gpu-power

Here is 1 public repository matching this topic...

Power cap vs prefill and decode for a 35B MoE on vLLM, TP=2 on two RTX 3090: method, scripts, results.

benchmark moe ampere tensor-parallelism vllm llm-inference qwen speculative-decoding gpu-power rtx-3090

To associate your repository with the gpu-power topic, visit your repo's landing page and select "manage topics."