Conditional DDPM using PyTorch - Example with MNIST dataset

In this video, 𝐌𝐍𝐈𝐒𝐓 dataset images have been used to train a 𝘊𝘰𝘯𝘥𝘪𝘵𝘪𝘰𝘯𝘢𝘭_𝘋𝘋𝘗𝘔 along with their labels to conditionally generate the output. 𝐌𝐍𝐈𝐒𝐓 dataset has 10 classes: 𝗢𝗻𝗲 - 𝟭, 𝗧𝘄𝗼 - 𝟮, 𝗧𝗵𝗿𝗲𝗲 - 𝟯, 𝗙𝗼𝘂𝗿 - 𝟰, 𝗙𝗶𝘃𝗲 - 𝟱, 𝗦𝗶𝘅 - 𝟲, 𝗦𝗲𝘃𝗲𝗻 - 𝟳, 𝗘𝗶𝗴𝗵𝘁 - 𝟴, 𝗡𝗶𝗻𝗲 - 𝟵, 𝗧𝗲𝗻 - 𝟭𝟬.
The labels of 𝐌𝐍𝐈𝐒𝐓 images were fed to an embedding table with 10 embeddings (for 10 classes), each with an embedding size of 256 (the same as time embeddings). These embeddings were then added with time embeddings to conditionally train the 𝐃𝐃𝐏𝐌.
𝙂𝙞𝙩𝙃𝙪𝙗 𝙖𝙙𝙙𝙧𝙚𝙨𝙨: github.com/randomaccess2023/M...
#datascience #conditionalddpm #python #pythonprogramming #deeplearning #generativemodels #generativeai #diffusionmodels #jupyternotebook #denoisingdiffusionprobabilisticmodels

[@user-id1cp5ef5v]

good job