All Posts
2026
October
August
- Neural Network Quantization
- Activations, Normalization and Optimizers, in Plain Terms
- PyTorch Distributed Training Internals
- Tiling and Collective Operations
- The KV Cache: Sizing It, and the Attention Variants That Shrink It
- LLM Training: The End-to-End Flow
- FlashInfer: A Customizable Attention Engine for LLM Serving
- FlashAttention in Practice: Four Schedules and One Branch