9e4e6fe207
[CI] Improve the readability of performance benchmarking results and prepare for upcoming performance dashboard (#5571)
32 lines
867 B
JSON
32 lines
867 B
JSON
[
|
|
{
|
|
"test_name": "latency_llama8B_tp1",
|
|
"parameters": {
|
|
"model": "meta-llama/Meta-Llama-3-8B",
|
|
"tensor_parallel_size": 1,
|
|
"load_format": "dummy",
|
|
"num_iters_warmup": 5,
|
|
"num_iters": 15
|
|
}
|
|
},
|
|
{
|
|
"test_name": "latency_llama70B_tp4",
|
|
"parameters": {
|
|
"model": "meta-llama/Meta-Llama-3-70B-Instruct",
|
|
"tensor_parallel_size": 4,
|
|
"load_format": "dummy",
|
|
"num-iters-warmup": 5,
|
|
"num-iters": 15
|
|
}
|
|
},
|
|
{
|
|
"test_name": "latency_mixtral8x7B_tp2",
|
|
"parameters": {
|
|
"model": "mistralai/Mixtral-8x7B-Instruct-v0.1",
|
|
"tensor_parallel_size": 2,
|
|
"load_format": "dummy",
|
|
"num-iters-warmup": 5,
|
|
"num-iters": 15
|
|
}
|
|
}
|
|
] |