Merge pull request #1361 from PrincetonUniversity/devel

Devel

.appveyor.yml

@@ -35,9 +35,8 @@ install:
   - pip install --user -U certifi
   - pip install --user git+https://github.com/benureau/leabra.git@master
 
-  # pytorch does not distribute windows packages over pypi.
-  # Install it directly, or remove from requirements if not available (win32).
-  - if "%ARCH%" == "" (findstr /V torch < dev_requirements.txt > tmp_req && move /Y tmp_req dev_requirements.txt) else (pip install --user torch -f https://download.pytorch.org/whl/cpu/torch_stable.html)
+  # pytorch does not distribute windows packages over pypi. Install it directly.
+  - if not "%ARCH%" == "" (pip install --user torch -f https://download.pytorch.org/whl/cpu/torch_stable.html)
 
   - pip install --user -e .[dev]
 

.travis.yml

@@ -5,53 +5,38 @@ branches:
     - /devel-.*/
     - /travis.*/
 
-language: python
+language: shell
 
-python:
-    - 3.6
-    - 3.7
-
-os: linux
-dist: xenial
+os:
+  - linux
+  - osx
+dist: bionic
 
 env:
+  matrix:
+    - PYTHON=3.6.8
+    - PYTHON=3.7.4
   global:
     - PYTHONWARNINGS="ignore::DeprecationWarning"
+    - PIP_PROGRESS_BAR="off"
 
 # Cache downloaded(built) python packages
+# and homebrew downloads
 cache:
-  pip
+  directories:
+    - $HOME/.cache/pip
+    - $HOME/Library/Caches/Homebrew
+    - $HOME/Library/Caches/pip
 
 addons:
   apt:
     packages:
       - graphviz
-
-matrix:
-  include:
-  - os: osx
-    python: 3.6
-    language: minimal
-    env: PYTHON=3.6.8
-    # Cache pip and homebrew downloads
-    cache:
-        directories:
-          - $HOME/Library/Caches/Homebrew
-          - $HOME/Library/Caches/pip
-
-  - os: osx
-    python: 3.7
-    language: minimal
-    env: PYTHON=3.7.4
-    # Cache pip and homebrew downloads
-    cache:
-        directories:
-          - $HOME/Library/Caches/Homebrew
-          - $HOME/Library/Caches/pip
+      - python3-pip
 
 before_install:
   - |
-    # OSX Python is not directly supported on travis
+    # Hombrew doesn't provide older python versions.
     # Install it manually from python.org
     if [ "$TRAVIS_OS_NAME" == "osx" ]; then
       # homebrew plugin is not working correctly
@@ -70,20 +55,23 @@ before_install:
       echo "Deploying new python venv"
       python3 -m pip  install virtualenv
       python3 -m venv $HOME/venv
-      source $HOME/venv/bin/activate
-
-      # Upgrade pip
-      pip install -U pip
+    elif [ "$TRAVIS_OS_NAME" == "linux" ]; then
+      sudo apt-get install -y python${PYTHON%.*}-dev python${PYTHON%.*}-venv
+      python${PYTHON%.*} -m venv $HOME/venv
+      # Provide fake xdg-open
+      echo "#!/bin/sh" > $HOME/venv/bin/xdg-open
+      chmod +x $HOME/venv/bin/xdg-open
     fi
+  - source $HOME/venv/bin/activate
+
+  # Upgrade pip
+  - pip install -U pip
   - python --version
   - pip --version
 
 install:
   - pip install coveralls
   - pip install git+https://github.com/benureau/leabra.git@master
-  # Travis bundles pytest 4.3.1 in Linux Xenial, new pytest-xdist
-  # requires pytest>=4.4.0, and the dependencies are broken
-  - pip install -U pytest
   - pip install -e .[dev]
 
 

CONVENTIONS.md

@@ -26,10 +26,10 @@ Made up of two types of classes:
         - **LearningProjection**
         - **ControlProjection**
         - **GatingProjection**
-- *State*
-    - **InputState**
-    - **ParameterState**
-    - **OutputState**
+- *Port*
+    - **InputPort**
+    - **ParameterPort**
+    - **OutputPort**
     - *ModulatorySignal*
         - **LearningSignal**
         - **ControlSignal**
@@ -65,7 +65,7 @@ Extensions of Core objects
 - Component names always end in their type (e.g., TransferMechanism, LearningProjection)
   (the only exception is the DDM)
 - Components and Compositions should *always* be referred to in caps
-  (e.g., All Mechanisms have Projections; the receiver for a Projection is an InputState; etc.).
+  (e.g., All Mechanisms have Projections; the receiver for a Projection is an InputPort; etc.).
 
