[GH-ISSUE #89] Non-disaggregated performance 怎么测试 #49

New issue

Closed

opened 2026-10-29 21:08:25 +00:00 by bsakaguchi · 2 comments

bsakaguchi commented

2026-10-29 21:08:25 +00:00

Owner

Originally created by @ToBeResumed on GitHub (Jan 22, 2025).
Original GitHub issue: https://github.com/kvcache-ai/Mooncake/issues/89

在vllm_benchmark result中给出的几种测试对比组，请问Non-disaggregated模式是怎么设置的，对应步骤里哪步？

Originally created by @ToBeResumed on GitHub (Jan 22, 2025). Original GitHub issue: https://github.com/kvcache-ai/Mooncake/issues/89 在vllm_benchmark result中给出的几种测试对比组，请问Non-disaggregated模式是怎么设置的，对应步骤里哪步？

bsakaguchi closed this issue

2026-10-29 21:08:25 +00:00

bsakaguchi commented

2026-10-29 21:08:26 +00:00

Author

Owner

@ShangmingCai commented on GitHub (Jan 22, 2025):

@ToBeResumed Non-disaggregated的实验你直接通过vllm起实例就可以了，通过传入参数开启chunked-prefill，比如

CUDA_VISIBLE_DEVICES=0 python3 \
    -m vllm.entrypoints.openai.api_server \
    --model $model \
    --port 8100 \  
    --max-model-len 10000 \ 
    --enable-chunked-prefill \
    --gpu-memory-utilization 0.8

@ShangmingCai commented on GitHub (Jan 22, 2025): @ToBeResumed Non-disaggregated的实验你直接通过vllm起实例就可以了，通过传入参数开启chunked-prefill，比如 ``` CUDA_VISIBLE_DEVICES=0 python3 \ -m vllm.entrypoints.openai.api_server \ --model $model \ --port 8100 \ --max-model-len 10000 \ --enable-chunked-prefill \ --gpu-memory-utilization 0.8 ```

bsakaguchi commented

2026-10-29 21:08:26 +00:00

Author

Owner

@ToBeResumed commented on GitHub (Jan 23, 2025):

谢谢解答，新年快乐

@ToBeResumed commented on GitHub (Jan 23, 2025): 谢谢解答，新年快乐