Hello Flyte Community! Im running flyte on a sandbox deployment for experimentation. I´ve created a becnhmark file with the tasks that I need to perform. Also ran

docker_build.sh

to create the custom image that I would like use to run the workflow, instead of the flyte default one. Im running the pyflyte run command on the virtual environment that has my dependencies but it still fails, and I have no clue why.

try

pyflyte run --remote image

Will try it

You do not have mlflow installed locally

Hello! Thank you kindly for all your responses. I'm Lucas' partner in crime. We do have mlflow installed into the local conda env which is active while executing that command. If, per example, we start an interactive session and try to import it, we can make use of it no problem.

Ohh sorry missed the message

@Ketan (kumare3) let me know if you need any more information regarding this

I would love as much as I can get

I don't see any

no module named mlflow

error. I just tried running an example. @Lucas Parisi, can you try reinstalling your virtual environment?

@Samhita Alla just paste your example here instead of back and forth

I ran the example present in the docs.

@Samhita Alla the

no module named mlflow

is shown on the first image attached.

That error shouldn't occur when you install the mlflow library, and hence, I suggest recreating your virtual environment. Have you tried that?

Hello! Sorry for the late response. Just wanted to update you and thank you, the reconstruction of the environment indeed worked! First I tried manually, and then from a compiled txt from a requirements.in and now it's working perfectly. Still I have no idea how I messed up for it to no detect mlflow in the other env while being installed right there, but that was that. Again, thank you very much! 🤗