fallenshock / sinddm Goto Github PK

hi, what does sigma_t means in model.py? issigma_t = np.sqrt(1. - alphas_cumprod) / np.sqrt(alphas_cumprod) # sigma_t = sqrt_one_minus_alphas_cumprod_div_sqrt_alphas_cumprod.

Hi, thanks for sharing this great work. I have some questions about the code.

1. I wonder how many steps the program takes to train a model for a 248×186 image like balloons.png.
2. In my training environment, I can only pip install certifi==2022.9.24, PyWavelets==1.3.0 and tqdm==4.63.1. However, the requirements in your repo are certifi==2022.12.7, PyWavelets==1.4.1 and tqdm==4.64.1. So I wonder if the downgraded version of the three packages will affect my training results. All the other packages are the same as yours.

Thank you in advance!

Hello, I'm Jennie who is studying Diffusion.
I was curious about the number 6~8 line code.
Why you make x_s,mix_t? also, I think you regard x_s,mix_t as x_0.
I'm not good enough to understand to this sampling code.
Please make sure this code! Thank you

Hi, I'm trying to understand the math part of your algorithm. And I wonder which interpolation you use in order to upsample/downsample the image. I saw both LANCZOS and BILINEAR interpolation in function.py, so I'm a little confused. Could you explain how you use these two interpolations? Thank you in advance.