Rename examples directories

examples/README.md

@@ -1,8 +1,8 @@
 # Examples
 
-- `ale`: examples for general Atari games
-- `atari`: examples with benchmark scores for reproducing published results on Atari
+- `atari`: examples for general Atari games
+- `atari/reproduction`: examples with benchmark scores for reproducing published results on Atari
 - `gym`: examples for OpenAI Gym environments
 - `grasping`: examples for a Bullet-based robotic grasping environment
-- `mujoco`: examples with benchmark scores for reproducing published results on MuJoCo tasks
+- `mujoco/reproduction`: examples with benchmark scores for reproducing published results on MuJoCo tasks
 - `quickstart`: a quickstart guide of ChainerRL

examples_tests/atari/reproduction/test_dqn.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# atari/reproduction/dqn
+python examples/atari/reproduction/dqn/train_dqn.py --env PongNoFrameskip-v4 --steps 100 --replay-start-size 50 --outdir $outdir/atari/reproduction/dqn --eval-n-steps 200 --eval-interval 50 --n-best-episodes 1 --gpu $gpu
+model=$(find $outdir/atari/reproduction/dqn -name "*_finish")
+python examples/atari/reproduction/dqn/train_dqn.py --env PongNoFrameskip-v4 --demo --load $model --outdir $outdir/temp --eval-n-steps 200 --gpu $gpu

examples_tests/atari/reproduction/test_iqn.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# atari/reproduction/iqn
+python examples/atari/reproduction/iqn/train_iqn.py --env PongNoFrameskip-v4 --steps 100 --replay-start-size 50 --outdir $outdir/atari/reproduction/iqn --eval-n-steps 200 --eval-interval 50 --n-best-episodes 1  --gpu $gpu
+model=$(find $outdir/atari/reproduction/iqn -name "*_finish")
+python examples/atari/reproduction/iqn/train_iqn.py --env PongNoFrameskip-v4 --demo --load $model --outdir $outdir/temp --eval-n-steps 200 --gpu $gpu

examples_tests/atari/reproduction/test_rainbow.sh

@@ -0,0 +1,13 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+echo "outdir: $outdir"
+
+gpu="$1"
+
+# atari/reproduction/rainbow
+python examples/atari/reproduction/rainbow/train_rainbow.py --env PongNoFrameskip-v4 --steps 100 --replay-start-size 50 --outdir $outdir/atari/reproduction/rainbow --eval-n-steps 200 --eval-interval 50 --n-best-episodes 1 --gpu $gpu
+model=$(find $outdir/atari/reproduction/rainbow -name "*_finish")
+python examples/atari/reproduction/rainbow/train_rainbow.py --env PongNoFrameskip-v4 --demo --load $model --outdir $outdir/temp --eval-n-steps 200 --gpu $gpu

examples_tests/atari/test_a2c.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# atari/a2c
+python examples/atari/train_a2c_ale.py --env PongNoFrameskip-v4 --steps 100 --update-steps 50 --outdir $outdir/atari/a2c
+model=$(find $outdir/atari/a2c -name "*_finish")
+python examples/atari/train_a2c_ale.py --env PongNoFrameskip-v4 --demo --load $model --eval-n-runs 1

examples_tests/atari/test_a3c.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# atari/a3c (only for cpu)
+if [[ $gpu -lt 0 ]]; then
+  python examples/atari/train_a3c_ale.py 4 --env PongNoFrameskip-v4 --steps 100 --outdir $outdir/atari/a3c
+  model=$(find $outdir/atari/a3c -name "*_finish")
+  python examples/atari/train_a3c_ale.py 4 --env PongNoFrameskip-v4 --demo --load $model --eval-n-runs 1 --outdir $outdir/temp
+fi

examples_tests/atari/test_acer.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# atari/acer (only for cpu)
+if [[ $gpu -lt 0 ]]; then
+  python examples/atari/train_acer_ale.py 4 --env PongNoFrameskip-v4 --steps 100 --outdir $outdir/atari/acer
+  model=$(find $outdir/atari/acer -name "*_finish")
+  python examples/atari/train_acer_ale.py 4 --env PongNoFrameskip-v4 --demo --load $model --eval-n-runs 1 --outdir $outdir/temp
+fi

examples_tests/atari/test_categorical_dqn.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# atari/categorical_dqn
+python examples/atari/train_categorical_dqn_ale.py --env PongNoFrameskip-v4 --steps 100 --replay-start-size 50 --outdir $outdir/atari/categorical_dqn --gpu $gpu
+model=$(find $outdir/atari/categorical_dqn -name "*_finish")
+python examples/atari/train_categorical_dqn_ale.py --env PongNoFrameskip-v4 --demo --load $model --eval-n-runs 1 --outdir $outdir/temp --gpu $gpu

examples_tests/atari/test_dqn.sh

@@ -7,6 +7,6 @@ outdir=$(mktemp -d)
 gpu="$1"
 
