How to build Tensorrt-LLM without docker?

[beginlner]

[juney-nvidia]

In theory, as long as you have the required dependency(CUDA, Python, TRT, etc) properly set up, you can build TensorRT-LLM without docker. However, it is still suggested to build TensorRT-LLM with the docker-based approach to reduce the potential tricky environment issue.

In our daily development, we mainly build TensorRT-LLM inside docker.

June