Jae-Won Chung
New leaderboard prototype
b10121d
raw
history blame contribute delete
326 Bytes
{
"Model": "codellama/CodeLlama-34b-hf",
"GPU": "NVIDIA A100-SXM4-40GB",
"TP": 4,
"PP": 1,
"Energy/req (J)": 61.37366100182621,
"Avg TPOT (s)": 0.20305942186377088,
"Token tput (tok/s)": 1084.6096537850954,
"Avg Output Tokens": 86.33231707317073,
"Avg BS (reqs)": 316.72967614011895,
"Max BS (reqs)": 320
}