From ad43fd9a35f3848f5c22aea7a91dc4021ca3f9fd Mon Sep 17 00:00:00 2001
From: Peter Zhokhov <peterz@openai.com>
Date: Wed, 1 Aug 2018 16:15:59 -0700
Subject: [PATCH] add defaults

---
 baselines/acer/defaults.py     |  4 ++++
 baselines/ppo2/defaults.py     | 22 ++++++++++++++++++++++
 baselines/trpo_mpi/defaults.py | 30 ++++++++++++++++++++++++++++++
 3 files changed, 56 insertions(+)
 create mode 100644 baselines/acer/defaults.py
 create mode 100644 baselines/ppo2/defaults.py
 create mode 100644 baselines/trpo_mpi/defaults.py

diff --git a/baselines/acer/defaults.py b/baselines/acer/defaults.py
new file mode 100644
index 0000000..0334bae
--- /dev/null
+++ b/baselines/acer/defaults.py
@@ -0,0 +1,4 @@
+def atari():
+    return dict(
+        lrschedule='constant'
+    )
diff --git a/baselines/ppo2/defaults.py b/baselines/ppo2/defaults.py
new file mode 100644
index 0000000..f8c23a5
--- /dev/null
+++ b/baselines/ppo2/defaults.py
@@ -0,0 +1,22 @@
+def mujoco():
+    return dict(
+        nsteps=2048,
+        nminibatches=32,
+        lam=0.95,
+        gamma=0.99,
+        noptepochs=10,
+        log_interval=1,
+        ent_coef=0.0,
+        lr=3e-4,
+        cliprange=0.2,
+        value_network='copy'
+    )
+
+def atari():
+    return dict(
+        nsteps=128, nminibatches=4,
+        lam=0.95, gamma=0.99, noptepochs=4, log_interval=1,
+        ent_coef=.01,
+        lr=lambda f : f * 2.5e-4,
+        cliprange=lambda f : f * 0.1,
+    )
diff --git a/baselines/trpo_mpi/defaults.py b/baselines/trpo_mpi/defaults.py
new file mode 100644
index 0000000..96b6cb3
--- /dev/null
+++ b/baselines/trpo_mpi/defaults.py
@@ -0,0 +1,30 @@
+from rl_common.models import mlp, cnn_small
+
+
+def atari():
+    return dict(
+        network = cnn_small(),
+        timesteps_per_batch=512, 
+        max_kl=0.001,
+        cg_iters=10,
+        cg_damping=1e-3,
+        gamma=0.98,
+        lam=1.0,
+        vf_iters=3,
+        vf_stepsize=1e-4,
+        entcoeff=0.00,
+    )
+
+def mujoco():
+    return dict(
+        network = mlp(num_hidden=32, num_layers=2),
+        timesteps_per_batch=1024,
+        max_kl=0.01,
+        cg_iters=10,
+        cg_damping=0.1,
+        gamma=0.99,
+        lam=0.98,
+        vf_iters=5,
+        vf_stepsize=1e-3,
+        normalize_observations=True, 
+    )