diff --git a/ggml/src/ggml-cuda/fattn-vec-f16.cu b/ggml/src/ggml-cuda/fattn-vec-f16.cu index 89d3d9a2..d75a5601 100644 --- a/ggml/src/ggml-cuda/fattn-vec-f16.cu +++ b/ggml/src/ggml-cuda/fattn-vec-f16.cu @@ -110,7 +110,6 @@ bool ggml_cuda_fattn_vec_f16_is_supported([[maybe_unused]] ggml_backend_cuda_con if (K->type != V->type) return false; return K->type == GGML_TYPE_F16 || K->type == GGML_TYPE_Q8_0; } - bool supported = false; #ifdef GGML_CUDA_FA_ALL_QUANTS if (K->ne[0] == 64) { return K->type == GGML_TYPE_F16 && diff --git a/ggml/src/ggml-cuda/fattn-vec-f32.cu b/ggml/src/ggml-cuda/fattn-vec-f32.cu index e7e918f2..c1a5ad2c 100644 --- a/ggml/src/ggml-cuda/fattn-vec-f32.cu +++ b/ggml/src/ggml-cuda/fattn-vec-f32.cu @@ -110,7 +110,6 @@ bool ggml_cuda_fattn_vec_f32_is_supported([[maybe_unused]] ggml_backend_cuda_con if (K->type != V->type) return false; return K->type == GGML_TYPE_F16 || K->type == GGML_TYPE_Q8_0; } - bool supported = false; #ifdef GGML_CUDA_FA_ALL_QUANTS if (K->ne[0] == 64) { return K->type == GGML_TYPE_F16 &&