From 26600e5d58e92e9c9906242fbb4bb0b481f2547e Mon Sep 17 00:00:00 2001
From: slaren <slarengh@gmail.com>
Date: Thu, 21 Mar 2024 13:59:53 +0100
Subject: [PATCH] cuda : fix LLAMA_CUDA_F16 build (#6197)

---
 ggml-cuda.cu | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ggml-cuda.cu b/ggml-cuda.cu
index 280839ea475cd..04c6f5d07d4ca 100644
--- a/ggml-cuda.cu
+++ b/ggml-cuda.cu
@@ -9453,7 +9453,7 @@ static void ggml_cuda_op_dequantize_mul_mat_vec(
 
     // on some GPUs it is faster to convert src1 to half and to use half precision intrinsics
 #ifdef GGML_CUDA_F16
-    cuda_pool_alloc<half> src1_dfloat_a;
+    ggml_cuda_pool_alloc<half> src1_dfloat_a(ctx.pool());
     half * src1_dfloat = nullptr; // dfloat == half
 
     bool src1_convert_f16 =