From 167a46128ae18a5778cecd8e5452c9abe93160aa Mon Sep 17 00:00:00 2001 From: Ting Lu Date: Sat, 18 May 2024 20:20:37 -0700 Subject: [PATCH 1/3] add nvidia driver for cuda arm wheel --- .github/actions/setup-nvidia/action.yml | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/.github/actions/setup-nvidia/action.yml b/.github/actions/setup-nvidia/action.yml index 34f0308cb3..26cdc0ba48 100644 --- a/.github/actions/setup-nvidia/action.yml +++ b/.github/actions/setup-nvidia/action.yml @@ -8,6 +8,10 @@ inputs: required: false type: string default: "550.54.15" # https://docs.nvidia.com/datacenter/tesla/tesla-release-notes-550-54-15/index.html + GPU_ARCH_TYPE: + required: true + type: string + description: GPU Arch type runs: using: composite @@ -16,6 +20,11 @@ runs: uses: nick-fields/retry@3e91a01664abd3c5cd539100d10d33b9c5b68482 env: DRIVER_VERSION: ${{ inputs.driver-version }} + GPU_ARCH_TYPE: ${{ inputs.GPU_ARCH_TYPE }} + ${{ if eq(inputs.GPU_ARCH_TYPE , 'cuda-aarch64') }}: + ARCH_TYPE: sbsa + ${{ else }}: + ARCH_TYPE: x86_64 with: timeout_minutes: 10 max_attempts: 3 @@ -25,7 +34,7 @@ runs: set -eou pipefail DISTRIBUTION=$(. /etc/os-release;echo $ID$VERSION_ID) - DRIVER_FN="NVIDIA-Linux-x86_64-${DRIVER_VERSION}.run" + DRIVER_FN="NVIDIA-Linux-${GPU_ARCH_TYPE}-${DRIVER_VERSION}.run" YUM_REPO_URL="https://nvidia.github.io/nvidia-docker/${DISTRIBUTION}/nvidia-docker.repo" install_nvidia_docker2_amzn2() { From 55b882f07c33701aabd5dec2ed92f74de32213c9 Mon Sep 17 00:00:00 2001 From: Ting Lu Date: Sat, 18 May 2024 22:02:18 -0700 Subject: [PATCH 2/3] update var name --- .github/actions/setup-nvidia/action.yml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.github/actions/setup-nvidia/action.yml b/.github/actions/setup-nvidia/action.yml index 26cdc0ba48..01e955a54c 100644 --- a/.github/actions/setup-nvidia/action.yml +++ b/.github/actions/setup-nvidia/action.yml @@ -34,7 +34,7 @@ runs: set -eou pipefail DISTRIBUTION=$(. /etc/os-release;echo $ID$VERSION_ID) - DRIVER_FN="NVIDIA-Linux-${GPU_ARCH_TYPE}-${DRIVER_VERSION}.run" + DRIVER_FN="NVIDIA-Linux-${ARCH_TYPE}-${DRIVER_VERSION}.run" YUM_REPO_URL="https://nvidia.github.io/nvidia-docker/${DISTRIBUTION}/nvidia-docker.repo" install_nvidia_docker2_amzn2() { From 63aa40bc725027bb29df08f5d9cb9ac915c0b12f Mon Sep 17 00:00:00 2001 From: Ting Lu Date: Mon, 20 May 2024 07:57:05 -0700 Subject: [PATCH 3/3] Add var ARCH_TYPE and default to x86_64 --- .github/actions/setup-nvidia/action.yml | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) diff --git a/.github/actions/setup-nvidia/action.yml b/.github/actions/setup-nvidia/action.yml index 01e955a54c..b7f421355b 100644 --- a/.github/actions/setup-nvidia/action.yml +++ b/.github/actions/setup-nvidia/action.yml @@ -9,9 +9,15 @@ inputs: type: string default: "550.54.15" # https://docs.nvidia.com/datacenter/tesla/tesla-release-notes-550-54-15/index.html GPU_ARCH_TYPE: - required: true + description: GPU Arch type #"cuda", "cuda-aarch64" + required: false + type: string + ARCH_TYPE: + description: CPU Arch Type + required: false type: string - description: GPU Arch type + default: "x86_64" + runs: using: composite @@ -21,10 +27,9 @@ runs: env: DRIVER_VERSION: ${{ inputs.driver-version }} GPU_ARCH_TYPE: ${{ inputs.GPU_ARCH_TYPE }} + ARCH_TYPE: ${{ inputs.ARCH_TYPE }} ${{ if eq(inputs.GPU_ARCH_TYPE , 'cuda-aarch64') }}: ARCH_TYPE: sbsa - ${{ else }}: - ARCH_TYPE: x86_64 with: timeout_minutes: 10 max_attempts: 3