First of all thank you for a great work
I have a question about your denoising sampling process
In your code, it looks like after predicting epsilon from noise prediction model, it compute X_0, and computes the mean and variance of the diffusion posterior q(x_{t-1} | x_t, x_0), and use this diffusion posterior mean and variance to sample x_{t-1}.
Can I know the reason why not directly sample x_{t-1} using the below equation?
Thanks.
First of all thank you for a great work
I have a question about your denoising sampling process
In your code, it looks like after predicting epsilon from noise prediction model, it compute X_0, and computes the mean and variance of the diffusion posterior q(x_{t-1} | x_t, x_0), and use this diffusion posterior mean and variance to sample x_{t-1}.
Can I know the reason why not directly sample x_{t-1} using the below equation?
Thanks.