Dear Docile'23 competitors,
We provide baseline implementations for KILE and LIR tasks that you can use for inspiration and to have some numbers to compare against before the competition ends and benchmark leaderboard becomes public.
The
docile repository now contains code to run training and inference for the provided baselines, instructions to download corresponding checkpoints together with predictions on the validation set (you can visualize those with
dataset browser) and evaluation results on both validation and test sets.
Kind regards,
On behalf of the organizers
Stepan Simsa