testImgs = images[trainSize:]
testOneHotVecs = oneHotVecs[trainSize:]
trainingAccuracy = []
trainingLoss = []
testingAccuracy = []
testingLoss = []
epochs = list(range(0, cfg.epochs))
#training
for epoch in range(0, cfg.epochs):
#records weights to see if they change after training
weights1 = []
for param in cnn.parameters():
weights1.append(param.clone())
#training
print("Training Epoch", str(epoch + 1), "of", str(cfg.epochs) + ":")
trainAcc, trainLoss = cnn.train(trainImgs, trainOneHotVecs,
cfg.trainBatchSize)
trainingAccuracy.append(trainAcc)
trainingLoss.append(trainLoss)
#prints if weights are unchanged
weights2 = []
unchangedWeights = 0
for param in cnn.parameters():
weights2.append(param.clone())
for i in zip(weights1, weights2):
cnn.to(device)
print(cnn)
dummy_input = torch.randn(1, 1, 48, 48, device=device)
torch.onnx.export(cnn,
dummy_input,
"convnet.onnx",
verbose=True,
input_names=['input'],
output_names=['output'])
# %%
learn_rate = 3e-4
los = nn.CrossEntropyLoss()
loader = Loader(rtrain_x, 64, label=rtrain_y)
optim = torch.optim.Adam(cnn.parameters(), lr=learn_rate)
for ep in range(24):
print('epoches: %d' % ep)
for i, (x, y) in enumerate(loader):
optim.zero_grad()
x = x.to(device, dtype=torch.float)
# print(x.shape)
y = y.to(device, dtype=torch.long)
# print(y)
outt = cnn(x)
loss = los(outt, y)
# print(loss)
loss.backward()
optim.step()
if (i + 1) % (int(len(loader) / loader.bsz / 100)) == 0:
print('\r',
