search

triton-inference-server / dali_backend

The Triton backend that allows running GPU-accelerated data pre-processing pipelines implemented in DALI's python API.
https://docs.nvidia.com/deeplearning/dali/user-guide/docs/index.html
MIT License
123 stars 29 forks source link

Remove conda package manager #235

Closed mc-nv closed 6 months ago

mc-nv commented 6 months ago

This PR removes the conda package manager from the Triton system.

Conda has been used to isolate DALI wheel (which needs to be installed apart from being unpacked in the container) from the rest of Triton packages. Since we're moving away from this requirement, DALI can be installed bare-metal inside the Triton container and conda is no longer necessary.

dali-automaton commented 6 months ago

CI MESSAGE: [13959349]: BUILD STARTED

mzient commented 6 months ago

@mc-nv Please provide some description and motivation for this change. If it's part of a broader effort in TRITON, please describe it and/or provide some links to the related changes or relevant documentation.

szalpal commented 6 months ago

@mc-nv Please provide some description and motivation for this change. If it's part of a broader effort in TRITON, please describe it and/or provide some links to the related changes or relevant documentation.

I'll do it

nv-kmcgill53 commented 6 months ago

I just want to clarify that there is no time dependence on this change making it into this release, so if there needs to be more work done to make this better, let's do that. The LTS branches are fine with the previous changes we made and the next LTS release which may need this constraint is not for another couple of years.

dali-automaton commented 6 months ago

CI MESSAGE: [13959349]: BUILD FAILED