Data Science platform#
Welcome to KSL’s Data Science platform. Here you will find the documentation on how to run Data Science and AI workloads on computational resources of KSL systems. The pages listed here walks you through from getting started on KSL systems, preparing your datasets, developing your models, and running the production jobs on multiple GPUs, and everything in between. The aim of this documentation is not to teach you Data Science by to focus on how to leverage KSL computational resources to increase the productivity in your workflows.
Getting started#
Computational resources on KSL systems#
Storage on KSL systems#
Running Jobs on KSL systems#
Software environment#
Accelerating workloads#
Debugging and profiling#
- In-flight job telemetry with NVDashboard
- Profiling GPU workloads with NVIDIA Nsight
- Nsight Systems with NVTX instrumentation
- Profiling a CUDA kernel written in C++ using Nsight Systems
- Profiling a CUDA kernel written in Fortran using Nsight Systems
- Profiling C++ code with OpenACC directives using Nsight Systems
- Profiling an NVIDIA RAPIDS workflow using Nsight Systems
- Profiling a CUDA Ufunc written in Python using Nsight Systems
- Profiling PyTorch workloads on single GPU with NVIDIA Nsight