From fde1d38586bacb03c588827aeea39c181c02c228 Mon Sep 17 00:00:00 2001 From: Andreas Boltres Date: Fri, 13 Nov 2020 08:45:02 +0100 Subject: [PATCH] Update README.md --- README.md | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index 9b2018e7..d800d19d 100644 --- a/README.md +++ b/README.md @@ -101,7 +101,8 @@ The _TacticSelector_ defined in `scripts/tactic_selector.py` provides a class wr ## PPO-GoalRGCN -I am currently implementing a PPO agent that uses a relational GCN (see [the DGL library](https://docs.dgl.ai/index.html)). See `agents/goal_rgcn_agent.py` for the WIP code. +I have implemented a PPO agent that uses a relational GCN (see [the DGL library](https://docs.dgl.ai/index.html)). See `agents/goal_rgcn_agent.py` for the code. +It is able to train on simple training sets, although it needs **a lot** of RAM (>= 16GB for small Graphs). Future work might tackle this issue or test the performance on bigger machines than the ones I have access to. # Attribution