Kevin Hu

Not sure what's going on with the unknown schema content warning, but this should fix the decimal128 mapping: https://github.com/datahub-project/datahub/pull/4624

It is thrown by `sqlalchemy`, but we are able to patch their type mapping this way. The warning is the same type as was raised in https://github.com/datahub-project/datahub/issues/3704

What do you mean by this? > Some information of the table structure, such as' distributed by hash() ',' aggregate key() ',' primary key() ',' unique key() ',' 'duplicate key()'...

I see—why are you trying to ingest Doris with the MySQL connector though?

I can't speak for the partitioning error, but the issue with UUID datatypes should now be resolved (by https://github.com/linkedin/datahub/pull/4179)!

I was running into this error too, so I tried vanilla pip to see if that could help diagnose the error. It looks like several of the Nvidia packages simply...

I'm only using pypi—I encountered the timeout when installing torch alone: ```sh uv venv source .venv/bin/activate uv pip install torch ```

I'm running into this as well—it seems to be more prevalent with larger models and also shows up when using `best_of`.

After some digging, the bug seems to be related to calling `_preempt_by_recompute` from `_preempt`, which inserts sequence groups at the front of the waiting queue. (But based on the TODO...