Couple of questions
#5
by
abhisheklalwani
- opened
- What is the exact submission deadline (including the time)?
- I can see that there are some test examples which do not contain a [MASK] token. What is the deal with them? Are they used for calculating our score? Or can we use them as additional training data?
Hi @abhisheklalwani Please see answers below
- The exact deadline is 27th may 0:00:00 UTC. You can see it when you click on private leaderboard.
- Some test samples don't have [MASK] (some of the data in training also didn't have them). Results are calculated only sentences which have [MASK]. It is important to ensure that you 'image_id' for the sentence is correct because it is used for calculating metrics. You can see some reference in the helper notebook we provide here (https://huggingface.co/datasets/generative-newsai/news-unmasked/blob/main/.extras/helper_notebook.ipynb). You can use the sentences with no [MASK] as you want but we anticipate they will not many samples (as compared to large training data) and they are not used for final scores:
