Lesson 21: Deep Learning Foundations to Stable Diffusion

Jeremy explores ways to make the model faster without sacrificing quality. The Denoising Diffusion Implicit Model (DDIM) is introduced as a faster alternative to DDPM, and Jeremy demonstrates how to build a custom DDIM from scratch. The lesson concludes with a discussion on the differences between DDPM and DDIM, as well as the benefits of using DDIM for rapid sampling.

0:00:00 - A super cool demo with miniai and CIFAR-10
0:02:55 - The notebook
0:07:12 - Experiment tracking and W&B callback
0:16:09 - Fitting
0:17:15 - Comments on experiment tracking
0:20:50 - FID and KID, metrics for generated images
0:31:07 - Get the FID from an existing model
0:37:22 - Covariance matrix
0:42:21 - Matrix square root
0:46:17 - Why it is called Fréchet Inception Distance (FID)
0:47:54 - Some FID caveats
0:50:13 - KID: Kernel Inception Distance
0:55:30 - FID and KID plots
0:57:09 - Real FID - The Inception network
1:01:16 - Fixing (?) UNet feeding - DDPM_v3
1:08:49 - Schedule experiments
1:14:52 - Train DDPM_v3 and testing with FID
1:19:01 - Denoising Difussion Implicit Models - DDIM
1:26:12 - How does DDIM works?
1:30:15 - Notation in Papers
1:32:21 - DDIM paper
1:53:49 - Wrapping up

Timestamps and transcript thanks to Francisco Mussari.

[bayesianmonk]

Sometimes explaining the math helps more than escaping it, no heavy math is used anyway. I found the explanation of DDIM not very clear. Thanks for the course and videos.

[satirthapaulshyam]

Samples in these diffusion models r b2in -1 and 1 29:59

[frankchieng]

i thought in the class of WandBCB(MetricsCB)
def _log(self, d):
if self.train:
should be modified with if d['train']=='train'

[kettensaegenschlucker]

1:27:35 - Wondering what made you cheer, Johno ... 😂

Edit: spoiler alert, but there is a resolution shortly after