memory : handle kv_unified for hybrid models (#15050)

2025-08-18 05:56:00 -04:00 · 2025-08-03 15:43:07 -04:00
parent 97366dc6ab
commit 11a3811164
3 changed files with 4 additions and 1 deletions
--- a/src/llama-memory-hybrid.cpp
+++ b/src/llama-memory-hybrid.cpp
@@ -25,6 +25,7 @@ llama_memory_hybrid::llama_memory_hybrid(
                         /* common */
             uint32_t    n_seq_max,
                 bool    offload,
+                 bool    unified,
                         /* layer filters */
      layer_filter_cb && filter_attn,
      layer_filter_cb && filter_recr) :
@@ -38,7 +39,7 @@ llama_memory_hybrid::llama_memory_hybrid(
        type_v,
        v_trans,
        offload,
-        1,
+        unified,
        kv_size,
        n_seq_max,
        n_pad,