Hello Flyte Team! I want to use custom built docker image to use our custom dependencies. I tested on flyte sandbox on my local macbook. (Running from

flytectl demo start

) My workflow was successfully run on local (

pyflyte run cifar10.py wf

) But my workflow needs

torch

packages, so I built docker image with Dockerfile like below:

FROM python:3.9-slim-buster

RUN python -m pip install torch torchvision

RUN python -m pip install flytekit pandas numpy awscli

I pushed the image to dockerhub, it can be accessed publicly. And I submit flyte workflow with

pyflyte

cli

pyflyte run --remote --image {username}/flyte:0.0.4 cifar10.py wf

But it doesn’t work on remote (sandbox). How can I do?

Hey Seung-Woo, can you please test this to see whether it is a flyte or a kubernetes problem?

kubectl run  -i --tty --rm test --pod-running-timeout=20m0s --image={username}/flyte:0.0.4 --restart=Never -- bash

If this pod starts and gives you a shell, kubernetes can pull the image. If this doesn’t start, can you please post the events at the bottom of

kubectl describe pod test

?

I got my code error. i missed to add /root to PYTHONPATH env variable. After setting it, it works!

Perfect 🙂

Should I add /root to PYTHONPATH and copy my python file to /root?

adding /root to PYTHONPATH is enough, and you should copy python file to

WORKDIR

WORKDIR /tmp
COPY python_file /tmp

Originally I put the file(e.g.

train.py

) to docker image, i run workflow with

pyflyte run --remote --image {image_name} train.py wf

. It seems that the change of file (

train.py

) affected without rebuild the image. Is it intended?

Yes, it is.

pyflyte run

does fast registration. So you need not include code in your Docker image.