net output dimension mismatch on cifar10

[lucaskyle]

hey there:

when i run python train_CIFAR-10.py

Batch_size=128
so the output of network should be (128,10)
i got (512,10)

ValueError: Expected input batch_size (512) to match target batch_size (128).

[lucaskyle]

should i modify the Network?

[lucaskyle]

ok i checked official magvii code
seems like u r working on training supernet on cifar10？

[lucaskyle]

the gap_size should be 2

[ShunLu91]

Thanks! I will fix these bugs.

[lucaskyle]

this is so strange
the train loss is still Loss: 2.303
it didnt converge

i try other models that the train loss reduced easily...

maybe sth wrong in the supernetwork.
i will check magvii codes and their supermodel.

[ShunLu91]

Sorry, there are some bugs in the model.py and it has been fixed now. You can refer to the latest version of the code.

[lucaskyle]

its the learning rate
seems like u fixed it.

thx