From e371bea04fae22b43a24d3a30904b9830ed88ae3 Mon Sep 17 00:00:00 2001 From: juejuezi Date: Fri, 6 Sep 2024 01:34:35 +0800 Subject: [PATCH] feat: change minimal supported CUDA version to 11.7 (#1206) --- README.md | 2 +- setup.py | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index 3e2e066..054af18 100644 --- a/README.md +++ b/README.md @@ -98,7 +98,7 @@ MAX_JOBS=4 pip install flash-attn --no-build-isolation ### NVIDIA CUDA Support **Requirements:** -- CUDA 11.6 and above. +- CUDA 11.7 and above. We recommend the [Pytorch](https://catalog.ngc.nvidia.com/orgs/nvidia/containers/pytorch) diff --git a/setup.py b/setup.py index fd67f64..1f149c5 100644 --- a/setup.py +++ b/setup.py @@ -156,9 +156,9 @@ if not SKIP_CUDA_BUILD and not IS_ROCM: cc_flag = [] if CUDA_HOME is not None: _, bare_metal_version = get_cuda_bare_metal_version(CUDA_HOME) - if bare_metal_version < Version("11.6"): + if bare_metal_version < Version("11.7"): raise RuntimeError( - "FlashAttention is only supported on CUDA 11.6 and above. " + "FlashAttention is only supported on CUDA 11.7 and above. " "Note: make sure nvcc has a supported version by running nvcc -V." ) # cc_flag.append("-gencode")