I understand that metaflow typically loads the current working directory and subdirectories in flows so that they're available. Is there a file size limit where directories won't be loaded in the context? I have a situation where a tensorflow model exists in a directory relative to the flow of

model/tensorflow_model/__various_files__

. Locally I'm able to load this model, but when running on kubernetes that directory is absent. I've confirmed this by using pathlib to list the files in the

model

directory and some smaller python files are there, but other pkl files and the

tensorflow_model

directory are absent

yes - you can add

--package-suffixes=.txt,.json

to ensure those get packaged up as well

That worked, thanks! Also we're using keras, and longer term will leverage this method: https://outerbounds-community.slack.com/archives/C02116BBNTU/p1680624622388349?thread_ts=1680608119.325749&cid=C02116BBNTU