Allowing port mapping with triton start

[AvivSham]

Hi All,
Hope you're doing well.

We're attempting to run a LLaMA model on a machine that already has ports 8000, 8001, and 8002 in use. However, when using the triton start --help command, we couldn't find any option to change the HTTP, gRPC, or metrics ports.

Is there a recommended way to work around this?

We also tried port mapping using Docker with the following command, but as expected, it didn’t resolve the issue:

docker run --gpus '"device=2"' --rm -p 9000:8000 -p 9001:8001 -p 9002:8002 \
  -v /ds/users/artifacts/llama_models:/models \
  model:latest \
  tritonserver --model-repository=/models

Cheers!