Refactor #15
Refactor #15
Conversation
…tributes to InputLayer
| for batch in self.dataloader: | ||
| batch_x = batch[0] if self.have_sensory else None | ||
| batch_loc = batch[self.have_sensory] if self.have_location else None | ||
| batch_y = batch[-1] if self.have_label else None |
There was a problem hiding this comment.
Isn't this better, might add other things to dataloader in near future
| batch_y = batch[-1] if self.have_label else None | |
| batch_y = batch[1] if self.have_label else None |
There was a problem hiding this comment.
No, The label should always be at the end. data loader can pass extra items before the label and after sensory and location data.
|
|
||
| if batch_y is not None: | ||
| batch_y = batch_y.to(self.device) | ||
| each_instance = num_instance // torch.numel(batch_y) |
There was a problem hiding this comment.
It is not clear what variable num_instance refers to
There was a problem hiding this comment.
it's the number of iterations a batch will last.
conex/nn/Modules/io.py
Outdated
| if LAYER_TIMESTAMPS["InputDataset"] not in behavior: | ||
| behavior[LAYER_TIMESTAMPS["InputDataset"]] = SpikeNdDataset( | ||
| dataloader=input_dataloader, | ||
| N_sensory=sensory_data_dim, |
There was a problem hiding this comment.
Some have changed in the new commit. any recommendation for 'N_sensory' and 'N_location' ?
There was a problem hiding this comment.
How about sensory_size and location_size?
There was a problem hiding this comment.
It's not the size of the population. It's N as N-Dimension. e.g. 2d image for the sensory population...
There was a problem hiding this comment.
Then if size makes confusion, we can use ndim
conex/nn/Modules/io.py
Outdated
|
|
||
| def __get_ng(self, net, size, tag, trace, setter, user_defined=None): | ||
| behavior = { | ||
| NEURON_TIMESTAMPS["Fire"]: setter(), |
There was a problem hiding this comment.
could we add a simple comment as a doc for future doc reference
| from conex.nn.Structure.Layers import Layer | ||
| from conex.nn.Modules.topological_connections import StructuredSynapseGroup | ||
|
|
||
| from pymonntorch import SynapseGroup, NeuronGroup, TaggableObject | ||
|
|
||
| # TODO: handle multi-scale | ||
| # TODO: str.removesuffix needs python 3.9 |
There was a problem hiding this comment.
update the project base config and force devs to use python 3.9+
Major changes:
makeorupdate_makeon default config classes to generate all and/or update some config keys, respectively.