furyhawk/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-07-21 02:05:35 +00:00
Code Issues Packages Projects Releases Wiki Activity

GLM DSA: reduce indexer cache size (#2093)

Browse Source
This commit is contained in:
Kawrakow
2026-07-07 09:35:42 +02:00
committed by GitHub
parent ba62eaffe7
commit 05cba319e0
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama.cpp
+1 -1
View File
@@ -1030,7 +1030,7 @@ static bool llama_kv_cache_init(
cache.k_l.push_back(kv);
// DSA lightning-indexer key cache (MQA, single head). Store the Hadamard-rotated
// indexer keys in F16 so a decoded token can score against ALL past keys.
if (has_dsa_indexer && model.layers[i].indexer_attn_k && !is_mtp_tail_layer) {
if (has_dsa_indexer && model.layers[i].indexer_attn_k && hparams.indexer_is_full[i] && !is_mtp_tail_layer) {
ggml_tensor * kr = ggml_new_tensor_2d(ctx, idx_type_k, hparams.indexer_head_size, kv_size);
ggml_format_name(kr, "cache_kr_l%d", i);
cache.kr_l[i] = kr;