PixelML/club-170hx

GLM-5.3-Flash on 4× NVIDIA CMP 170HX with vLLM: up to 482.6 tok/s single-user, up to 943.7 tok/s at 8 users, 262K context, TP4 or PP4, one-command container install

Exemplar Performance provides recipes in ready-to-use templates for evaluating performance of specific AI use cases across hardware and software combinations.

Tuning and qualification harness for the NVIDIA CMP 170HX (GA100). Measures, gates and recovers clock/voltage settings, and treats a completed benchmark as evidence of nothing.

Local AI registry: one validated recipe per machine, with the evidence attached

DeepSeek-V4-Flash-0731 on 4x CMP 170HX (sm_80): 98 tok/s decode, ~5300 tok/s prefill. DSpark speculative decoding under pipeline parallelism, which vLLM does not support upstream.

Single-request LLM serving recipes, patches, and benchmarks for modified RTX 2080 Ti 22GB / SM75 systems

Collective Knowledge (CK), Collective Mind (CM/CMX) and MLPerf automations: community-driven projects to learn how to run AI, ML, and other emerging workloads more efficiently and cost-effectively across diverse models, datasets, software, and hardware using MLPerf methodology and benchmarks.

A tool to unlobotomize your NVIDIA card!

本仓库整理了在 HCU 硬件上部署、调优和运行 AI 模型的经验与最佳实践