Popular repositories Loading

1. deepseek-v4-flash-gb10 deepseek-v4-flash-gb10 Public

   Serve DeepSeek-V4-Flash-0731 on 2x NVIDIA GB10 / DGX Spark (sm_121) with vLLM — tuned dual-Spark config (TP=2+EP, DSpark spec decode, NCCL 2.30.4/RDMA, 384K), runbook + verify harness. ~1.7k tok/s …

   Python 9 1

2. vllm-gfx1201-qwen3.6-256k vllm-gfx1201-qwen3.6-256k Public

   Qwen3.6-27B-FP8 at 256k context on dual AMD RDNA4 (gfx1201) — working vLLM recipe, audited measurement harness, and the full debugging record

   Python 4

3. vllm-metrics-dashboard vllm-metrics-dashboard Public

   Real-time single-screen dashboard for vLLM servers — throughput, latency, KV/prefix cache, and a cost/ROI panel vs hosted APIs. Standalone or fleet.

   HTML 1

4. flashnext-rocm10-r9700 flashnext-rocm10-r9700 Public

   Experimental ROCm 10 compatibility build for Qwen3.8-Flash-Next on 4x AMD R9700

   Dockerfile 1

5. hello-world hello-world Public

   This is a new repo

6. VS2013 VS2013 Public

   C#