PortaSpeech / utils /nn /model_utils.py
RayeRen's picture
init
d1b91e7
raw
history blame
472 Bytes
import numpy as np
def print_arch(model, model_name='model'):
print(f"| {model_name} Arch: ", model)
num_params(model, model_name=model_name)
def num_params(model, print_out=True, model_name="model"):
parameters = filter(lambda p: p.requires_grad, model.parameters())
parameters = sum([np.prod(p.size()) for p in parameters]) / 1_000_000
if print_out:
print(f'| {model_name} Trainable Parameters: %.3fM' % parameters)
return parameters