Closed wangpatrick57 closed 4 months ago
I added some instrumentation to see why "root" wasn't being populated. In about 20 trials on my laptop, it was all because of this check: https://github.com/cmu-db/dbgym/blob/66cb2d43220d01c3a7996cb03a1c870f5e767686/tune/protox/embedding/select.py#L119-L120
@17zhangw suggested removing the check -- there should be no downside to doing so.
Details: select.py:36
Reproduce: Ran
pat_test.sh
on the embedding-datagen branch here: https://github.com/wangpatrick57/dbgym/tree/embedding-datagen. A specific .parquet file causing this error can be found inrun_2024-02-28_17-42-33
on dev4