TorchDR 0.5 lands with native multi-GPU support, beating cuML by 3.71x on a single GPU
leland_mcinnes · x · 2026-09-22
Leland McInnes (UMAP author) boosted the release of TorchDR 0.5, a dimensionality-reduction library:
- Native multi-GPU support via sharded FAISS search; modular architecture built on FAISS + PyTorch
- Beats cuML by 3.71x on 1M Tahoe single-cell samples even on a single GPU
- Adds neural-network encoders, out-of-sample transforms for neighbor embeddings (incl. UMAP), and faster, more memory-efficient UMAP optimization
More from Research
- Physicist pushes back on AI math proof criticism: strategy allowed by Clay criteria — skdh · 2026-09-22
- VR mocap pipeline captures 1,185 human trajectories to teach Unitree G1 to navigate clutter — Scobleizer · 2026-09-22
- AI memory system built with Jev claims 10x speed, 6x lower cost — realferrari · 2026-09-22
- AF3-style confidence heads surprisingly insensitive to diffusion module coordinates — chaitjo · 2026-09-22
- riderless: A Zero-Token Decision API on Gemma 4 Hits 93.1% Accuracy on One RTX 5090 — Pale-Soil-2524 · 2026-09-22
- Shenzhen University and HKUST propose ECA: evidence checks before agent actions — jiqizhixin · 2026-09-22