 #### Format:
 - class names:
@@ -233,8 +233,8 @@ Terminology used here:
         the module must be explicitly referenced (e.g., `ControlMechanism <ControlMechanism>`);
         [this appears to be redundant, but it is necessary]
     - conversely, for classes without subclasses, the title in the rst file is singular;
-        therefore, to refer to the plural of such a class (e.g., InputState),
-        the module must be explicitly referenced (e.g., `InputStates <InputState>`);
+        therefore, to refer to the plural of such a class (e.g., InputPort),
+        the module must be explicitly referenced (e.g., `InputStates <InputPort>`);
     - to flag references to sections that have not yet been documented (or labelled), 
         use the following construction: `section <LINK>` (so that <LINK> can be searched for replace these later).
 

Scripts/Debug/Gilbert_Shallice_debugging_Interactive_activation

@@ -21,7 +21,7 @@ WORD_OUTPUT_LAYER = pnl.RecurrentTransferMechanism(size = 3,
                                                    # integrator_function= pnl.InteractiveActivation(rate = 0.0015, decay=0.0, offset=-6),
                                                    name='WORD OUTPUT LAYER')
 WORD_OUTPUT_LAYER.set_log_conditions('value')
-WORD_OUTPUT_LAYER.set_log_conditions('InputState-0')
+WORD_OUTPUT_LAYER.set_log_conditions('InputPort-0')
 
 
 

Scripts/Debug/Markus Stroop.py

@@ -201,8 +201,8 @@ def trial_dict(red_color, green_color, red_word, green_word, CN, WR):
 #   CREATE THRESHOLD FUNCTION
 # first value of DDM's value is DECISION_VARIABLE
 def pass_threshold(mech1, mech2, thresh):
-    results1 = mech1.output_states[0].value
-    results2 = mech2.output_states[0].value
+    results1 = mech1.output_ports[0].value
+    results2 = mech2.output_ports[0].value
     for val in results1:
         if val >= thresh:
             return True

Scripts/Debug/Predator-Prey Sebastian.py

@@ -88,7 +88,7 @@ def new_episode():
     initial_observation = ddqn_agent.env.reset()
     new_episode_flag = True
 
-    # Initialize both states to verdical state based on first observation
+    # Initialize both ports to verdical state based on first observation
     perceptual_state = veridical_state = ddqn_agent.buffer.next(initial_observation, is_new_episode=True)
 
 def get_optimal_action(observation):
@@ -143,7 +143,7 @@ def get_action(variable=[[0,0],[0,0],[0,0]]):
 #    note: unitization is done in main loop, to allow compilation of LinearCombination function in ObjectiveMech) (TBI)
 action_mech = ProcessingMechanism(default_variable=[[0,0],[0,0],[0,0]],
                                   function=get_action, name='ACTION',
-                                  output_states='agent action')
+                                  output_ports='agent action')
 
 # ************************************** BASIC COMPOSITION *************************************************************
 
@@ -156,9 +156,9 @@ def get_action(variable=[[0,0],[0,0],[0,0]]):
 
 agent_comp.add_node(action_mech, required_roles=[NodeRole.OUTPUT])
 
-a = MappingProjection(sender=player_percept, receiver=action_mech.input_states[0])
-b = MappingProjection(sender=predator_percept, receiver=action_mech.input_states[1])
-c = MappingProjection(sender=prey_percept, receiver=action_mech.input_states[2])
+a = MappingProjection(sender=player_percept, receiver=action_mech.input_ports[0])
+b = MappingProjection(sender=predator_percept, receiver=action_mech.input_ports[1])
+c = MappingProjection(sender=prey_percept, receiver=action_mech.input_ports[2])
 agent_comp.add_projections([a,b,c])
 
 

Scripts/Debug/StabilityFlexibility.py

@@ -7,7 +7,7 @@
 # the associated accuracy of the trial will be the probability that the DDM hits the upper threshold
 def computeAccuracy(variable):
 
