The ZeroDivisionError issue was fixed and training models for the 5th round of wikis has been completed successfully.

Sorry for the late response. @elukey regarding cleaningup/deprecating, the last I heard was wikilabels was to be replaced by our own labeling tool called pilot-flag proposed by @calbon. I'll leave him to clarify.

Generating datasets and training models for all wikis in this round went well besides Tibetan Wikipedia - bowiki.

The migration of draftquality models has been completed.

Inference services were created for all the 2 draftquality models and they are all up and running in KServe on both eqiad and codfw.

2/2 draftquality models were uploaded successfully to Thanos Swift.

As I was thinking through async, I noticed the feature extraction code is being called twice here: https://github.com/wikimedia/machinelearning-liftwing-inference-services/blob/main/revscoring/editquality/model-server/model.py#L47-L48

The migration of articlequality models has been completed.

Inference services were created for all the 13 articlequality models and they are all up and running in KServe on both eqiad and codfw.

13/13 articlequality models were uploaded successfully to Thanos Swift.

@Isaac thank you so much for the detailed break down.

In order to deploy the content translation recommendation model to Lift Wing, we need to upload the model files to storage, so our Inference Services can download the binaries and mount them into the pod.

I have opened a separate task here T309263 and shared the location of the models.

@kostajh, thank you for the confirmation. We have published the datasets for the 17/19 wikis that passed the evaluation.

Since @MGerlach mentioned that some heuristics are failing when working with the skipped wikis, I think the next steps would be to manually inspect the models with users who have experience with these languages or use google-translate as the link-recommendation algorithm is iteratively improved until the models pass the backtesting evaluation.

If you would like to check wiki models before they are deployed, I think the backtesting evaluation can be used for this.

@kostajh, we completed training models for the fourth round of wikis (listed in the task description) and shared the models' evaluation above that suggested we exclude jawiki and aswiki for the time being. We are now ready to publish the datasets for the wikis that passed the model evaluation, should we proceed?

Training models for the 4th round of wikis has been completed successfully.

Generating datasets and training models for the first 18 wikis in this round went well. When I reached the Japanese Wiki, I kept running into the error shown in the screenshot below. To keep archives happy, this issue was solved by installing both mecab-python3 and unidic-lite in a venv.

@Trizek-WMF, thank you for creating all the rounds. I am working on generating datasets and models round by round and will be sharing updates on the sub-tasks.

13/13 articlequality models were uploaded successfully to Thanos Swift.

The migration of editquality models has been completed.