Merge pull request #501 from mahbodnr/master

Minor changes

bindsnet/learning/learning.py

@@ -1,5 +1,6 @@
 from abc import ABC
 from typing import Union, Optional, Sequence
+import warnings
 
 import torch
 import numpy as np
@@ -26,7 +27,7 @@ def __init__(
         nu: Optional[Union[float, Sequence[float]]] = None,
         reduction: Optional[callable] = None,
         weight_decay: float = 0.0,
-        **kwargs
+        **kwargs,
     ) -> None:
         # language=rst
         """
@@ -49,13 +50,19 @@ def __init__(
         # Learning rate(s).
         if nu is None:
             nu = [0.0, 0.0]
-        elif isinstance(nu, float) or isinstance(nu, int):
+        elif isinstance(nu, (float, int)):
             nu = [nu, nu]
 
         self.nu = torch.zeros(2, dtype=torch.float)
         self.nu[0] = nu[0]
         self.nu[1] = nu[1]
 
+        if (self.nu == torch.zeros(2)).all() and not isinstance(self, NoOp):
+            warnings.warn(
+                f"nu is set to [0., 0.] for {type(self).__name__} learning rule. "
+                + "It will disable the learning process."
+            )
+
         # Parameter update reduction across minibatch dimension.
         if reduction is None:
             if self.source.batch_size == 1:
@@ -96,7 +103,7 @@ def __init__(
         nu: Optional[Union[float, Sequence[float]]] = None,
         reduction: Optional[callable] = None,
         weight_decay: float = 0.0,
-        **kwargs
+        **kwargs,
     ) -> None:
         # language=rst
         """
@@ -113,7 +120,7 @@ def __init__(
             nu=nu,
             reduction=reduction,
             weight_decay=weight_decay,
-            **kwargs
+            **kwargs,
         )
 
     def update(self, **kwargs) -> None:
@@ -137,7 +144,7 @@ def __init__(
         nu: Optional[Union[float, Sequence[float]]] = None,
         reduction: Optional[callable] = None,
         weight_decay: float = 0.0,
-        **kwargs
+        **kwargs,
     ) -> None:
         # language=rst
         """
@@ -155,7 +162,7 @@ def __init__(
             nu=nu,
             reduction=reduction,
             weight_decay=weight_decay,
-            **kwargs
+            **kwargs,
         )
 
         assert (
@@ -253,7 +260,7 @@ def __init__(
         nu: Optional[Union[float, Sequence[float]]] = None,
         reduction: Optional[callable] = None,
         weight_decay: float = 0.0,
-        **kwargs
+        **kwargs,
     ) -> None:
         # language=rst
         """
@@ -271,7 +278,7 @@ def __init__(
             nu=nu,
             reduction=reduction,
             weight_decay=weight_decay,
-            **kwargs
+            **kwargs,
         )
 
         assert self.source.traces, "Pre-synaptic nodes must record spike traces."
@@ -391,7 +398,7 @@ def __init__(
         nu: Optional[Union[float, Sequence[float]]] = None,
         reduction: Optional[callable] = None,
         weight_decay: float = 0.0,
-        **kwargs
+        **kwargs,
     ) -> None:
         # language=rst
         """
@@ -409,7 +416,7 @@ def __init__(
             nu=nu,
             reduction=reduction,
             weight_decay=weight_decay,
-            **kwargs
+            **kwargs,
         )
 
         assert (
@@ -496,7 +503,7 @@ def __init__(
         nu: Optional[Union[float, Sequence[float]]] = None,
         reduction: Optional[callable] = None,
         weight_decay: float = 0.0,
-        **kwargs
+        **kwargs,
     ) -> None:
         # language=rst
         """
@@ -520,7 +527,7 @@ def __init__(
             nu=nu,
             reduction=reduction,
             weight_decay=weight_decay,
-            **kwargs
+            **kwargs,
         )
 
         if isinstance(connection, (Connection, LocalConnection)):
@@ -690,7 +697,7 @@ def __init__(
         nu: Optional[Union[float, Sequence[float]]] = None,
         reduction: Optional[callable] = None,
         weight_decay: float = 0.0,
-        **kwargs
+        **kwargs,
     ) -> None:
         # language=rst
         """
@@ -715,7 +722,7 @@ def __init__(
             nu=nu,
             reduction=reduction,
             weight_decay=weight_decay,
-            **kwargs
+            **kwargs,
         )
 
         if isinstance(connection, (Connection, LocalConnection)):
@@ -788,7 +795,7 @@ def _connection_update(self, **kwargs) -> None:
         self.p_minus += a_minus * target_s
 
