Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions spark-rapids/mig.sh
Original file line number Diff line number Diff line change
Expand Up @@ -212,9 +212,9 @@ fi
readonly OS_NAME=$(lsb_release -is | tr '[:upper:]' '[:lower:]')

# CUDA version and Driver version config
CUDA_VERSION=$(get_metadata_attribute 'cuda-version' '12.4.1') #12.2.2
NVIDIA_DRIVER_VERSION=$(get_metadata_attribute 'driver-version' '550.54.15') #535.104.05
CUDA_VERSION_MAJOR="${CUDA_VERSION%.*}" #12.2
CUDA_VERSION=$(get_metadata_attribute 'cuda-version' '13.1.1') #13.1.1
NVIDIA_DRIVER_VERSION=$(get_metadata_attribute 'driver-version' '590.48.01') #590.48.01
CUDA_VERSION_MAJOR="${CUDA_VERSION%.*}" #13.1

# EXCEPTIONS
# Change CUDA version for Ubuntu 18 (Cuda 12.1.1 - Driver v530.30.02 is the latest version supported by Ubuntu 18)
Expand Down
15 changes: 7 additions & 8 deletions spark-rapids/spark-rapids.sh
Original file line number Diff line number Diff line change
Expand Up @@ -226,7 +226,7 @@ else
fi

# Update SPARK RAPIDS config
readonly DEFAULT_SPARK_RAPIDS_VERSION="26.04.2"
readonly DEFAULT_SPARK_RAPIDS_VERSION="26.06.0"
readonly SPARK_RAPIDS_VERSION=$(get_metadata_attribute 'spark-rapids-version' ${DEFAULT_SPARK_RAPIDS_VERSION})
readonly XGBOOST_VERSION=$(get_metadata_attribute 'xgboost-version' ${DEFAULT_XGBOOST_VERSION})

Expand All @@ -236,9 +236,9 @@ readonly MASTER=$(/usr/share/google/get_metadata_value attributes/dataproc-maste
readonly RUNTIME=$(get_metadata_attribute 'rapids-runtime' 'SPARK')

# CUDA version and Driver version config
CUDA_VERSION=$(get_metadata_attribute 'cuda-version' '12.4.1') #12.2.2
NVIDIA_DRIVER_VERSION=$(get_metadata_attribute 'driver-version' '550.54.15') #535.104.05
CUDA_VERSION_MAJOR="${CUDA_VERSION%.*}" #12.2
CUDA_VERSION=$(get_metadata_attribute 'cuda-version' '13.1.1') #13.1.1
NVIDIA_DRIVER_VERSION=$(get_metadata_attribute 'driver-version' '590.48.01') #590.48.01
CUDA_VERSION_MAJOR="${CUDA_VERSION%.*}" #13.1

# EXCEPTIONS
# Change CUDA version for Ubuntu 18 (Cuda 12.1.1 - Driver v530.30.02 is the latest version supported by Ubuntu 18)
Expand Down Expand Up @@ -450,7 +450,7 @@ function install_nvidia_gpu_driver() {
execute_with_retries "apt-get install -y -q nvidia-kernel-open-dkms"
clear_dkms_key
execute_with_retries \
"apt-get install -y -q --no-install-recommends cuda-drivers-${NVIDIA_DRIVER_VERSION_PREFIX}"
"apt-get install -y -q --no-install-recommends nvidia-driver-cuda"
execute_with_retries \
"apt-get install -y -q --no-install-recommends cuda-toolkit-${CUDA_VERSION_MAJOR//./-}"

Expand Down Expand Up @@ -571,8 +571,7 @@ function install_nvidia_gpu_driver() {
execute_with_retries "apt-get install -y -q --no-install-recommends nvidia-driver-${NVIDIA_DRIVER_VERSION_PREFIX}-open"
else
# Standard installation from local CUDA repo
for pkg in "nvidia-driver-${NVIDIA_DRIVER_VERSION_PREFIX}-open" \
"cuda-drivers-${NVIDIA_DRIVER_VERSION_PREFIX}" \
for pkg in "nvidia-driver-cuda" \
"cuda-toolkit-${CUDA_VERSION_MAJOR//./-}" ; do
execute_with_retries "apt-get install -y -q --no-install-recommends ${pkg}"
done
Expand Down Expand Up @@ -626,7 +625,7 @@ function download_agent(){
if [[ ${OS_NAME} == rocky ]]; then
execute_with_retries "dnf -y -q install git"
else
execute_with_retries "apt-get install git -y"
execute_with_retries "apt-get install -y -q git python3-venv python3-pip"
fi
mkdir -p /opt/google
chmod 777 /opt/google
Expand Down
2 changes: 1 addition & 1 deletion spark-rapids/test_spark_rapids.py
Original file line number Diff line number Diff line change
Expand Up @@ -114,7 +114,7 @@ def test_spark_rapids_sql(self, configuration, machine_suffixes, accelerator):
# Only need to do this once
self.verify_spark_job_sql()

@parameterized.parameters(("STANDARD", ["w-0"], GPU_T4, "12.4.0", "550.54.14"))
@parameterized.parameters(("STANDARD", ["w-0"], GPU_T4, "12.4.0", "590.48.01"))
def test_non_default_cuda_versions(self, configuration, machine_suffixes,
accelerator, cuda_version, driver_version):

Expand Down