 # atari/dqn
-python examples/atari/dqn/train_dqn.py --env PongNoFrameskip-v4 --steps 100 --replay-start-size 50 --outdir $outdir/atari/dqn --eval-n-steps 200 --eval-interval 50 --n-best-episodes 1 --gpu $gpu
+python examples/atari/train_dqn_ale.py --env PongNoFrameskip-v4 --steps 100 --replay-start-size 50 --outdir $outdir/atari/dqn --gpu $gpu
 model=$(find $outdir/atari/dqn -name "*_finish")
-python examples/atari/dqn/train_dqn.py --env PongNoFrameskip-v4 --demo --load $model --outdir $outdir/temp --eval-n-steps 200 --gpu $gpu
+python examples/atari/train_dqn_ale.py --env PongNoFrameskip-v4 --demo --load $model --eval-n-runs 1 --outdir $outdir/temp --gpu $gpu

examples_tests/atari/test_dqn_batch.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# atari/dqn batch
+python examples/atari/train_dqn_batch_ale.py --env PongNoFrameskip-v4 --steps 100 --replay-start-size 50 --outdir $outdir/atari/dqn_batch --gpu $gpu
+model=$(find $outdir/atari/dqn_batch -name "*_finish")
+python examples/atari/train_dqn_batch_ale.py --env PongNoFrameskip-v4 --demo --load $model --eval-n-runs 1 --outdir $outdir/temp --gpu $gpu

examples_tests/atari/test_nsq.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# atari/nsq (only for cpu)
+if [[ $gpu -lt 0 ]]; then
+  python examples/atari/train_nsq_ale.py 4 --env PongNoFrameskip-v4 --steps 100 --outdir $outdir/atari/nsq
+  model=$(find $outdir/atari/nsq -name "*_finish")
+  python examples/atari/train_nsq_ale.py 4 --env PongNoFrameskip-v4 --demo --load $model --eval-n-runs 1 --outdir $outdir/temp
+fi

examples_tests/atari/test_ppo.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# atari/ppo
+python examples/atari/train_ppo_ale.py --env PongNoFrameskip-v4 --steps 100 --update-interval 50 --batchsize 16 --epochs 2 --outdir $outdir/atari/ppo --gpu $gpu
+model=$(find $outdir/atari/ppo -name "*_finish")
+python examples/atari/train_ppo_ale.py --env PongNoFrameskip-v4 --demo --load $model --eval-n-runs 1 --outdir $outdir/temp --gpu $gpu

examples_tests/mujoco/reproduction/test_ddpg.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# mujoco/reproduction/ddpg (specify non-mujoco env to test without mujoco)
+python examples/mujoco/reproduction/ddpg/train_ddpg.py --env Pendulum-v0 --gpu $gpu --steps 10 --replay-start-size 5 --batch-size 5 --outdir $outdir/mujoco/reproduction/ddpg
+model=$(find $outdir/mujoco/reproduction/ddpg -name "*_finish")
+python examples/mujoco/reproduction/ddpg/train_ddpg.py --env Pendulum-v0 --demo --load $model --eval-n-runs 1 --outdir $outdir/temp --gpu $gpu

examples_tests/mujoco/reproduction/test_ppo.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# mujoco/reproduction/ppo (specify non-mujoco env to test without mujoco)
+python examples/mujoco/reproduction/ppo/train_ppo.py --steps 10 --update-interval 5 --batch-size 5 --epochs 2 --outdir $outdir/mujoco/reproduction/ppo --env Pendulum-v0 --gpu $gpu
+model=$(find $outdir/mujoco/reproduction/ppo -name "*_finish")
+python examples/mujoco/reproduction/ppo/train_ppo.py --demo --load $model --eval-n-runs 1 --env Pendulum-v0 --outdir $outdir/temp --gpu $gpu

examples_tests/mujoco/reproduction/test_td3.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# mujoco/reproduction/td3 (specify non-mujoco env to test without mujoco)
+python examples/mujoco/reproduction/td3/train_td3.py --env Pendulum-v0 --gpu $gpu --steps 10 --replay-start-size 5 --batch-size 5 --outdir $outdir/mujoco/reproduction/td3
+model=$(find $outdir/mujoco/reproduction/td3 -name "*_finish")
+python examples/mujoco/reproduction/td3/train_td3.py --env Pendulum-v0 --demo --load $model --eval-n-runs 1 --outdir $outdir/temp --gpu $gpu