Miles ships Day-0 RL support for DeepSeek-V4.1-Flash with KL held at 0.0012–0.0017
ying11231 · x · 2026-09-10
radixark's RL training framework Miles added Day-0 support for DeepSeek-V4.1-Flash:
- Trainer stays close to SGLang sampling; parallelism and shared state let the new architecture scale intact across GPUs
- Quantization-aware training mirrors SGLang's FP4/FP8 rounding, with routing replay reusing rollout expert choices
- FP32 and deterministic reductions ensure numerical consistency; colocated training/rollout fits full-parameter RL on 16 GPUs
Over steps 0–80 of a DAPO run, per-token trainer–rollout KL stayed at 0.0012–0.0017 while reward rose from 0.51 to 0.78. Blog and cookbook in the comments.
More from Infra
- Anthropic S-1 reveals 48 cents of compute cost per dollar of Q2 revenue — nmasc_ · 2026-10-26
- Persistent AI joins UR network to route private AI inference with encrypted, IP-hidden traffic — 0xSammy · 2026-10-26
- Anthropic's disclosed compute commitments: $114.5B Google, $161.2B Broadcom top the list — BenBajarin · 2026-10-26
- SpaceX strikes $40B deal with Apollo Global to lease AI chips and data center hardware — Polymarket · 2026-10-26
- SpaceX AI revenue up 830%, Claude designs 318K antivenom proteins in 64 hours — PeterDiamandis · 2026-10-26
- Meta plans up to $95 billion stock sale to fund AI buildout — The Decoder · 2026-10-26