Dependencies & Tools

The following table lists NVIDIA software packages with Spark support status, minimum supported version and a link to the latest supported release. Note that support might extend over time, so make sure to always check the latest version of this porting guide. Last Updated: March 9, 2026.

Package	Supported	Min. Version	Notes
AI Aerial	No	n/a
Compute Sanitizer	Yes	n/a	Included in CUDA Toolkit
Container Toolkit	Yes	1.17.7
cuCIM	No	n/a
cuBLAS	Yes	n/a	Included in HPC SDK and CUDA Toolkit
CUDA Toolkit	Yes	13.0
CUDA-GDB	Yes	n/a	Included in CUDA Toolkit
CUDA Math API	Yes	n/a	Included in CUDA Toolkit
cuDNN	Yes	9.11
cuDSS	Yes	0.7.0
cuEquivariance	Yes	0.9.0
cuFFT	Yes	n/a	Included in HPC SDK and CUDA Toolkit
cuOpt	Yes	25.12
CUPTI	Yes	n/a	Included in CUDA Toolkit
cuPyNumeric	Yes	26.01
cuQuantum	Yes	25.06
cuRAND	Yes	n/a	Included in HPC SDK and CUDA Toolkit
cuSolver	Yes	n/a	Included in HPC SDK and CUDA Toolkit
cuSPARSE	Yes	n/a	Included in CUDA Toolkit
cuTensor	Yes	2.3
CUTLASS	Yes	4.2.0	C++ Supported
cuVS	Yes	25.12
CV-CUDA	No	n/a
DeepStream	Yes	8.0
HPC SDK	No	n/a
Deep Learning Frameworks	Yes	25.09
DOCA	No	n/a
Model Optimizer	Yes	n/a	Dependent on TensorRT 10.14.1+
MONAI	Yes	TBD
NCCL	Yes	2.28.9-1
NPP	Yes	n/a	Included in CUDA Toolkit
Nsight Aftermath SDK	No	n/a
Nsight Cloud	No	n/a
Nsight Compute	Yes	n/a	Included in CUDA Toolkit
Nsight Deep Learning Designer	No	n/a
Nsight Graphics	No	n/a
Nsight Perf SDK	No	n/a
Nsight Systems	Yes	TBD
Nsight Tools JupyterLab Extension	No	n/a
Nsight Visual Studio Code Edition	Yes	2025.1
nvImageCodec	No	n/a
nvJPEG2000	No	n/a
nvmath-python	Yes	0.8
NVRTC	Yes	n/a	Included in CUDA Toolkit
nvTIFF	No	n/a
NVTX	Yes	n/a
OptiX	Yes	9.0
RAPIDS	Yes	25.10
Streamline	No	n/a
TensorRT	Yes	10.14.1
TensorRT for RTX	No	n/a
TensorRT-LLM	Yes	1.2
Triton	Yes	25.08
Warp	Yes	1.10.0