ArjunSubramonian

Have you discovered anything else with your extra experiments? Assuming that the validation loss converges fine and doesn't immediately spike due to overfitting, warm-up shouldn't make a huge difference. The...

@niansong1996 would you be interested in implementing the sparse attention mechanism used by Longformer? I think it would be a great addition to our library. There are a variety of...

@dirkgr thoughts on implementing the positional embedding extension as an option for the BART model in AllenNLP?

I think this feature is a great idea! The latter design (passing a flag) seems better to me. I am adding @epwalsh here to get his input as well.

> @ArjunSubramonian, apparently, the prompt [convert_to_stereotype](https://github.com/bigscience-workshop/promptsource/blob/e3a22e09d0131a6ca6810ad8684c59eab3ede13d/promptsource/templates/BigScienceBiasEval/bias-shades/spanish/templates.yaml#L48) is raising some issues in [lm-eval-harness](https://github.com/bigscience-workshop/lm-evaluation-harness/pull/37#pullrequestreview-1171340441). Could you have a look at it? I took a look and don't know what the issue is...