raullenchai/twinspark

Production-grade recipe for DeepSeek-V4-Flash-0731 (284B MoE) on 2x NVIDIA DGX Spark: self-healing 2-node vLLM cluster, reboot-verified, tuned DSpark speculative decoding (~75 tok/s), full benchmarks, OpenAI Codex CLI integration

语言：Shell · GitHub 星标：176 · Fork：20

话题：deepseek · dgx-spark · llm-inference · local-llm · vllm

在 GitHub 上查看原仓库