llama : fix K-shift with quantized K and BLAS backend (#13113)

2025-10-29 08:41:22 +00:00 · 2025-04-25 19:40:11 +02:00
parent 558a764713
commit 295354ea68
2 changed files with 4 additions and 16 deletions
--- a/src/llama-context.h
+++ b/src/llama-context.h
@@ -170,8 +170,7 @@ private:
        ggml_tensor * shift,
        ggml_tensor * factors,
              float   freq_base,
-              float   freq_scale,
-        ggml_backend_buffer * bbuf) const;
+              float   freq_scale) const;

    llm_graph_result_ptr build_kv_self_shift(
            ggml_context * ctx0,