From a6ec5ed43c91e3caececf71a86726b567e5dca14 Mon Sep 17 00:00:00 2001 From: Brian DuSell Date: Fri, 6 Oct 2017 09:29:04 -0400 Subject: [PATCH] Average in some more tests: ~85% -> ~80% --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 03f0f31..4b23dc3 100644 --- a/README.md +++ b/README.md @@ -45,7 +45,7 @@ sequence reversal transduction task as described in the paper. Note that it is rather typical for this architecture to get permanently stuck in a local minimum during training, so you may need to try more than once to get a successful run. Using the default values for `train.py`, you should be able to -successfully train a model about 85% of the time. +successfully train a model about 80% of the time. Note that the default size of the model is quite large, and the script is best run on GPU in this case. If you want to test out a smaller version of the