tonyd2wild/DeepSeek-V4.1-Flash-vLLM-DGX-Spark

DeepSeek-V4.1-Flash (552B MoE, MXFP4 experts, 1M ctx) on four NVIDIA DGX Sparks with vLLM TP4: Engram-on-disk patch, sm121 kernel build, launchers, measured numbers

13 stars 1 forks 13 watchers Python MIT License

View on GitHub

1 Open Issue Need Help Last updated: Sep 10, 2026

Open Issues Need Help

View All on GitHub

GB10 GPU flips between a fast and a slow hidden state (decode 63 vs 94 ms/step); nvidia-smi shows nothing 7 days ago

bug help wanted

tonyd2wild/DeepSeek-V4.1-Flash-vLLM-DGX-Spark

DeepSeek-V4.1-Flash (552B MoE, MXFP4 experts, 1M ctx) on four NVIDIA DGX Sparks with vLLM TP4: Engram-on-disk patch, sm121 kernel build, launchers, measured numbers

Python