Error running hierarchy_network_demo.py

[Containerhouse]

Running the stock file hierarchy_network_demo.py (nupic/examples/network) the following error is thrown before it is able to write the .csv:

Writing output to: hierarchy-demo-output.csv
Traceback (most recent call last):
  File "hierarchy_network_demo.py", line 368, in <module>
    runDemo()
  File "hierarchy_network_demo.py", line 362, in runDemo
    runNetwork(network, numRecords, writer)
  File "hierarchy_network_demo.py", line 305, in runNetwork
    network.run(1)
  File "/usr/local/lib/python2.7/dist-packages/nupic-0.5.5.dev0-py2.7.egg/nupic/engine/__init__.py", line 693, in run
    engine.Network.run(self, n)
  File "/usr/local/lib/python2.7/dist-packages/nupic.bindings-0.4.5.dev0-py2.7-linux-x86_64.egg/nupic/bindings/engine_internal.py", line 1147, in run
    return _engine_internal.Network_run(self, *args, **kwargs)
  File "/usr/local/lib/python2.7/dist-packages/nupic.bindings-0.4.5.dev0-py2.7-linux-x86_64.egg/nupic/bindings/regions/PyRegion.py", line 178, in guardedCompute
    return self.compute(inputs, DictReadOnlyWrapper(outputs))
  File "/usr/local/lib/python2.7/dist-packages/nupic-0.5.5.dev0-py2.7.egg/nupic/regions/SDRClassifierRegion.py", line 352, in compute
    categories = [category for category in inputs["categoryIn"]
KeyError: 'categoryIn'

[rhyolight]

I can replicate this error. @scottpurdy can @andrewmalta13 look into this?

[scottpurdy]

@mrcslws can you take a look

[rhyolight]

@mrcslws Just told me that there is some history around this issue, but now I'm more confused. We probably just need to fix this demo so an error is not thrown, perhaps we don't need to crawl down that categoryIn rabbit hole.

[rhyolight]

I tried running the example by switching the params spatialImp & temporalImp to py instead of cpp, and it ran for a long while before having the same error.

However with cpp it crashed right away. Using py it was able to run a lot of records through before crashing. 🤷‍♂️

[scottpurdy]

There appear to be a couple issues. One is recent changes to do away with customCompute in the classifier region. I've made those changes and will push a new version of the hierarchy demo soon to close out this issue. The predictions will still be incorrect though until we address #3651