hard20 pts · first answer countsmodel-servingparallelism
Splitting a model across two nodes
A model needs 2 nodes of 8 GPUs. Nodes are linked by a network much slower than NVLink. Which layout usually gives the best latency?
Discussion 0
Join free to post in the discussion.
Related questions
- When batch inference fitsMLOps Engineer
- Traffic before the model has loadedMLOps Engineer
- Why export a model to ONNXMLOps Engineer
- Protocol for internal inference callsMLOps Engineer
- Latency and throughputMLOps Engineer
- What TTFT measuresMLOps Engineer
- Gap between streamed tokensMLOps Engineer
- Two phases of LLM generationMLOps Engineer