The training of this notebook on colab takes approximately 20 minutes.
- ddpm_nano_empty.ipynb is the notebook where you code the DDPM algorithm (a simple UNet is provided for the denoising network), its training and the sampling. Here is the corresponding solution: ddpm_nano_sol.ipynb
The training of this notebook on colab takes approximately 20 minutes (so do not expect high-quality pictures!). Still, after finetuning on specific classes, we see that the model learns features of the class.