Debugging CoordinateEncoder issues in NuPIC 1.0

Amitai · August 17, 2017, 4:54pm

@rhyolight, @scott, you can see the very simple code above, and here is a clearer trace (focusing on nupic libraries):

Traceback (most recent call last):
  File "C:\Program Files\JetBrains\PyCharm Community Edition 2017.2.1\helpers\pydev\_pydevd_bundle\pydevd_exec.py", line 3, in Exec
    exec exp in global_vars, local_vars
  File "<input>", line 1, in <module>
  File "C:\Users\amitai.perlstein\AppData\Local\Continuum\Anaconda2\envs\htm_101\lib\site-packages\nupic\frameworks\opf\htm_prediction_model.py", line 429, in run
    self._sensorCompute(inputRecord)
  File "C:\Users\amitai.perlstein\AppData\Local\Continuum\Anaconda2\envs\htm_101\lib\site-packages\nupic\frameworks\opf\htm_prediction_model.py", line 514, in _sensorCompute
    sensor.compute()
  File "C:\Users\amitai.perlstein\AppData\Local\Continuum\Anaconda2\envs\htm_101\lib\site-packages\nupic\engine\__init__.py", line 433, in compute
    return self._region.compute()
  File "C:\Users\amitai.perlstein\AppData\Local\Continuum\Anaconda2\envs\htm_101\lib\site-packages\nupic\bindings\engine_internal.py", line 1856, in compute
    return _engine_internal.Region_compute(self)
  File "C:\Users\amitai.perlstein\AppData\Local\Continuum\Anaconda2\envs\htm_101\lib\site-packages\nupic\bindings\regions\PyRegion.py", line 184, in guardedCompute
    return self.compute(inputs, DictReadOnlyWrapper(outputs))
  File "C:\Users\amitai.perlstein\AppData\Local\Continuum\Anaconda2\envs\htm_101\lib\site-packages\nupic\regions\record_sensor.py", line 401, in compute
    outputs["bucketIdxOut"][:] = encoder.getBucketIndices(actualValue)
  File "C:\Users\amitai.perlstein\AppData\Local\Continuum\Anaconda2\envs\htm_101\lib\site-packages\nupic\encoders\base.py", line 351, in getBucketIndices
    assert False, "Should be implemented in base classes that are not " \
AssertionError: Should be implemented in base classes that are not containers for other encoders

rhyolight · August 18, 2017, 12:32am

@Amitai Your predicted field must be called vector for this to work:

github.com

numenta/nupic/blob/1.0.1/src/nupic/regions/record_sensor.py#L383


sequenceId = data["_sequenceId"]
categories = data["_category"]


# Encode the processed records; populate outputs["dataOut"] in place
self.encoder.encodeIntoArray(data, outputs["dataOut"])


# If there is a field to predict, set bucketIdxOut and actValueOut.
# There is a special case where a predicted field might be a vector, as in
# the CoordinateEncoder. Since this encoder does not provide bucket
# indices for prediction, we will ignore it.
if self.predictedField is not None and self.predictedField != "vector":
  allEncoders = list(self.encoder.encoders)
  if self.disabledEncoder is not None:
    allEncoders.extend(self.disabledEncoder.encoders)
  encoders = [e for e in allEncoders
              if e[0] == self.predictedField]
  if len(encoders) == 0:
    raise ValueError("There is no encoder for set for the predicted "
                     "field: %s" % self.predictedField)
  # TODO: Figure out why there are sometimes multiple encoders with the
  # same name.

Amitai · August 21, 2017, 10:41am

Thanks Matt,
setting predictedField = ‘vector’ sorted it out.
This is a subtle point. It isn’t so intuitive that a name HAS to be a specific sting in order for it to work.

rhyolight · August 21, 2017, 3:53pm

No, it is not. The CoordinateEncoder certainly feels like a special case in the OPF, like it was shoehorned in. Because it was!

Topic		Replies	Views
Example usage of Coordinate Encoder NuPIC	52	6375	August 17, 2017
Multidimensional encodings using Java API HTM.Java	34	1587	January 16, 2019
NuPIC 1.0.1 bugfix release NuPIC release	2	630	August 8, 2017
Inconsistency between SpatialPooler in nupic and nupic.bindings NuPIC	17	1644	January 17, 2018
Help with a basic python error NuPIC newbie	23	1857	August 30, 2018

Debugging CoordinateEncoder issues in NuPIC 1.0

Related topics