diff options
| author | Kawrakow <iwankawrakow@gmail.com> | 2025-07-03 18:03:23 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-07-03 18:03:23 +0200 |
| commit | 8a0c38f496f60bb0f627521823de604ce10fdc16 (patch) | |
| tree | 646fa2f21c7436dcae8214d382dda1a2e4b5f994 /src/llama.cpp | |
| parent | 9534461c01e132672821e53ce6e5e560dc58e829 (diff) | |
Vulkan: add GGML_OP_FUSED_MUL_UNARY (#580)
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'src/llama.cpp')
| -rw-r--r-- | src/llama.cpp | 8 |
1 files changed, 1 insertions, 7 deletions
diff --git a/src/llama.cpp b/src/llama.cpp index 794dcca6..11a7060c 100644 --- a/src/llama.cpp +++ b/src/llama.cpp @@ -9688,13 +9688,7 @@ static struct ggml_tensor * llm_build_ffn( cur = tmp; } -#ifdef GGML_USE_VULKAN - constexpr bool use_fused_mul_unary = false; -#else - constexpr bool use_fused_mul_unary = true; -#endif - - if (use_fused_mul_unary && type_gate == LLM_FFN_PAR && + if (type_gate == LLM_FFN_PAR && (type_op == LLM_FFN_SILU || type_op == LLM_FFN_RELU || (type_op == LLM_FFN_GELU && !act_scales))) { cur = ggml_fused_mul_unary(ctx, cur, tmp, type_op == LLM_FFN_SILU ? GGML_UNARY_OP_SILU : type_op == LLM_FFN_RELU ? GGML_UNARY_OP_RELU : GGML_UNARY_OP_GELU); |