-    # variable is the list of values given by the monitored output states in the Objective Mechanism
+    # variable is the list of values given by the monitored output ports in the Objective Mechanism
 
     print("Inputs to ComputeAccuracy Function: ", variable)
 
@@ -79,7 +79,7 @@ def computeAccuracy(variable):
 inputLayer = pnl.TransferMechanism(#default_variable=[[0.0, 0.0]],
                                    size=2,
                                    function=pnl.Linear(slope=1, intercept=0),
-                                   output_states = [pnl.RESULT],
+                                   output_ports = [pnl.RESULT],
                                    name='Input')
 inputLayer.set_log_conditions([pnl.RESULT])
 
@@ -93,7 +93,7 @@ def computeAccuracy(variable):
                                             integrator_mode = True,
                                             integrator_function=pnl.AdaptiveIntegrator(rate=(tau)),
                                             initial_value=np.array([[0.0, 0.0]]),
-                                            output_states = [pnl.RESULT],
+                                            output_ports = [pnl.RESULT],
                                             name = 'Activity')
 
 activation.set_log_conditions([pnl.RESULT, "mod_gain"])
@@ -102,23 +102,23 @@ def computeAccuracy(variable):
 stimulusInfo = pnl.TransferMechanism(default_variable=[[0.0, 0.0]],
                                      size = 2,
                                      function = pnl.Linear(slope=1, intercept=0),
-                                     output_states = [pnl.RESULT],
+                                     output_ports = [pnl.RESULT],
                                      name = "Stimulus Info")
 
 stimulusInfo.set_log_conditions([pnl.RESULT])
 
 controlledElement = pnl.TransferMechanism(default_variable=[[0.0, 0.0]],
                                           size = 2,
                                           function=pnl.Linear(slope=1, intercept= 0),
-                                          input_states=pnl.InputState(combine=pnl.PRODUCT),
-                                          output_states = [pnl.RESULT],
+                                          input_ports=pnl.InputPort(combine=pnl.PRODUCT),
+                                          output_ports = [pnl.RESULT],
                                           name = 'Stimulus Info * Activity')
 
 controlledElement.set_log_conditions([pnl.RESULT])
 
 ddmCombination = pnl.TransferMechanism(size = 1,
                                        function = pnl.Linear(slope=1, intercept=0),
-                                       output_states = [pnl.RESULT],
+                                       output_ports = [pnl.RESULT],
                                        name = "DDM Integrator")
 ddmCombination.set_log_conditions([pnl.RESULT])
 
@@ -127,7 +127,7 @@ def computeAccuracy(variable):
                                                                  threshold = THRESHOLD,
                                                                  noise = NOISE,
                                                                  t0 = T0),
-                                                                 output_states = [pnl.DECISION_VARIABLE, pnl.RESPONSE_TIME,
+                                                                 output_ports = [pnl.DECISION_VARIABLE, pnl.RESPONSE_TIME,
                                                                                   pnl.PROBABILITY_UPPER_THRESHOLD, pnl.PROBABILITY_LOWER_THRESHOLD],
                                                                  name='DDM')
 
@@ -170,7 +170,7 @@ def computeAccuracy(variable):
                                                            function = computeAccuracy)
 
 meta_controller = pnl.OptimizationControlMechanism(agent_rep = stabilityFlexibility,
-                                                   features = [inputLayer.input_state,stimulusInfo.input_state],
+                                                   features = [inputLayer.input_port,stimulusInfo.input_port],
                                                    # features = {pnl.SHADOW_INPUTS: [inputLayer, stimulusInfo]},
                                                    # features = [(inputLayer, pnl.SHADOW_INPUTS),
                                                    #             (stimulusInfo, pnl.SHADOW_INPUTS)],

Scripts/Debug/Umemoto_Feb.py

@@ -42,7 +42,7 @@
 
 Target_Rep.set_log_conditions('value')#, log_condition=pnl.PROCESSING) # Log Target_Rep
 Target_Rep.set_log_conditions('mod_slope')#, log_condition=pnl.PROCESSING)
-Target_Rep.set_log_conditions('InputState-0')#, log_condition=pnl.PROCESSING)
+Target_Rep.set_log_conditions('InputPort-0')#, log_condition=pnl.PROCESSING)
 
 Distractor_Rep = pnl.TransferMechanism(name='Distractor Representation')
 
