Popular repositories Loading

1. legacy-gpu-llm-notes legacy-gpu-llm-notes Public

   Measured llama.cpp and quantization results on Tesla V100 (sm_70), Pascal GTX 1070, and RTX 4070 - hardware most projects don't test on

   Python 1 1

2. volta-gpt-oss volta-gpt-oss Public

   gpt-oss-20b MXFP4 runs on a Tesla V100 (sm_70) at 143 t/s, despite MXFP4 being documented as requiring compute capability 9.0+

   1

3. strata-vs-llamacpp strata-vs-llamacpp Public

   Strata vs llama.cpp's MoE expert cache, head to head on a V100 with 16 GB RAM

   Shell 1

4. volta-kernel-bakeoff volta-kernel-bakeoff Public archive

   Our own quantized matmul kernels for the Tesla V100 (sm_70) raced against cuBLAS and llama.cpp at real MoE shapes -- checked, sabotage-tested, 30-round clock-locked medians, and what a whole model …

   Cuda 1

5. EnvisionOS EnvisionOS Public

   envision os repository

6. volta-bonsai volta-bonsai Public

   Ternary-Bonsai-2-27B on a Tesla V100 (sm_70): ternary kernels and vision both work on Volta

   Shell