triton-inference-server / fastertransformer_backend

BSD 3-Clause "New" or "Revised" License
411 stars 133 forks source link

Some docker build fixes #94

Closed tanmayv25 closed 1 year ago

tanmayv25 commented 1 year ago

These changes seems to be needed to run "create_dockerfile_and_build.py" directly.

python3 docker/create_dockerfile_and_build.py --triton-version=22.12 --image-name tritonserver_with_ft -q

Above command will fail without these changes. I am testing a build with Triton' build.py and update soon.

byshiue commented 1 year ago

Can you post the error you encounter? We cannot reproduce this issue.

tanmayv25 commented 1 year ago

My bad. The issue comes when I am trying to build FasterTransformer Backend with Triton Min images.

docker/create_dockerfile_and_build.py --base-image="nvcr.io/nvidia/tritonserver:22.12-py3-min" --image-name tritonserver_with_ft -q

The build is getting stuck at:

Please select the geographic area in which you live. Subsequent configuration
questions will narrow this down by presenting a list of cities, representing
the time zones in which they are located.

  1. Africa      4. Australia  7. Atlantic  10. Pacific  13. Etc
  2. America     5. Arctic     8. Europe    11. SystemV
  3. Antarctica  6. Asia       9. Indian    12. US
Geographic area: 

After fixing it with DEBIAN_FRONTEND=noninteractive, it is unable to configure cmake as NVIDIA_TRITON_SERVER_VERSION is not defined.

May be we can modify our min containers to get this fixed there. I am open to suggestion. The reason of using min container is that user might not want to pull a container image with all the backends if they just want a container image with FT backend.