network.add_layer(input_layer, name="X")
network.add_layer(conv_layer, name="Y")
network.add_connection(conv_conn, source="X", target="Y")
# Train the network.
print("Begin training.\n")
if args.tensorboard:
analyzer = TensorboardAnalyzer("logs/conv")
else:
analyzer = MatplotlibAnalyzer()
for step, batch in enumerate(tqdm(train_dataloader)):
# batch contains image, label, encoded_image since an image_encoder
# was provided
# batch["encoded_image"] is in BxTxCxHxW format
inputs = {"X": batch["encoded_image"]}
# Run the network on the input.
# Specify the location of the time dimension
network.run(inputs=inputs, time=time, input_time_dim=1)
network.reset_state_variables() # Reset state variables.
analyzer.plot_conv2d_weights(conv_conn.w, step=step)
analyzer.finalize_step()
voltage_ims, voltage_axes = plot_voltages(
{
layer: voltages[layer].get("v").view(-1, time)
for layer in voltages
},
ims=voltage_ims,
axes=voltage_axes,
)
weights_im = plot_weights(get_square_weights(C1.w, 23, 28),
im=weights_im,
wmin=-2,
wmax=2)
weights_im2 = plot_weights(C2.w, im=weights_im2, wmin=-2, wmax=2)
plt.pause(1e-8)
network.reset_state_variables()
# Define logistic regression model using PyTorch.
class NN(nn.Module):
def __init__(self, input_size, num_classes):
super(NN, self).__init__()
# h = int(input_size/2)
self.linear_1 = nn.Linear(input_size, num_classes)
# self.linear_1 = nn.Linear(input_size, h)
# self.linear_2 = nn.Linear(h, num_classes)
def forward(self, x):
out = torch.sigmoid(self.linear_1(x.float().view(-1)))
# out = torch.sigmoid(self.linear_2(out))
return out