Model
stringclasses 3
values | Framework
stringclasses 1
value | Accuracy
stringclasses 1
value | Inferential Mode
stringclasses 1
value | TP
int64 1
4
| Max_Input_Output
stringclasses 4
values | Concurrency
int64 3
92
| TTFT
float64 144
4.46k
| TPOT
float64 49.7
55
| Generate TPS
float64 44.8
1.79k
| Total TPS
float64 1.33k
6.79k
| QPS
float64 0.11
4.5
|
|---|---|---|---|---|---|---|---|---|---|---|---|
Llama-3.1-8B-Instruct | vllm | FP8 | server | 1 | (1000, 500) | 92 | 144.027 | 49.7313 | 1,793.1533 | 5,836.0988 | 4.501 |
Llama-3.1-8B-Instruct | vllm | FP8 | server | 1 | (2000, 500) | 62 | 264.4247 | 49.7329 | 1,201.3084 | 6,617.8854 | 3.0138 |
Llama-3.1-8B-Instruct | vllm | FP8 | server | 1 | (3600, 500) | 38 | 445.2348 | 49.7002 | 731.0105 | 6,789.137 | 1.8406 |
Llama-3.1-8B-Instruct | vllm | FP8 | server | 1 | (20000, 500) | 7 | 2,493.5093 | 50.0167 | 122.3333 | 5,592.8262 | 0.3022 |
Qwen2.5-72B-Instruct_FP8_TP4 | vllm | FP8 | server | 4 | (1000, 500) | 23 | 326.0608 | 49.6567 | 407.1565 | 1,331.8259 | 1.0289 |
Qwen2.5-72B-Instruct_FP8_TP8 | vllm | FP8 | server | 4 | (3600, 500) | 9 | 802.3972 | 49.8334 | 164.919 | 1,548.7753 | 0.4186 |
Qwen2.5-72B-Instruct_FP8_TP8 | vllm | FP8 | server | 4 | (20000, 500) | 3 | 4,463.8949 | 55.0398 | 44.771 | 2,011.0025 | 0.1092 |
Subsets and Splits