Python Testing & Image Inference Improvements (Unity-Technologies#353)

* Reorganized python tests into separate folder, and make individiual test files for different (sub) modules.
* Add tests for trainer_controller, PPO, and behavioral cloning. More to come soon.
* Minor bug fixes discovered while writing tests.
* Reworked GirdWorld to reset much faster.
* Cleaned ObservationToTex and reworked GetObservationMatrixList to be 3x faster.

python/tests/__init__.py

@@ -0,0 +1,2 @@
+from unityagents import *
+from unitytrainers import *

python/tests/test_bc.py

@@ -0,0 +1,103 @@
+import mock
+import pytest
+
+import numpy as np
+import tensorflow as tf
+
+from unitytrainers.bc.models import BehavioralCloningModel
+from unityagents import UnityEnvironment
+
+
+def test_cc_bc_model():
+    c_action_c_state_start = '''{
+      "AcademyName": "RealFakeAcademy",
+      "resetParameters": {},
+      "brainNames": ["RealFakeBrain"],
+      "externalBrainNames": ["RealFakeBrain"],
+      "logPath":"RealFakePath",
+      "apiNumber":"API-2",
+      "brainParameters": [{
+          "stateSize": 3,
+          "stackedStates": 2,
+          "actionSize": 2,
+          "memorySize": 0,
+          "cameraResolutions": [],
+          "actionDescriptions": ["",""],
+          "actionSpaceType": 1,
+          "stateSpaceType": 1
+          }]
+    }'''.encode()
+
+    tf.reset_default_graph()
+    with mock.patch('subprocess.Popen'):
+        with mock.patch('socket.socket') as mock_socket:
+            with mock.patch('glob.glob') as mock_glob:
+                # End of mock
+                with tf.Session() as sess:
+                    with tf.variable_scope("FakeGraphScope"):
+                        mock_glob.return_value = ['FakeLaunchPath']
+                        mock_socket.return_value.accept.return_value = (mock_socket, 0)
+                        mock_socket.recv.return_value.decode.return_value = c_action_c_state_start
+                        env = UnityEnvironment(' ')
+
+                        model = BehavioralCloningModel(env.brains["RealFakeBrain"])
+                        init = tf.global_variables_initializer()
+                        sess.run(init)
+
+                        run_list = [model.sample_action, model.policy]
+                        feed_dict = {model.batch_size: 2,
+                                     model.sequence_length: 1,
+                                     model.state_in: np.array([[1, 2, 3, 1, 2, 3],
+                                                               [3, 4, 5, 3, 4, 5]])}
+                        sess.run(run_list, feed_dict=feed_dict)
+                        env.close()
+
+
+def test_dc_bc_model():
+    d_action_c_state_start = '''{
+      "AcademyName": "RealFakeAcademy",
+      "resetParameters": {},
+      "brainNames": ["RealFakeBrain"],
+      "externalBrainNames": ["RealFakeBrain"],
+      "logPath":"RealFakePath",
+      "apiNumber":"API-2",
+      "brainParameters": [{
+          "stateSize": 3,
+          "stackedStates": 2,
+          "actionSize": 2,
+          "memorySize": 0,
+          "cameraResolutions": [{"width":30,"height":40,"blackAndWhite":false}],
+          "actionDescriptions": ["",""],
+          "actionSpaceType": 0,
+          "stateSpaceType": 1
+          }]
+    }'''.encode()
+
+    tf.reset_default_graph()
+    with mock.patch('subprocess.Popen'):
+        with mock.patch('socket.socket') as mock_socket:
+            with mock.patch('glob.glob') as mock_glob:
+                with tf.Session() as sess:
+                    with tf.variable_scope("FakeGraphScope"):
+                        mock_glob.return_value = ['FakeLaunchPath']
+                        mock_socket.return_value.accept.return_value = (mock_socket, 0)
+                        mock_socket.recv.return_value.decode.return_value = d_action_c_state_start
+                        env = UnityEnvironment(' ')
+
+                        model = BehavioralCloningModel(env.brains["RealFakeBrain"])
+                        init = tf.global_variables_initializer()
+                        sess.run(init)
+
+                        run_list = [model.sample_action, model.policy]
+                        feed_dict = {model.batch_size: 2,
+                                     model.dropout_rate: 1.0,
+                                     model.sequence_length: 1,
+                                     model.state_in: np.array([[1, 2, 3, 1, 2, 3],
+                                                               [3, 4, 5, 3, 4, 5]]),
+                                     model.observation_in[0]: np.ones([2, 40, 30, 3])}
+                        sess.run(run_list, feed_dict=feed_dict)
+                        env.close()
+
+
+if __name__ == '__main__':
+    pytest.main()

