I was getting this error when running the Exercise 14 pipeline. This came in the preprocessing step.

OSError: Could not open parquet input source '': Invalid: Parquet magic bytes not found in footer. Either the file is corrupted or this is not a parquet file.

Looking into it I saw that the data was not being downloaded correctly. Moreover the link actually returns a 404 error.

Link in the config.yaml
https://raw.githubusercontent.com/udacity/nd0821-c2-build-model-workflow/master/lesson-2-data-exploration-and-preparation/exercises/exercise_4/starter/genres_mod.parquet?token=ABNEVO6ANSAW5NZCIJ5ZXV3AOVMZK

Changing it to the following link seems to solve this issue
https://github.com/udacity/nd0821-c2-build-model-workflow-exercises/blob/master/lesson-2-data-exploration-and-preparation/exercises/exercise_4/starter/genres_mod.parquet?raw=true

1. should be val_size instead of test_size
   https://github.com/udacity/nd0821-c2-build-model-workflow-exercises/blob/master/lesson-5-final-pipeline-release-and-deploy/exercises/exercise_14/solution/main.py#L92

2. random_state should be one of the input parameters
   https://github.com/udacity/nd0821-c2-build-model-workflow-exercises/blob/master/lesson-5-final-pipeline-release-and-deploy/exercises/exercise_14/solution/main.py#L71-L72

3. the type is not list but some custom list under omegaconf, assert will fail this case
   https://github.com/udacity/nd0821-c2-build-model-workflow-exercises/blob/master/lesson-5-final-pipeline-release-and-deploy/exercises/exercise_14/starter/main.py#L23

Please do rework an all exercises of part 3, they are not working anymore, even not the given solution. I tried it with Python 3.8 with a new virtual environment, after realising that the common Python 3.10 is not usable at all.

It is waisting most of the time with the MLOps course to get at least something to run, but only for the first exercises this was possible with a newer w&b version installed via conda and not via pip.