sklearn.model selection.StratifiedKFold either shuffling is wrong or documentation is misleading <!-- If…
sklearn.model selection.StratifiedKFold either shuffling is wrong or documentation is misleading <!-- If…: a task in Lego-RL-SWE-Bench-Verified (Harbor dataset). As you can see in the output below, 1 is always paired with 11, 2 with 12, 3 with 13, etc. regardless whether shuffle parameter is True…
The task
As you can see in the output below, 1 is always paired with 11, 2 with 12, 3 with 13, etc. regardless whether shuffle parameter is True or False. When shuffle=True, the batches are always the same for any random_state, but appear in a different order.
Part of Lego-X/Lego-RL-SWE-Bench-Verified.