diff --git a/.github/workflows/publish.yml b/.github/workflows/publish.yml
index 119014b..12c221a 100644
--- a/.github/workflows/publish.yml
+++ b/.github/workflows/publish.yml
@@ -45,7 +45,7 @@ jobs:
           os: [ubuntu-20.04]
           python-version: ['3.7', '3.8', '3.9', '3.10', '3.11']
           torch-version: ['1.12.1', '1.13.1', '2.0.1', '2.1.0.dev20230731']
-          cuda-version: ['11.6.2', '11.7.1', '11.8.0', '12.1.0']
+          cuda-version: ['11.6.2', '11.7.1', '11.8.0', '12.1.0', '12.2.0']
           # We need separate wheels that either uses C++11 ABI (-D_GLIBCXX_USE_CXX11_ABI) or not.
           # Pytorch wheels currently don't use it, but nvcr images have Pytorch compiled with C++11 ABI.
           # Without this we get import error (undefined symbol: _ZN3c105ErrorC2ENS_14SourceLocationESs)
@@ -145,7 +145,7 @@ jobs:
           export PATH=/usr/local/nvidia/bin:/usr/local/nvidia/lib64:$PATH
           export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:$LD_LIBRARY_PATH
           # Currently for this setting the runner goes OOM if we pass --threads 4 to nvcc
-          if [[ ${MATRIX_CUDA_VERSION} == "121" && ${MATRIX_TORCH_VERSION} == "2.1" ]]; then
+          if [[ ${MATRIX_CUDA_VERSION} =~ "12." && ${MATRIX_TORCH_VERSION} == "2.1" ]]; then
             export FLASH_ATTENTION_FORCE_SINGLE_THREAD="TRUE"
           fi
           # Limit MAX_JOBS otherwise the github runner goes OOM