TransAct V2: Production System for Lifelong User Sequence Modeling at Scale
TL;DR
This work presents TransAct V2, a production ranking model deployed at Pinterest serving over 630 million users, that addresses challenges through a dual-path training-serving architecture with request-level deduplication and int8 quantization and provides a reproducible blueprint for practitioners building large-scale sequential recommenders.
Abstract
Modeling lifelong user action sequences in CTR prediction faces three production challenges: computational complexity with quadratic transformer costs as sequences extend from hundreds to tens of thousands of actions, infrastructure overhead from storage and network costs that scale with both sequence length and number of candidates per request, and weak training supervision as sequence encoders sit far from CTR prediction heads. We present TransAct V2, a production ranking model deployed at Pinterest serving over 630 million users, that addresses these challenges through: (1) a dual-path training-serving architecture with request-level deduplication and int8 quantization achieving 1% logging cost, (2) fused Triton kernels with pinned memory management delivering 250 × p99 latency improvement, and (3) a Next Action Loss providing direct supervision for sequence modeling within CTR frameworks. Deployed in production, TransAct V2 demonstrates significant improvements in both engagement and recommendation diversity. To support industry adoption, we open-source our serving optimizations with comprehensive ablation studies. Our work provides a reproducible blueprint for practitioners building large-scale sequential recommenders.