alan-stokes
commented
9 years ago issue is due to master pop table impl. If you recompile your code with 2darray master pop, it works perfectly.
To do so you ened to go to spynnaker.cfg and change [MasterPopTable]
algorithm: {2dArray, BinarySearch, HashTable}
generator = BinarySearch
generator = 2dArray
to
[MasterPopTable]
algorithm: {2dArray, BinarySearch, HashTable}
generator = BinarySearch
generator = 2dArray
and then go to spynnaker/neural_models/src/neuron/builds/makefile.common and change
POPULATION_TABLE_IMPL ?= fixed
POPULATION_TABLE_IMPL ?= binary_search
to POPULATION_TABLE_IMPL ?= fixed
POPULATION_TABLE_IMPL ?= binary_search
and recompile binaires and wal la.
this means the master pop binary search is a bit more buggered than we thought. Anna has some issues with get spikes becuase of it as well. So some core issue with that version of master pop
Alan
I just ported the vogels 2011 inhibitory plasticity rule across to the re-factored tools and I'm having problems getting the examples to work. I get one or other of the following errors when I try and get plastic weights:
File "vogels_2011.py", line 96, in
plastic_weights = plastic_ie_projection.getWeights(format="array")
File "/home/jamie/manchester/spinnaker_git/sPyNNaker/spynnaker/pyNN/models/pynn_projection.py", line 363, in getWeights
self._retrieve_synaptic_data_from_machine()
File "/home/jamie/manchester/spinnaker_git/sPyNNaker/spynnaker/pyNN/models/pynn_projection.py", line 439, in _retrieve_synaptic_data_from_machine
routing_infos=self._spinnaker.routing_infos)
File "/home/jamie/manchester/spinnaker_git/sPyNNaker/spynnaker/pyNN/models/neural_projections/projection_partitionable_edge.py", line 117, in get_synaptic_list_from_machine
routing_infos, subedge.weight_scales).get_rows()
File "/home/jamie/manchester/spinnaker_git/sPyNNaker/spynnaker/pyNN/models/abstract_models/abstract_synaptic_manager.py", line 664, in get_synaptic_list_from_machine
weight_scales)
File "/home/jamie/manchester/spinnaker_git/sPyNNaker/spynnaker/pyNN/models/abstract_models/abstract_synaptic_manager.py", line 692, in _translate_synaptic_block_from_memory
weight_scales)
File "/home/jamie/manchester/spinnaker_git/sPyNNaker/spynnaker/pyNN/models/neural_properties/synapse_dynamics/plastic_weight_synapse_row_io.py", line 133, in create_row_info_from_elements
"plastic synapses cannot create row ios from fixed entries.")
spynnaker.pyNN.exceptions.SynapticBlockGenerationException: plastic synapses cannot create row ios from fixed entries.
File "vogels_2011.py", line 96, in
plastic_weights = plastic_ie_projection.getWeights(format="array")
File "/home/jamie/manchester/spinnaker_git/sPyNNaker/spynnaker/pyNN/models/pynn_projection.py", line 363, in getWeights
self._retrieve_synaptic_data_from_machine()
File "/home/jamie/manchester/spinnaker_git/sPyNNaker/spynnaker/pyNN/models/pynn_projection.py", line 439, in _retrieve_synaptic_data_from_machine
routing_infos=self._spinnaker.routing_infos)
File "/home/jamie/manchester/spinnaker_git/sPyNNaker/spynnaker/pyNN/models/neural_projections/projection_partitionable_edge.py", line 117, in get_synaptic_list_from_machine
routing_infos, subedge.weight_scales).get_rows()
File "/home/jamie/manchester/spinnaker_git/sPyNNaker/spynnaker/pyNN/models/abstract_models/abstract_synaptic_manager.py", line 664, in get_synaptic_list_from_machine
weight_scales)
File "/home/jamie/manchester/spinnaker_git/sPyNNaker/spynnaker/pyNN/models/abstract_models/abstract_synaptic_manager.py", line 682, in _translate_synaptic_block_from_memory
position_in_block)
File "/home/jamie/manchester/spinnaker_git/sPyNNaker/spynnaker/pyNN/models/abstract_models/abstract_synaptic_manager.py", line 717, in _extract_row_data_from_memory_block
no_fixed_fixed = synaptic_block[position_in_block]
IndexError: index out of bounds
It's quite possible I have done something wrong as loads of stuff has changed, but it did used to work great on the branch from the last SpiNNaker workshop...