Thoughts on one sided truncations?

[hesenp]

Hey there,

Thanks a lot for writing this wonderful module. While it meets most of my modeling needs, I wonder if you have ever considered enabling the module for one sided truncation? I was thinking that using np.inf might solve the problem. But it does not seem to work as I imagined:

import tensorflow as tf 
import numpy as np 
from truncatedDistribution import TruncatedDistribution as TD
import matplotlib.pyplot as plt 

session = tf.Session()

base_normal = tf.distributions.Gamma(100.,2.)
trunc_normal = TD(base_normal, - np.inf, 10.)

plt.hist(trunc_normal.sample(100).eval(session = session))

Here the output is a vector full of Nan.