What did you find confusing? Please describe.
Different versions of Triton support different Python versions depending on which version of Ubuntu the container is based on (i.e. up to 23.02 ships with py38 as the default, later versions ship with py310)
Describe how documentation can be improved
Explicitly/correctly document which system python is shipped with the image.
Additional context
Unlike most sagemaker containers where additional packages are installed via requirements.txt Triton requires custom packages to be installed into a conda venv and then compressed using conda pack. This requires explicit knowledge of which system python the Triton image contains and this isn't py38 since 23.02.
The documentation for the Triton image (https://github.com/aws/deep-learning-containers/blob/master/available_images.md#nvidia-triton-inference-containers-sm-support-only) mentions
py38as the python version option. However the Example URL containspy3.What did you find confusing? Please describe. Different versions of Triton support different Python versions depending on which version of Ubuntu the container is based on (i.e. up to 23.02 ships with py38 as the default, later versions ship with py310)
Describe how documentation can be improved Explicitly/correctly document which system python is shipped with the image.
Additional context Unlike most sagemaker containers where additional packages are installed via
requirements.txtTriton requires custom packages to be installed into acondavenv and then compressed usingconda pack. This requires explicit knowledge of which system python the Triton image contains and this isn't py38 since 23.02.