improved-diffusion
improved-diffusion copied to clipboard
openai
Release for Improved Denoising Diffusion Probabilistic Models
Hi, Where can I find details about the format of the images (for custom dataset) such as type(unit8, float..), size, number of channel and etc. thanks
What is the no. of epochs by default because i have been training the model with 16 batch size and has been training for quite a long time . and...
I am trying to train the improved diffusion rack with my denoiser. However, the model is not getting saved anywhere on the system. I have searched the entire storage, there...
My training stopped abruptly in between and I have 3 checkpoints saved - ema_0.9999_020000.pt, model020000.pt and opt020000.pt. Now to again start training I want to resume from same state but...
Has anyone used large dataset such as laion-5b etc to train an improved ddpm model? Is there any work like this?
I can train normally, but I encounter a key value missing problem when sampling. What is the reason? 
After sampling, a .npz file is being created. Please let me know - how to handle that and how to create images out of it so that I can see...
My dataset consists of 8 thousand grayscale images of 256 * 256 size，the follow is my train script: MODEL_FLAGS="--image_size 256 --num_channels 128 --num_res_blocks 3" ``` DIFFUSION_FLAGS="--diffusion_steps 1000 \ --noise_schedule cosine...
Based on the latest state-of-the-art studies, the sigmoid scheduler can be a good and stable choice for diffusion models, adding more flexibility to them.
