Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Devel #1399

Merged
merged 43 commits into from
Nov 18, 2019
Merged

Devel #1399

merged 43 commits into from
Nov 18, 2019

Conversation

dillontsmith
Copy link
Contributor

No description provided.

jvesely and others added 30 commits November 8, 2019 01:04
LLVM IR doesn't have void* type.
Signed-off-by: Jan Vesely <[email protected]>
The new name better captures function semantics.

Signed-off-by: Jan Vesely <[email protected]>
Don't call the ir conversion routine twice on the same argument.
Signed-off-by: Jan Vesely <[email protected]>
The ctypes data types should match.
Signed-off-by: Jan Vesely <[email protected]>
Use cfunc argument types directly instead of relying on byref_arg_types.
Signed-off-by: Jan Vesely <[email protected]>
Signed-off-by: Jan Vesely <[email protected]>
Everything can be done in __init__.
Signed-off-by: Jan Vesely <[email protected]>
* • TRANSFER_OUTPUT:
  - automated assignment of attributes from Keywords.OUTPUTS

• RecurrentTransferMechanism
  - renamed definition of local ENERGY to ENERGY_OUTPUT_PORT
  - renamed definition of local ENTROPY to ENTROPY_OUTPUT_PORT

* • Mechanism_Base
  - added standard_output_ports and standard_output_port_names

• TransferMechanism
  - standard_output_ports and standard_output_port_names now inherited

* • Project
  Names of standard_output_states: OUTPUT_<*> -> <*>

• OutputPort
  - _instantiate_output_port: added instantiation of owner.standard_output_ports

* • Mechanism subclasses:
  - eliminated instantiation of owner.standard_output_ports
    (moved to OutputPort._instantiate_output_port

* • ProcessingMechanism, Mechanism_Base
  - moved definition of standard_output_ports from ProcessingMechanism to Mechanism_Base

* • TransferMechanisms
  - If TransferMechanism has > 1 InputPort, RESULT renamed as RESULT-0

* -

* • Merged with devel

* • Project
  RESULTS -> RESULT

* • Project
  RESULTS -> RESULT

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* docs styling: fixed various spacing issues

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* • ContrastiveHebbianMechanism
  - fixed bug in which target_size=0 caused crash

* =

* • ComparatorMechanism
  - _validate_params: fixed bug causing crash if sample and/or target input_port not specified

* -

* -

* -

* • DDM
  removed DDM_OUTPUT;  replaced with DDM.standard_output_ports
  NOTE:  all prior references to pnl.DDM_OUTPUT.<OUTPUT_PORT_NAME>
         can now use use just pnl.<<OUTPUT_PORT_NAME>

* -

* -

* -

* -

* • TransferMechanism
  - added COMBINE to standard_output_ports

* • tests
  - test_transfer_mechanism
    added test_combine_standard_output_port
  - test_processing_mechanism
    added tests for standard_ouput_ports

* • SelectionFunctions
  - OneHot:  fixed bug in *_ABS_VAL value modes

• ProcessingMechanism
  - modified MAX_VAL and MAX_ABS_VAL standard_output_ports
    to return scalar rather than OneHot values
  - added MAX_ONE_HOT and MAX_ABS_ONE_HOT standard_output_ports

* • SelectionFunctions
  - OneHot:  fixed bug in *_ABS_VAL value modes

• ProcessingMechanism
  - modified MAX_VAL and MAX_ABS_VAL standard_output_ports
    to return scalar rather than OneHot values
  - added MAX_ONE_HOT and MAX_ABS_ONE_HOT standard_output_ports
Signed-off-by: Jan Vesely <[email protected]>
This addresses two problems:
1.) Copies of RandomState continue generating the same random sequence
based on the seed of the original. This makes all copies of original
Function with random state produce the same random sequence.

2.) Global instances of Functions are initialized (and seeded) before
any access to global seed can be made. This makes them use the default
time-based seed. Reseeding the copies makes them follow global seed
setting and thus more amenable for testing.

Signed-off-by: Jan Vesely <[email protected]>
Add typing information.
Simplify generating of ctypes.cfunctype arguments.
Codestyle.
Python3.8 complains about this.
Signed-off-by: Jan Vesely <[email protected]>
Python3.8 raises TypeError in check whether None is a valid enum value:
def __contains__(cls, member):
    if not isinstance(member, Enum):
>       raise TypeError(
            "unsupported operand type(s) for 'in': '%s' and '%s'" % (
                type(member).__qualname__, cls.__class__.__qualname__))
E       TypeError: unsupported operand type(s) for 'in': 'NoneType' and 'EnumMeta'
Signed-off-by: Jan Vesely <[email protected]>
Ubuntu only for now
The former is just an alias.
Signed-off-by: Jan Vesely <[email protected]>
…-system-comp

Convert several tests from System to Composition
jvesely and others added 13 commits November 15, 2019 14:26
Convert to ms before casting to integer.
Fixes random seed initialization on macos.
Add regression test.

Signed-off-by: Jan Vesely <[email protected]>
This is now handled in github-actions
Signed-off-by: Jan Vesely <[email protected]>
Allow ci- prefixed branches in Appveyor and Travis.
Restrict Appveyor to 32-bit builds.
@coveralls
Copy link

Coverage Status

Coverage decreased (-0.04%) to 78.321% when pulling 332c260 on devel into 0e564cd on master.

@dillontsmith dillontsmith merged commit 83772c8 into master Nov 18, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants