nrehiew_ dumps the unique RL training infra tricks: trajectory masking to HBM estimation
nrehiew_ · x · 2026-09-23
nrehiew lists the most unique infra techniques from the technical report:
- Mask out infra failures (e.g., unavailable tools) from trajectories
- Train on many harnesses but keep one group on the same harness for normalization
- Online scheduling based on rollout length and time
- Estimate trajectory length to determine HBM headroom for KV cache
He also questions the MOPD timeline and why MixRL is used for teacher training, which should be orthogonal.
More from Infra
- Agent Substrate Calculator Estimates Session Resume Speed Under Load — pjausovec · 2026-11-12
- Dell expands AI Data Platform: agents have a data-context problem, not a model problem — DavidLinthicum · 2026-11-12
- SGLang Megakernel Pushes Muse Spark to 1,109 tok/s, 2.1x Over Baseline — hsu_byron · 2026-11-12
- Potion UI's local login system turns your PC into a multi-user AI server — cocktailpeanut · 2026-11-12
- AWS Finally Built Its Own AWS Emulator — super_linear · 2026-11-12
- Megakernel is multiprogramming for GPUs: 'GPUs are entering the CPU age' — pranjalssh · 2026-11-12