Deploying DeepSeek-V4 on Blackwell: Fixing 3 critical SGLang bugs
shrug_hellifino · reddit · 2026-09-02
The author shared their experience running DeepSeek-V4-Flash-Vision-Exp via SGLang on dual RTX PRO 6000 Blackwell (SM120) cards, successfully handling a 269k-token context. They resolved three critical issues: SM120 sparse-MLA Vision prefill crashes (fixed by adding capability checks and falling back to Triton), a CUDA OOM in the indexer caused by 269k context (resolved via row slicing and chunking to bound memory), and a bug where Vision preview corrupted multi-turn tool-call history.
More from coding & agent
- LLM one-shots a Minecraft 3x3 piston door by building its own Linux test infra — Vjeux · 2026-10-18
- Dev builds Mumbai-inspired 3D manga background generator from OpenStreetMap data — CurieuxExplorer · 2026-10-18
- Solo founder builds a money-printing data service on Claude's Pi Durable harness — MicahBerkley · 2026-10-18
- Dev builds Arch-based OS with custom desktop and browser using Claude Code — Snoo-10493 · 2026-10-18
- Five SaaS agent workflows fail the same way — and it's never model intelligence — usehive · 2026-10-18
- Claude skill generates hand-drawn interactive diagrams that teach kids how things work — CatAstro_Piyush · 2026-10-18