64 lines
2.4 KiB
Diff
Executable file
64 lines
2.4 KiB
Diff
Executable file
diff --git a/mnist_classifier.py b/mnist_classifier.py
|
|
index 8eac991..4578ce3 100644
|
|
--- a/mnist_classifier.py
|
|
+++ b/mnist_classifier.py
|
|
@@ -3,7 +3,10 @@ import torch.nn as nn
|
|
import torch.optim as optim
|
|
import torch.nn.functional as F
|
|
from torchvision import transforms, datasets
|
|
+from tqdm import tqdm
|
|
+import wandb
|
|
|
|
+wandb.init(project='pytorch_ai')
|
|
train = datasets.MNIST('./datasets', train=True, download=True,
|
|
transform=transforms.Compose([
|
|
transforms.ToTensor()
|
|
@@ -37,31 +40,32 @@ class Net(nn.Module):
|
|
|
|
|
|
net = Net()
|
|
+wandb.watch(net)
|
|
|
|
loss_function = nn.CrossEntropyLoss()
|
|
optimizer = optim.Adam(net.parameters(), lr=0.001)
|
|
|
|
-for epoch in range(10): # 3 full passes over the data
|
|
- for data in trainset: # `data` is a batch of data
|
|
+for epoch in range(10): # 10 full passes over the data
|
|
+ for data in tqdm(trainset): # `data` is a batch of data
|
|
X, y = data # X is the batch of features, y is the batch of targets.
|
|
net.zero_grad() # sets gradients to 0 before loss calc. You will do this likely every step.
|
|
output = net(X.view(-1, 784)) # pass in the reshaped batch (recall they are 28x28 atm)
|
|
- loss = F.nll_loss(output, y) # calc and grab the loss value
|
|
+ loss = loss_function(output, y) # calc and grab the loss value
|
|
loss.backward() # apply this loss backwards thru the network's parameters
|
|
optimizer.step() # attempt to optimize weights to account for loss/gradients
|
|
+ wandb.log({'loss': loss})
|
|
|
|
- print(loss) # print loss. We hope loss (a measure of wrong-ness) declines!
|
|
- torch.save(net, './nets/net_' + str(epoch) + ".pt")
|
|
+ # torch.save(net, './nets/net_' + str(epoch) + ".pt")
|
|
correct = 0
|
|
total = 0
|
|
with torch.no_grad():
|
|
for data in testset:
|
|
X, y = data
|
|
output = net(X.view(-1, 784))
|
|
- # print(output)
|
|
for idx, i in enumerate(output):
|
|
- # print(torch.argmax(i), y[idx])
|
|
if torch.argmax(i) == y[idx]:
|
|
correct += 1
|
|
total += 1
|
|
+ wandb.log({'test_accuracy': correct / total})
|
|
print("Accuracy: ", round(correct / total, 3))
|
|
+ wandb.log({'epoch': epoch})
|
|
diff --git a/wandb/settings b/wandb/settings
|
|
index 8dae664..73c329e 100644
|
|
--- a/wandb/settings
|
|
+++ b/wandb/settings
|
|
@@ -1,2 +1,4 @@
|
|
[default]
|
|
+project = pytorch_ai
|
|
+entity = cdautermann
|
|
|