PyTorch ships a JAX-style stateless RNG: SM-independent and free of GPU fingerprinting
ezyang · x · 2026-11-17
PyTorch core dev ezyang highlights the new stateless RNG API: ideal when you need careful management of randomness, and SM-independent, so no more fingerprinting GPU types. He recounts the history — the upstream issue was never fixed, leading to many bodgy userland workarounds — until @giffmana pushed attention to stateless RNG and jbschlosser implemented the API based on JAX's design. Worth a try.
Related event: PyTorch Ships Stateless RNG API to Fix Reproducibility Woes(2 posts)→
More from Research
- Study: word-based output encoding cuts small-model silent errors 2.3x — ITMO · 2026-11-17
- Physicist: AI now beats most junior theorists, and researchers should ride the dragon — IgorCarron · 2026-11-17
- "We formalized it in Lean" — researcher questions what formal proofs actually verify — DimitrisPapail · 2026-11-17
- CVPR 2027 Announced for Seattle, June 20-25, 2027 — CSProfKGD · 2026-11-17
- BeyondBinding challenge launches: AI-designed antibodies get wet-lab tested — QuanquanGu · 2026-11-17
- Singularity-Bench: Opus 5.5 with Claude Code beats human baselines on building AI systems — armand_ruiz · 2026-11-17