python/tests/test_ppo.py

@@ -0,0 +1,105 @@
+import mock
+import pytest
+
+import numpy as np
+import tensorflow as tf
+
+from unitytrainers.ppo.models import PPOModel
+from unityagents import UnityEnvironment
+
+
+def test_ppo_model_continuous():
+    c_action_c_state_start = '''{
+      "AcademyName": "RealFakeAcademy",
+      "resetParameters": {},
+      "brainNames": ["RealFakeBrain"],
+      "externalBrainNames": ["RealFakeBrain"],
+      "logPath":"RealFakePath",
+      "apiNumber":"API-2",
+      "brainParameters": [{
+          "stateSize": 3,
+          "stackedStates": 2,
+          "actionSize": 2,
+          "memorySize": 0,
+          "cameraResolutions": [],
+          "actionDescriptions": ["",""],
+          "actionSpaceType": 1,
+          "stateSpaceType": 1
+          }]
+    }'''.encode()
+
+    tf.reset_default_graph()
+    with mock.patch('subprocess.Popen'):
+        with mock.patch('socket.socket') as mock_socket:
+            with mock.patch('glob.glob') as mock_glob:
+                # End of mock
+                with tf.Session() as sess:
+                    with tf.variable_scope("FakeGraphScope"):
+                        mock_glob.return_value = ['FakeLaunchPath']
+                        mock_socket.return_value.accept.return_value = (mock_socket, 0)
+                        mock_socket.recv.return_value.decode.return_value = c_action_c_state_start
+                        env = UnityEnvironment(' ')
+
+                        model = PPOModel(env.brains["RealFakeBrain"])
+                        init = tf.global_variables_initializer()
+                        sess.run(init)
+
+                        run_list = [model.output, model.probs, model.value, model.entropy,
+                                    model.learning_rate]
+                        feed_dict = {model.batch_size: 2,
+                                     model.sequence_length: 1,
+                                     model.state_in: np.array([[1, 2, 3, 1, 2, 3],
+                                                               [3, 4, 5, 3, 4, 5]])}
+                        sess.run(run_list, feed_dict=feed_dict)
+                        env.close()
+
+
+def test_ppo_model_discrete():
+    d_action_c_state_start = '''{
+      "AcademyName": "RealFakeAcademy",
+      "resetParameters": {},
+      "brainNames": ["RealFakeBrain"],
+      "externalBrainNames": ["RealFakeBrain"],
+      "logPath":"RealFakePath",
+      "apiNumber":"API-2",
+      "brainParameters": [{
+          "stateSize": 3,
+          "stackedStates": 2,
+          "actionSize": 2,
+          "memorySize": 0,
+          "cameraResolutions": [{"width":30,"height":40,"blackAndWhite":false}],
+          "actionDescriptions": ["",""],
+          "actionSpaceType": 0,
+          "stateSpaceType": 1
+          }]
+    }'''.encode()
+
+    tf.reset_default_graph()
+    with mock.patch('subprocess.Popen'):
+        with mock.patch('socket.socket') as mock_socket:
+            with mock.patch('glob.glob') as mock_glob:
+                # End of mock
+                with tf.Session() as sess:
+                    with tf.variable_scope("FakeGraphScope"):
+                        mock_glob.return_value = ['FakeLaunchPath']
+                        mock_socket.return_value.accept.return_value = (mock_socket, 0)
+                        mock_socket.recv.return_value.decode.return_value = d_action_c_state_start
+                        env = UnityEnvironment(' ')
+                        model = PPOModel(env.brains["RealFakeBrain"])
+                        init = tf.global_variables_initializer()
+                        sess.run(init)
+
+                        run_list = [model.output, model.all_probs, model.value, model.entropy,
+                                    model.learning_rate]
+                        feed_dict = {model.batch_size: 2,
+                                     model.sequence_length: 1,
+                                     model.state_in: np.array([[1, 2, 3, 1, 2, 3],
+                                                               [3, 4, 5, 3, 4, 5]]),
+                                     model.observation_in[0]: np.ones([2, 40, 30, 3])
+                                     }
+                        sess.run(run_list, feed_dict=feed_dict)
+                        env.close()
+
+
+if __name__ == '__main__':
+    pytest.main()