ds4-on-spark
Entrpi/ds4, a Blackwell CUDA perf fork of antirez/ds4 on NVIDIA DGX Spark: one-command install, ~3x upstream prefill, ~1.5x decode, DSpark, and full continuous batch support
- 416
- Shell
- MIT
- 3 mo ago
Entrpi
Building the future with you.
Built from Entrpi's public GitHub profile. Not on Take My Repo yet, so there are no posts or updates here.
Entrpi/ds4, a Blackwell CUDA perf fork of antirez/ds4 on NVIDIA DGX Spark: one-command install, ~3x upstream prefill, ~1.5x decode, DSpark, and full continuous batch support
The most extreme way to train a GPT in pure, dependency-free C. 84000x faster than Python. Optimized for Apple Silicon with SME2.
Qwen3.5-122B-A10B on a DGX Spark with DFlash speculative decode. One-shot Docker/vLLM installer. 80+ tok/s!
GLM-5.3-Flash (EXL3 4bpw + DFlash2) on 2x NVIDIA DGX Spark: one-shot installer, 33-74 tok/s c1, 1.3M+ context, vision
Headless / multi-user serving-mode for the DGX Spark: free unified memory for vLLM by paring back the desktop. Companion to qwen3.5-122B-A10B-on-spark.
antirez/ds4-style hybrid quant DeepSeek V4 Flash on a single DGX Spark via vLLM
vLLM fork branch serving GLM-5.3-Flash on 2x DGX Spark (GB10): glm5_next port + sm121 fixes + EXL3 fused serving + DFlash2 speculative decode + ring draft-KV. Community derivative; setup repo: Entrpi/glm-5.3-flash-exl3-2x-spark
Autonomy Golf is a game you play with your projects: hole-in-one is total automation of a commit.
AC power meter monitor for Matrix MPM-1010B with terminal graphs and native GUI