Triton at the Core: Building Portable LLM Operator Libraries with FlagGems and Benchmarking Against Native CUDA
Build a portable LLM operator library with Triton and FlagGems, then benchmark it against native CUDA to quantify portability vs. performance trade-offs.

