Evgeny Pavlov
Evgeny Pavlov
I assumed if a user trains on a local machine, not a cluster, they have root access. We can make it a setting on the top of the makefile, but...
It's essentially the same as https://github.com/mozilla/firefox-translations-training/issues/472. Now we train everything in one run with OpusTrainer and don't have a problem using the worse fine-tuned model since the pre-trained checkpoint will...
Closing this for now since we're going to migrate to OpusCleaner for language and dataset-specific fixes.
We want to clean everything with OpusCleaner eventually including the custom fixes of the datasets.
This change and dataset are not critical regardless and I prefer not to have very old PRs hanging. We'll add this if needed as a part of the OpusCleaner config.
I didn't see this for some time and I assume it's fixed by using OpusTrainer.
This work was a part of project Bergamot, we can revisit it later if we want to implement this approach as a part of the TaskCluster pipeline.
I believe this is meant to solve specific issues on HPC that we're not using at the moment. Closing
Let's wait until we ship some models in Nightly.
Ok, it might take a while since we're focusing on general cleaning now and only when it's done we'll start a full retraining. I think let's close some of those...