sparverius

See: https://github.com/EleutherAI/lm-evaluation-harness/blob/7411947112117e0339fe207fb620a70bcec22690/lm_eval/__main__.py#L129-L136 add ability to modify temperature, top_k, top_p, and most relevant = repetition_penalty 🫠 https://github.com/tcapelle/llm_recipes/blob/cf5b38b0dad0bd188b4d8d3111a46bff54b9d163/eval_harness/run.py#L21