From cc55425770f3dec6b48f2c31dfc856624bd2d519 Mon Sep 17 00:00:00 2001 From: ndvb <845175+ndvbd@users.noreply.github.com> Date: Wed, 25 Jan 2023 12:46:53 +0200 Subject: [PATCH] Update released_checkpoints.md --- released_checkpoints.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/released_checkpoints.md b/released_checkpoints.md index a5583da2..c751c980 100644 --- a/released_checkpoints.md +++ b/released_checkpoints.md @@ -8,7 +8,7 @@ These checkpoints are not officially supported - use at your own risk! Similar to the models described in our [paper][paper], with the following improvements: -* GEGLU activation in feed-forward hidden layer, rather than ReLU - see https://arxiv.org/abs/2002.05202 . +* GELU activation in feed-forward hidden layer, rather than ReLU - see https://arxiv.org/abs/2002.05202 . * Dropout was turned off in pre-training (quality win). Dropout should be re-enabled during fine-tuning.