realNVP

realNVP copied to clipboard

Here is possible typo `x, size()` https://github.com/fmu2/realNVP/blob/8d36691df215af3678440ccb7c01a13d2b441a4a/realnvp.py#L253 Should be `x.size()`

rnoxy

Hi. What's the purpose of [concating](https://github.com/fmu2/realNVP/blob/8d36691df215af3678440ccb7c01a13d2b441a4a/realnvp.py#L316) x, -x and mask in CheckerboardAffineCoupling layer ? `x_ = self.in_bn(x * mask) x_ = torch.cat((x_, -x_), dim=1) x_ = torch.cat((x_, mask), dim=1)` I...

blatr

Hi. Would you please upload your pretrained models or checkpoints?

isharifi

按照初始化的设置 训练了100个epoch ===> Average train loss: nan ===> Average train log-likelihood: nan ===> Average train bit_per_dim: nan ===> Average validation loss: nan ===> Average validation log-likelihood: nan ===> Average validation...

cymqqqq