From f3c3eafa6e604bba7d3224fb429a339d59d5ccf1 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Johannes=20G=C3=A4=C3=9Fler?= <johannesg@5d6.de>
Date: Sat, 11 May 2024 15:32:46 +0200
Subject: [PATCH] fixup! fixup! CUDA: add FP32 FlashAttention vector kernel

---
 ggml-cuda/fattn-vec-f16.cu | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ggml-cuda/fattn-vec-f16.cu b/ggml-cuda/fattn-vec-f16.cu
index e79fb9b7a..54307fcfb 100644
--- a/ggml-cuda/fattn-vec-f16.cu
+++ b/ggml-cuda/fattn-vec-f16.cu
@@ -312,7 +312,7 @@ void ggml_cuda_flash_attn_ext_vec_f16(ggml_backend_cuda_context & ctx, ggml_tens
 
     ggml_tensor * KQV = dst;
 
-    const int32_t precision = KQV->op_params[1];
+    const int32_t precision = KQV->op_params[2];
     GGML_ASSERT(precision == GGML_PREC_DEFAULT);
 
     constexpr int cols_per_block = 1;