From 3ec2f01c776561d6e15365b05c72ac94908b51ff Mon Sep 17 00:00:00 2001 From: Laura Graesser Date: Sun, 23 Sep 2018 00:23:27 -0700 Subject: [PATCH] Benchmark result for A2C cartpole (#180) * add benchmark spec files * make dqn polyak * specify resources * update demo spec * Updating a2c gae specs * Update max cpus for a2c cartpole * speedrun * lower trial * try save deepcopy of spec and info_space * lower trial * Updated a2c mlp specs * Switch nstep separate to linear decay * ramp up trial. problem confirmed fixed * update lunar a2c search specs * cleanup cartpole and lunar benchmark specs * update demo spec to good params * placeholder for experiment link --- CONTRIBUTING.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md index 02423c4d7..60ca721bb 100644 --- a/CONTRIBUTING.md +++ b/CONTRIBUTING.md @@ -20,7 +20,7 @@ Reproducibility is crucial to deep RL, thus it is very valuable to have multiple 1. a spec file 2. a git SHA: check out the SLM Lab to the appropriate branch or commit -An example experiment is [here](todo) +An example experiment is [here](todo-pending) >The only aspect of experiments that are not reproducible are the random seeds. This is by design. Good algorithms and hyper-parameter settings should be reasonably stable across different random seeds.