search

2017-fall-DL-training-program / PyTorchWarmUp

3 stars 0 forks source link

exercise#1 in the WarmUp #4

Open conlinwang opened 6 years ago

conlinwang commented 6 years ago

Dear TAs, One silly question, from me. in the WarmUp LeNet

self.conv1 = nn.Conv2d(1, 6, kernel_size=(5,5),stride=1, padding=0) self.conv2 = nn.Conv2d(6, 16, kernel_size=(5,5),stride=1, padding=0) self.fc1 = nn.Linear(1644, 120) <---

when I was reading the LeNet paper I thought it was 16 (55), because the input data is 32x32 however, I've output the size of from the code and found out the size was 28x28 why is it not 32x32 do datasets.MNIST on PyTorch chop down the size of the datasets.MNIST?? Thank you so much

ref: (128L, 1L, 28L, 28L) (128L, 6L, 24L, 24L) (128L, 6L, 12L, 12L) (128L, 16L, 8L, 8L) (128L, 16L, 4L, 4L) (128L, 256L) (128L, 120L) (128L, 84L) (128L, 10L)

JiaRenChang commented 6 years ago

Hi, The original image size of MNIST which Lecun released is 28x28. Jia-Ren

conlinwang commented 6 years ago

I see thx U!!