         # Calculate point eligibility value.
-        self.eligibility = torch.ger(self.p_plus, target_s) + torch.ger(
+        self.eligibility = torch.outer(self.p_plus, target_s) + torch.outer(
             source_s, self.p_minus
         )
 
@@ -894,7 +901,7 @@ def __init__(
         nu: Optional[Union[float, Sequence[float]]] = None,
         reduction: Optional[callable] = None,
         weight_decay: float = 0.0,
-        **kwargs
+        **kwargs,
     ) -> None:
         # language=rst
         """
@@ -919,7 +926,7 @@ def __init__(
             nu=nu,
             reduction=reduction,
             weight_decay=weight_decay,
-            **kwargs
+            **kwargs,
         )
 
         # Trace is needed for computing epsilon.

bindsnet/models/models.py

@@ -391,7 +391,7 @@ def __init__(
         )
         self.add_connection(input_output_conn, source="X", target="Y")
 
-        # add internal inhibetory connections
+        # add internal inhibitory connections
         w = torch.ones(self.n_neurons, self.n_neurons) - torch.diag(
             torch.ones(self.n_neurons)
         )

bindsnet/network/network.py

@@ -231,7 +231,7 @@ def _get_inputs(self, layers: Iterable = None) -> Dict[str, torch.Tensor]:
                             self.batch_size,
                             target.res_window_size,
                             *target.shape,
-                            device=target.s.device
+                            device=target.s.device,
                         )
                     else:
                         inputs[c[1]] = torch.zeros(
@@ -275,6 +275,7 @@ def run(
             learning.
         :param Dict[Tuple[str], torch.Tensor] masks: Mapping of connection names to
             boolean masks determining which weights to clamp to zero.
+        :param Bool progress_bar: Show a progress bar while running the network.
 
         **Example:**
 
@@ -306,6 +307,11 @@ def run(
             plt.title('Input spiking')
             plt.show()
         """
+        # Check input type
+        assert type(inputs) == dict, (
+            "'inputs' must be a dict of names of layers "
+            + f"(str) and relevant input tensors. Got {type(inputs).__name__} instead."
+        )
         # Parse keyword arguments.
         clamps = kwargs.get("clamp", {})
         unclamps = kwargs.get("unclamp", {})

bindsnet/network/nodes.py

@@ -70,10 +70,9 @@ def __init__(
             self.register_buffer(
                 "tc_trace", torch.tensor(tc_trace)
             )  # Time constant of spike trace decay.
-            if self.traces_additive:
-                self.register_buffer(
-                    "trace_scale", torch.tensor(trace_scale)
-                )  # Scaling factor for spike trace.
+            self.register_buffer(
+                "trace_scale", torch.tensor(trace_scale)
+            )  # Scaling factor for spike trace.
             self.register_buffer(
                 "trace_decay", torch.empty_like(self.tc_trace)
             )  # Set in compute_decays.
@@ -101,7 +100,7 @@ def forward(self, x: torch.Tensor) -> None:
             if self.traces_additive:
                 self.x += self.trace_scale * self.s.float()
             else:
-                self.x.masked_fill_(self.s.bool(), 1)
+                self.x.masked_fill_(self.s.bool(), self.trace_scale)
 
         if self.sum_input:
             # Add current input to running sum.
@@ -1538,7 +1537,7 @@ def ExponentialKernel(self, dt):
 
     def RectangularKernel(self, dt):
         t = torch.arange(0, self.res_window_size, dt)
-        kernelVec = 1 / (selftau * 2)
+        kernelVec = 1 / (self.tau * 2)
         return torch.flip(kernelVec, [0])
 
     def TriangularKernel(self, dt):