LLM inference in C/C++
C++ ggml
context : disable non-fused GDN and LID ops (#27877)
Georgi Gerganov86632248188c
diff --git a/src/llama-context.cpp b/src/llama-context.cppindex 179c526c29..9aed801332 100644--- a/src/llama-context.cpp+++ b/src/llama-context.cpp@@ -231,10 +231,10 @@ llama_context::llama_context(cparams.fused_gdn_ar = true;cparams.fused_gdn_ch = true;- cparams.auto_fgdn = true;+ cparams.auto_fgdn = false;- cparams.fused_lid = true;- cparams.auto_flid = true;+ cparams.fused_lid = true;+ cparams.auto_flid = false;cparams.fused_dsv4_hc_pre = true;cparams.fused_dsv4_hc_comb = true;