/lm-evaluation-harness-fast

speedup for lm-evaluation-harness; support tensor-parallel inference and data-parallel inference; support gptq, bitsandbytes, peft and exllamav2.

Primary LanguagePythonMIT LicenseMIT

Stargazers