Pinned Loading

1. NTHU-2025PP NTHU-2025PP Public

   C++

2. glm-5.2-sglang glm-5.2-sglang Public

   Serving GLM-5.2 (704 GB, FP8) with SGLang + Apptainer on a Slurm HPC cluster — 8x H200 deployment record

   Shell

3. 5090-inference-playbook 5090-inference-playbook Public

   Tuning playbook: Gemma 4 31B on a single RTX 5090 with vLLM — quantization, KV-cache dtype, context length, MTP speculative decoding

   JavaScript