Merge pull request #472 from muupan/split-test-examples

Split test_examples.sh

examples_tests/ale/test_a2c.sh

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

examples_tests/ale/test_a3c.sh

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

examples_tests/ale/test_acer.sh

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

examples_tests/ale/test_categorical_dqn.sh

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

examples_tests/ale/test_dqn.sh

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

examples_tests/ale/test_dqn_batch.sh

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

examples_tests/ale/test_nsq.sh

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

examples_tests/ale/test_ppo.sh

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

examples_tests/atari/test_dqn.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+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 $gpu
+model=$(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

examples_tests/atari/test_iqn.sh

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

examples_tests/atari/test_rainbow.sh

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

examples_tests/grasping/test_dqn.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# grasping/dqn
+python examples/grasping/train_dqn_batch_grasping.py --gpu $gpu --steps 100 --outdir $outdir/grasping/dqn
+model=$(find $outdir/grasping/dqn -name "*_finish")
+python examples/grasping/train_dqn_batch_grasping.py --demo --load $model --eval-n-runs 1 --outdir $outdir/temp --gpu $gpu

examples_tests/gym/test_a2c.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# gym/a2c
+python examples/gym/train_a2c_gym.py --steps 100 --update-steps 50 --outdir $outdir/gym/a2c --gpu $gpu
+model=$(find $outdir/gym/a2c -name "*_finish")
+python examples/gym/train_a2c_gym.py --demo --load $model --eval-n-runs 1 --gpu $gpu

examples_tests/gym/test_a3c.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# gym/a3c
+python examples/gym/train_a3c_gym.py 4 --steps 100 --outdir $outdir/gym/a3c
+model=$(find $outdir/gym/a3c -name "*_finish")
+python examples/gym/train_a3c_gym.py 4 --demo --load $model --eval-n-runs 1 --outdir $outdir/temp

examples_tests/gym/test_acer.sh

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

examples_tests/gym/test_categorical_dqn.sh

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

examples_tests/gym/test_ddpg.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# gym/ddpg (specify non-mujoco env to test without mujoco)
+python examples/gym/train_ddpg_gym.py --steps 100 --replay-start-size 50 --minibatch-size 32 --outdir $outdir/gym/ddpg --env Pendulum-v0 --gpu $gpu
+model=$(find $outdir/gym/ddpg -name "*_finish")
+python examples/gym/train_ddpg_gym.py --demo --load $model --eval-n-runs 1 --env Pendulum-v0 --outdir $outdir/temp --gpu $gpu

examples_tests/gym/test_ddpg_batch.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# gym/ddpg batch (specify non-mujoco env to test without mujoco)
+python examples/gym/train_ddpg_batch_gym.py --steps 100 --replay-start-size 50 --minibatch-size 32 --outdir $outdir/gym/ddpg_batch --env Pendulum-v0 --gpu $gpu
+model=$(find $outdir/gym/ddpg_batch -name "*_finish")
+python examples/gym/train_ddpg_batch_gym.py --demo --load $model --eval-n-runs 1 --env Pendulum-v0 --outdir $outdir/temp --gpu $gpu

examples_tests/gym/test_dqn.sh

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

examples_tests/gym/test_iqn.sh

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

examples_tests/gym/test_pcl.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# gym/pcl
+python examples/gym/train_pcl_gym.py --steps 100 --batchsize 2 --replay-start-size 2 --outdir $outdir/gym/pcl --gpu $gpu
+model=$(find $outdir/gym/pcl -name "*_finish")
+python examples/gym/train_pcl_gym.py --demo --load $model --eval-n-runs 1 --outdir $outdir/temp --gpu $gpu

examples_tests/gym/test_ppo.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# gym/ppo (specify non-mujoco env to test without mujoco)
+python examples/gym/train_ppo_gym.py --steps 100 --update-interval 50 --batchsize 16 --epochs 2 --outdir $outdir/gym/ppo --env Pendulum-v0 --gpu $gpu
+model=$(find $outdir/gym/ppo -name "*_finish")
+python examples/gym/train_ppo_gym.py --demo --load $model --eval-n-runs 1 --env Pendulum-v0 --outdir $outdir/temp --gpu $gpu

examples_tests/gym/test_ppo_batch.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# gym/ppo batch (specify non-mujoco env to test without mujoco)
+python examples/gym/train_ppo_batch_gym.py --steps 100 --update-interval 50 --batchsize 16 --epochs 2 --outdir $outdir/gym/ppo_batch --env Pendulum-v0 --gpu $gpu
+model=$(find $outdir/gym/ppo_batch -name "*_finish")
+python examples/gym/train_ppo_batch_gym.py --demo --load $model --eval-n-runs 1 --env Pendulum-v0 --outdir $outdir/temp --gpu $gpu

examples_tests/gym/test_reinforce.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# gym/reinforce
+python examples/gym/train_reinforce_gym.py --steps 100 --batchsize 1 --outdir $outdir/gym/reinforce --gpu $gpu
+model=$(find $outdir/gym/reinforce -name "*_finish")
+python examples/gym/train_reinforce_gym.py --demo --load $model --eval-n-runs 1 --outdir $outdir/temp --gpu $gpu

examples_tests/gym/test_trpo.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+set -Ceu
+
+outdir=$(mktemp -d)
+
+gpu="$1"
+
+# gym/trpo (specify non-mujoco env to test without mujoco)
+python examples/gym/train_trpo_gym.py --steps 100 --trpo-update-interval 50 --outdir $outdir/gym/trpo --env Pendulum-v0 --gpu $gpu
+model=$(find $outdir/gym/trpo -name "*_finish")
+python examples/gym/train_trpo_gym.py --demo --load $model --eval-n-runs 1 --env Pendulum-v0 --outdir $outdir/temp --gpu $gpu

examples_tests/mujoco/test_ddpg.sh

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

examples_tests/mujoco/test_ppo.sh

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

examples_tests/mujoco/test_td3.sh

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