fix: define shape variables in forward method of SpikingSelfAttention class

Author: Khushiyant

src/tether/nn/attention.py

@@ -57,6 +57,7 @@ def forward(self, x_seq):
         torch.Tensor
             Output sequence of shape (T, B, N, D).
         """
+        T, B, N, D = x_seq.shape
         x_flat = x_seq.view(T, B * N, D)
 
         # Q, K, V Generation through LIF neurons