I’m using a Resnet50 model
Even after the pruning step, the weights doesn’t become 0’s.
Can anyone help me?
I’m using a Resnet50 model
Even after the pruning step, the weights doesn’t become 0’s.
Can anyone help me?
Hi MOUNIKA, Sorry for late reply
You can reduce the pruning frequency to < 80 (40 steps * 2 epochs) to make pruning happen within the training steps. Simply set
…PolynomialDecay( … , frequency = 10)
The default frequency is 100, so without setting it explicitly, the pruning does not take place during your training process.