Ragged Paged Attention: A High-Performance and Flexible LLM Inference Kernel for TPU
## Episode Summary In this episode, we cover: - **Ragged Paged Attention: A High-Performance and Flexible LLM Inference Kernel for TPU** (arXiv) - **Overmind NSA: A Unified Neuro-Symbolic Computing Architecture with Approximate Nonlinear Activations and Preemptive Memory Bypass** (arXiv) - **RISC-V Vector Extension: Between standardization and tailor-made accelerators - eeNews Europe** (google_riscv) - **RISC-V set to announce 25% market penetration — open-standard ISA is ahead of schedule, securing fast-growing silicon footprint - Tom's Hardware** (google_riscv) - **QuMA: Researchers Develop…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.