Tag: #재순위
Kubernetes, DevOps and MLOps, written from things we ran and measured ourselves · 1 posts
Parallelization, Multimodal, and Pooling: What You Can and Cannot Benchmark with a Single GPU
We have grouped the remaining major feature sets of vLLM. Tensor, pipeline, data, and expert parallelism divide models across multiple GPUs. Because we only have a single GPU, these were not benchmarked; instead, we have
2026-10-27 · 16 min read #vLLM#텐서 병렬#파이프라인 병렬#데이터 병렬#전문가 병렬