1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
|
import torch
from files.models.snn import SimpleSNN
from files.data_io.dataset_loader import get_dataloader
import torch.nn as nn
import torch.optim as optim
def _one_step(model, xb, yb, lyapunov=False):
model.train()
opt = optim.Adam(model.parameters(), lr=1e-3)
ce = nn.CrossEntropyLoss()
opt.zero_grad(set_to_none=True)
logits, lyap = model(xb, compute_lyapunov=lyapunov)
loss = ce(logits, yb)
if lyapunov and lyap is not None:
loss = loss + 0.1 * (lyap - 0.0) ** 2
loss.backward()
opt.step()
assert torch.isfinite(loss).all()
def test_train_step_baseline_and_lyapunov():
train_loader, _ = get_dataloader("data_io/configs/shd.yaml")
xb, yb = next(iter(train_loader))
B, T, D = xb.shape
C = 20
model = SimpleSNN(input_dim=D, hidden_dim=64, num_classes=C)
# baseline
_one_step(model, xb, yb, lyapunov=False)
# lyapunov-regularized
_one_step(model, xb, yb, lyapunov=True)
|