tsne-cuda 3.0.1-foss-2023a-CUDA-12.1.1

A optimized CUDA version of FIt-SNE algorithm with associated python modules. We find that our implementation of t-SNE can be up to 1200x faster than Sklearn, or up to 50x faster than Multicore-TSNE when used with the right GPU. The paper describing our approach, as well as the results below, is available at https://arxiv.org/abs/1807.11824.

Accessing tsne-cuda 3.0.1-foss-2023a-CUDA-12.1.1

To load the module for tsne-cuda 3.0.1-foss-2023a-CUDA-12.1.1 please use this command on the BEAR systems (BlueBEAR and BEAR Cloud VMs):

📋 module load bear-apps/2023a
module load tsne-cuda/3.0.1-foss-2023a-CUDA-12.1.1

BEAR Apps Version

2023a

Architectures

EL8-icelake (GPUs: NVIDIA A100, NVIDIA A30)

The listed architectures consist of two parts: OS-CPU. The OS used is represented by EL and there are several different processor (CPU) types available on BlueBEAR. More information about the processor types on BlueBEAR is available on the BlueBEAR Job Submission page.

More Information

For more information visit the tsne-cuda website.

Dependencies

This version of tsne-cuda has a direct dependency on: CUDA/12.1.1 Faiss/1.7.4-foss-2023a-CUDA-12.1.1 foss/2023a gflags/2.2.2-GCCcore-12.3.0 Python/3.11.3-GCCcore-12.3.0 ZeroMQ/4.3.4-GCCcore-12.3.0

Required By

This version of tsne-cuda is a direct dependent of: DynaMight/0.0.0.20240319.eef4aa6-foss-2023a-CUDA-12.1.1

Last modified on 3rd March 2025