From aef65609e1ae2bf5d03c4a2827b5fea626b433a9 Mon Sep 17 00:00:00 2001
From: Andrii Staikov <andrii.staikov@intel.com>
Date: Wed, 13 Nov 2024 18:44:39 +0100
Subject: [PATCH] skip cuda

---
 .github/workflows/job_pytorch_layer_tests.yml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.github/workflows/job_pytorch_layer_tests.yml b/.github/workflows/job_pytorch_layer_tests.yml
index dbf0e1913c3db1..4eeed787d1678e 100644
--- a/.github/workflows/job_pytorch_layer_tests.yml
+++ b/.github/workflows/job_pytorch_layer_tests.yml
@@ -122,8 +122,8 @@ jobs:
       - name: Install flash_attn module
         run: |
           # due to flash_attn issues, it needs to be installed separately from other packages
-          # pip install flash_attn --no-build-isolation
-          pip install https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.0.post1/flash_attn-2.7.0.post1+cu12torch2.5cxx11abiTRUE-cp312-cp312-linux_x86_64.whl
+          export FLASH_ATTENTION_SKIP_CUDA_BUILD=TRUE
+          pip install flash_attn --no-build-isolation
 
       - name: PyTorch Layer Tests
         if: ${{ fromJSON(inputs.affected-components).PyTorch_FE.test && runner.arch != 'ARM64' }} # Ticket: 126287, 142196