learning rate #18
Description
In the paper you said "The model is trained using SGD for 600 epochs with a batch size of 8, momentum of 0.9 and learning rate of 10−7". Why we need to set learning rate so small?
{{ message }}
In the paper you said "The model is trained using SGD for 600 epochs with a batch size of 8, momentum of 0.9 and learning rate of 10−7". Why we need to set learning rate so small?