PreprocessBuildJob is translation-specific, not generic to all engine types

[Enkidu93]

No description provided.

[johnml1135]

I would argue that there are many commonalities - that we are generally taking corpora and placing them on the S3 bucket for processing by a ClearML job. I believe we can maintain a inheritance relationship between the various preprocess build jobs.

What would be your rationale for splitting it up to be entirely separate entities?

[Enkidu93]

I would argue that there are many commonalities - that we are generally taking corpora and placing them on the S3 bucket for processing by a ClearML job. I believe we can maintain a inheritance relationship between the various preprocess build jobs.

What would be your rationale for splitting it up to be entirely separate entities?

This is a place-holder issue connected to the word alignment PR. We can still keep a PrerprocessBuilldJob, it's just that currently the WriteDataFilesAsync function is overwritten in the word alignment preprocessing job since the generic PreprocessBuildJob still does things specific to pretranslation - for example, writes to the pretranslate.json files. It would be cleaner for that to be moved to a translation-specific job.

If you'd prefer, I can do this as part of the current PR.

[johnml1135]

Ok - good point. We can wait to do this for the next release.

[BiancaPaw]

Technical debt.

[Enkidu93]

This was completed here: #506