Meta open-sources Component Benchmark, a hierarchical profiler for TB-scale recommender models
_reachsumit · x · 2026-09-28
Meta introduces Component Benchmark (CB), a profiling system for large-scale recommendation models. RecSys architectures mix memory-bandwidth-bound ops, small dense layers, and dynamic shapes from jagged categorical features, and standard tools only offer end-to-end throughput or operator-level traces.
CB independently profiles each submodule and presents latency, memory, and MFU in an interactive tree view, with a plugin architecture for extensible hierarchical analysis. It was validated on TB-scale models running across thousands of GPUs and ingesting 100B examples per day.
More from Infra
- Chamath breaks down AI compute: prefill is compute-bound, decode is bandwidth-bound — rohanpaul_ai · 2026-09-28
- Taiwan companies scramble for advanced packaging talent, says analyst — LIWEI_TWCapital · 2026-09-28
- Developer says local AI is shifting from nice-to-have to infrastructure: control beats privacy — Aiden_Tech_Ai · 2026-09-28
- apple-llm: Node/Python wrapper for the free local LLM built into Apple Silicon Macs — light_2earth · 2026-09-28
- Running 8 watercooled GPUs for local AI: one user's case for watercooling over air cooling — HanchungLee · 2026-09-28
- HF: transformers backend now matches native vLLM speed, no porting needed — ariG23498 · 2026-09-28