From 25bd85c17dd503af097e3b7355b6dc46b073f477 Mon Sep 17 00:00:00 2001 From: allanjie Date: Wed, 11 Mar 2020 17:23:02 +0800 Subject: [PATCH] remind users to use glove embeddings for reproducing the results --- docs/benchmark.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/docs/benchmark.md b/docs/benchmark.md index 33324f4..a86ca9b 100644 --- a/docs/benchmark.md +++ b/docs/benchmark.md @@ -1,6 +1,8 @@ ### Benchmark Performance Empirically, although `ADAM` optimizer converges faster, we found that using `SGD` with learning rate of `0.01` and `100` epochs is better than `ADAM`. +The following results are obtained with the Glove embeddings (`glove.6B.100d.txt`). +Download the embeddings and specify the path as arguments. * Experiments on the CoNLL-2003 dataset