models/spatial/attncnf.py [204:217]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
def gaussian_loglik(z, mean, log_std):
    mean = mean + torch.tensor(0.)
    log_std = log_std + torch.tensor(0.)
    c = torch.tensor([math.log(2 * math.pi)]).to(z)
    inv_sigma = torch.exp(-log_std)
    tmp = (z - mean) * inv_sigma
    return -0.5 * (tmp * tmp + 2 * log_std + c)


def gaussian_sample(mean, log_std):
    mean = mean + torch.tensor(0.)
    log_std = log_std + torch.tensor(0.)
    z = torch.randn_like(mean) * torch.exp(log_std) + mean
    return z
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



models/spatial/cond_gmm.py [121:134]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
def gaussian_loglik(z, mean, log_std):
    mean = mean + torch.tensor(0.)
    log_std = log_std + torch.tensor(0.)
    c = torch.tensor([math.log(2 * math.pi)]).to(z)
    inv_sigma = torch.exp(-log_std)
    tmp = (z - mean) * inv_sigma
    return -0.5 * (tmp * tmp + 2 * log_std + c)


def gaussian_sample(mean, log_std):
    mean = mean + torch.tensor(0.)
    log_std = log_std + torch.tensor(0.)
    z = torch.randn_like(mean) * torch.exp(log_std) + mean
    return z
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -