A collection of different benchmarks
Find a file
Simon Richter d8c8c95162 Add koboldcpp backend
Allows benchmarking koboldcpp alongside llamacpp using the same
GGUF model cache. Downloads models from HuggingFace via stdlib
urllib if not already cached by llamacpp.

Usage: --backends "llamacpp:auto koboldcpp:auto"
2026-03-04 14:45:46 +01:00
llm Add koboldcpp backend 2026-03-04 14:45:46 +01:00
yocto yocto/run-benchmark.sh: use https for cloning 2024-08-14 12:22:51 +02:00
.gitignore initial commit 2024-07-11 16:52:49 +02:00