[Usage]: Is there any document for using trt-llm to run pipeline parallel inference unevenly on multiple jetson nodes?

System Info

3 jetson Orin, 1 server with 4090
cuda-python:13.0
cuda:12.6
tensor-RT: v0.12.0-jetson

How would you like to use TensorRT-LLM

Is there any document for using trt-llm to run pipeline parallel inference unevenly on multiple jetson nodes?

For example, we may use pipeline parallel within 3 jetson Orin nodes and distribute uneven portions of the model to each node. More specifically, let's say a model has 80 decoder blocks. I want to deploy 50 on the first node, 20 on the second node, and 10 on the third node.

Is this supported by trt-llm (we not use triton)?

Thank you.

Before submitting a new issue...

• Make sure you already searched for relevant issues, and checked the documentation and examples for answers to frequently asked questions.

Start with the TensorRT-LLM documentation and examples linked in the issue, then inspect the existing pipeline-parallel documentation for Jetson or multi-node coverage. Determine whether uneven decoder-block placement across three nodes is supported without Triton. Done should be a documented answer with any relevant usage guidance or a clear limitation.