@@ -62,7 +62,7 @@
         starting_point=(x_0),
         t0=t0
     ),name='Decision',
-    output_states=[
+    output_ports=[
         pnl.DECISION_VARIABLE,
         pnl.RESPONSE_TIME,
         pnl.PROBABILITY_UPPER_THRESHOLD,
@@ -73,7 +73,7 @@
         }
     ],) #drift_rate=(1.0),threshold=(0.2645),noise=(0.5),starting_point=(0), t0=0.15
 
-Decision.set_log_conditions('InputState-0')#, log_condition=pnl.PROCESSING)
+Decision.set_log_conditions('InputPort-0')#, log_condition=pnl.PROCESSING)
 
 # Outcome Mechanisms:
 Reward = pnl.TransferMechanism(name='Reward')
@@ -119,10 +119,10 @@
                                                   allocation_samples=signalSearchRange)
 
 Umemoto_comp.add_model_based_optimizer(optimizer=pnl.OptimizationControlMechanism(agent_rep=Umemoto_comp,
-                                                                                  features=[Target_Stim.input_state, Distractor_Stim.input_state, Reward.input_state],
+                                                                                  features=[Target_Stim.input_port, Distractor_Stim.input_port, Reward.input_port],
                                                                                   feature_function=pnl.AdaptiveIntegrator(rate=1.0),
                                                                                   objective_mechanism=pnl.ObjectiveMechanism(monitor_for_control=[Reward,
-                                                                                                                                                 (Decision.output_states[pnl.PROBABILITY_UPPER_THRESHOLD], 1, -1)],
+                                                                                                                                                 (Decision.output_ports[pnl.PROBABILITY_UPPER_THRESHOLD], 1, -1)],
                                                                                                                              ),
                                                                                   function=pnl.GridSearch(),
                                                                                   control_signals=[Target_Rep_Control_Signal, Distractor_Rep_Control_Signal]

Scripts/Debug/Umemoto_Feb2.py

@@ -41,7 +41,7 @@
 
 Target_Rep.set_log_conditions('value')#, log_condition=pnl.PROCESSING) # Log Target_Rep
 Target_Rep.set_log_conditions('mod_slope')#, log_condition=pnl.PROCESSING)
-Target_Rep.set_log_conditions('InputState-0')#, log_condition=pnl.PROCESSING)
+Target_Rep.set_log_conditions('InputPort-0')#, log_condition=pnl.PROCESSING)
 
 Distractor_Rep = pnl.TransferMechanism(name='Distractor Representation')
 
@@ -64,7 +64,7 @@
         starting_point=(x_0),
         t0=t0
     ),name='Decision',
-    output_states=[
+    output_ports=[
         pnl.DECISION_VARIABLE,
         pnl.RESPONSE_TIME,
         pnl.PROBABILITY_UPPER_THRESHOLD,
@@ -75,7 +75,7 @@
         }
     ],) #drift_rate=(1.0),threshold=(0.2645),noise=(0.5),starting_point=(0), t0=0.15
 
-Decision.set_log_conditions('InputState-0')#, log_condition=pnl.PROCESSING)
+Decision.set_log_conditions('InputPort-0')#, log_condition=pnl.PROCESSING)
 Decision.set_log_conditions('PROBABILITY_UPPER_THRESHOLD')
 print(Decision.loggable_items)
 # Outcome Mechanisms:
@@ -129,13 +129,13 @@
 
 Umemoto_comp.add_model_based_optimizer(optimizer=pnl.OptimizationControlMechanism(
         agent_rep=Umemoto_comp,
-        features=[Target_Stim.input_state,
-                  Distractor_Stim.input_state,
-                  Reward.input_state],
+        features=[Target_Stim.input_port,
+                  Distractor_Stim.input_port,
+                  Reward.input_port],
         feature_function=pnl.AdaptiveIntegrator(rate=1.0),
         objective_mechanism=pnl.ObjectiveMechanism(
                 monitor_for_control=[Reward,
-                                     (Decision.output_states[pnl.PROBABILITY_UPPER_THRESHOLD], 1, -1)],
+                                     (Decision.output_ports[pnl.PROBABILITY_UPPER_THRESHOLD], 1, -1)],
         ),
         function=pnl.GridSearch(save_values=True),
         control_signals=[Target_Rep_Control_Signal,