PyTorch 2.13 Released
PyTorch · x · 2026-07-18
PyTorch 2.13 has been released, bringing updates spanning attention, compilation, distributed training, memory efficiency, Python compatibility, and accelerator platform support.
This release comprises 3,328 commits from 526 contributors. A live Q&A session is scheduled for July 22, where maintainers and contributors will discuss the release and answer community questions.
Key changes include:
- FlexAttention
- Inductor's CuTeDSL Native DSL backend
- nn.LinearCrossEntropyLoss
- torchcomms
- FSDP2 communication overlap improvements
- Linux Python 3.15 wheel, including free-threaded 3.15t builds
- Expanded support for ROCm, Arm, and Intel XPU
More from Infra
- SF Compute founder: buying compute is 'an absolutely awful experience' right now — IgorCarron · 2026-09-11
- SmolVM open-sources persistent computer infrastructure for agents that outlive chat sessions — aniketmaurya · 2026-09-11
- PyTorch Day Korea 2026 launches first offline conf, CFP closes Sept 13 — PyTorch · 2026-09-11
- Local LLM server dilemma: 4x CMP-170HX (price up 53% in 20 days) vs Mac Studio M5 Ultra — rumboll · 2026-09-11
- llama.cpp lands Flash Attention tuning for RDNA4, big prefill gains on AMD — pmttyji · 2026-09-11
- Your p99 latency benchmark may be lying: a deep dive into coordinated omission — Franc0Fernand0 · 2026-09-11