cuda

DeepGEMM: clean and efficient BLAS kernel library on GPU

Qwen3.8-27B on a single RTX 3090 with vLLM: ~1,000 tok/s at 64 concurrent (int8 tensor-core GEMMs, fp16 DeltaNet state), ~114 tok/s single-user at default sampling / ~124 greedy (MTP drafts, own-output draft vocab, calibrated int4 lm_head, split-KV verify attention), 150k-262k context; patches, requant scripts, benchmarks

Cross-vendor 3D Gaussian Splatting trainer - video to splat to mesh, Vulkan or CUDA.

Catches silent CPU fallback and mislabeled tok/s in local LLMs. llama.cpp and ollama, one file, no deps.