📄 文章列表

23 篇文章
TDS

Building a Production-Grade Multi-Node Training Pipeline with PyTorch DDP

✅ completed ⭐ ★★★★★★ 2026-03-27
TDS

Prefill Is Compute-Bound. Decode Is Memory-Bound. Why Your GPU Shouldn’t Do Both.

✅ completed ⭐ ★★★★★★ 2026-04-15
TDS

The Counterintuitive Networking Decisions Behind OpenAI’s 131,000-GPU Training Fabric

✅ completed ⭐ ★★★★★★ 2026-05-14