DeepFaceLab/utils/random_utils.py
2019-03-19 23:53:27 +04:00

15 lines
379 B
Python

import numpy as np
def random_normal( size=(1,), trunc_val = 2.5 ):
len = np.array(size).prod()
result = np.empty ( (len,) , dtype=np.float32)
for i in range (len):
while True:
x = np.random.normal()
if x >= -trunc_val and x <= trunc_val:
break
result[i] = (x / trunc_val)
return result.reshape ( size )