🧬 Related Insights?

- **Read more:** [The 30-Second Rollback: Why Deploynix's Release Strategy Actually Works (And Why It Matters)](https://opensourcebeat.com/article/the-30-second-rollback-why-deploynixs-release-strategy-actually-works-and-why-it-matters/) - **Read more:** [Friday's Linux Security Storm: Kernel Patches That Could Save Your Server](https://opensourcebeat.com/article/security-updates-for-friday/) Frequently Asked Questions What is PyTorch 2.11's biggest new feature? Differentiable Collectives for backpropping through distributed ops — huge for scaling research. Does PyTorch 2.11 support Apple Silicon better? Yes, MPS gets error reporting, new distributions, and more ops. Still not NVIDIA-level. Is TorchScript gone in PyTorch 2.11? Deprecated. Switch to torch.export for models, Executorch for runtime.

🤖 AI & Machine Learning

PyTorch 2.11: 2723 Commits Later, FlashAttention Speeds Up — But TorchScript's Dead

2723 commits from 432 contributors. PyTorch 2.11 promises distributed training breakthroughs and GPU speedups. But deprecating TorchScript? That's a gut punch for legacy code.

theAIcatchup Apr 07, 2026 4 min read

PyTorch 2.11 release graphic with GPU icons and commit stats

⚡ Key Takeaways

Differentiable Collectives enable backprop through distributed training ops without hacks. 𝕏
FlexAttention with FlashAttention-4 delivers up to 3.2x speedups on NVIDIA Hopper/Blackwell. 𝕏
TorchScript deprecated; migrate to torch.export. 2026 sees bi-monthly releases. 𝕏

Published by

theAIcatchup

Community-driven. Code-first.

#Apple Silicon MPS #FlexAttention #MPS Apple Silicon #PyTorch 2.11 #TorchScript deprecation #distributed training

Worth sharing?

Get the best Open Source stories of the week in your inbox — no noise, no spam.

Originally reported by PyTorch Blog

⚡ Key Takeaways

The 60-Second TL;DR

theAIcatchup

Share this article

Worth sharing?

Related Stories

The 429 Error That's Killing Your AI Fallback Chains Dead

DFlash Cracks Open Speculative Decoding's Parallel Future

Reclaiming the Wheel: How I Code with AI Without Handing Over the Keys

AI's Hidden Bills: Track Multi-Provider Costs Before They Bury You

Stay in the loop