const llm_arch target_arch, const size_t seed, const gg
if (arch == LLM_ARCH_GEMMA4 || arch == LLM_ARCH_GEMMA4_ASSISTANT) {
continue; // FIXME: ISWA KV cache initialization needs more fixture params
}
if (arch == LLM_ARCH_EAGLE3) {
if (arch == LLM_ARCH_EAGLE3 || arch == LLM_ARCH_DFLASH) {
continue;
}