Was GoogLeNet trained the same way as other torchvision models?
If so why is there additional preprocessing step via _transform_input function?
It’s called in forward() before the actual work in _forward() function is executed.
Was GoogLeNet trained the same way as other torchvision models?
If so why is there additional preprocessing step via _transform_input function?
It’s called in forward() before the actual work in _forward() function is executed.