From 44155ba538fd65d498498df639421c342aaaecc4 Mon Sep 17 00:00:00 2001 From: Sainbayar Sukhbaatar Date: Tue, 3 Jan 2017 15:18:47 -0500 Subject: [PATCH] Update README.md typo --- levers/README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/levers/README.md b/levers/README.md index 8ff2b73..8e77562 100644 --- a/levers/README.md +++ b/levers/README.md @@ -10,10 +10,10 @@ For more details about the task, see our [paper](https://arxiv.org/abs/1605.0773 ## Usage The code uses Torch and run on a GPU for speed-up. To start training, run - th lever.lua --reward_only --comm + th levers.lua --reward_only --comm Here, option `--comm` enables a communication between the agents, and `--reward_only` picks a reinforcement learning instead of supervised learning. The following script can be used to replicated the experiments in the paper. - th lever.lua --batchsize 512 --lr 10 --clip .01 --hdim 64 --apg 5 --nlevers 5 --reward_only --maxiter 100000 --comm + th levers.lua --batchsize 512 --lr 10 --clip .01 --hdim 64 --apg 5 --nlevers 5 --reward_only --maxiter 100000 --comm