From 49c4d154ae1f258d260a898fc1bd23013995deff Mon Sep 17 00:00:00 2001 From: Inforithmics Date: Tue, 12 Aug 2025 21:55:19 +0200 Subject: [PATCH] Enable Vulkan Flash attention in FlashAttentionSupported --- discover/gpu.go | 3 +-- discover/types.go | 3 ++- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/discover/gpu.go b/discover/gpu.go index 8a36db854..123177d3a 100644 --- a/discover/gpu.go +++ b/discover/gpu.go @@ -422,8 +422,7 @@ func GetGPUInfo() GpuInfoList { C.free(unsafe.Pointer(memInfo.err)) continue } - - gpuInfo.FlashAttention = true + gpuInfo.TotalMemory = uint64(memInfo.total) gpuInfo.FreeMemory = uint64(memInfo.free) gpuInfo.ID = C.GoString(&memInfo.gpu_id[0]) diff --git a/discover/types.go b/discover/types.go index 183c51ae2..39830dc5a 100644 --- a/discover/types.go +++ b/discover/types.go @@ -182,7 +182,8 @@ func (l GpuInfoList) FlashAttentionSupported() bool { supportsFA := gpu.Library == "cpu" || gpu.Library == "metal" || (gpu.Library == "cuda" && gpu.DriverMajor >= 7) || - gpu.Library == "rocm" + gpu.Library == "rocm" || + gpu.Library == "vulkan" if !supportsFA { return false