Wasn’t sure if anyone was working on this or planning on it but I’ll be working on chapter markers for Lesson 9 today. @jeremy do you know of any current efforts?
Also, I imagine 9A and 9B will need markers as well. @johnowhitaker @ilovescience @seem do you need help with those?
Link to Part 1 discussion on chapter markers and instructions. If I remember correctly, we put too many chapter markers last time so I’ll make markers with that in mind.
Chapter 9
0:00 - Introduction
6:38 - This course vs DALL-E 2
10:38 - How to take full advantage of this course
12:14 - Cloud computing options
14:58 - Getting started (Github, notebooks to play with, resources)
20:48 - Diffusion notebook from Hugging Face
26:59 - How stable diffusion works
30:06 - Diffusion notebook - guidance scale, negative prompts, starting diffusion with an image, textual inversion, Dreambooth
45:00 - Stable diffusion explained & how fast.ai will be teaching it
1:14:37 - Creating a neural network to predict noise in an image
1:27:46 - Working with images and compressing the data with autoencoders
1:40:12 - Explaining latents that will be input into the unet
1:43:54 - Adding text as one hot encoded input to the noise and drawing (aka guidance)
1:47:06 - How do you represent numbers vs text embeddings in our model with image and text encoders (CLIP)
1:53:13 - Encoder’s loss function
2:00:55 - Caveat regarding “time steps”. This term won’t be used in this course
2:07:04 Why don’t we do this all in one step?
The section from 45 mins on could do with more detail if possible. I think around 8-10 mins between timestamps is a good goal.
Absolutely agree with the timestamp spacing. I went with a bit of a compromise and just combined timestamps I would normally have so you’re still able to search for topics with text without overloading users with timestamps. After 45:00 is still being worked on. That was just my stopping point as I wasn’t sure if anyone else was working on this. Seems like I have the green light though so I’ll work on it more this afternoon.
Hi, how about the transcriptions? I’m waiting for the google doc files to start transcribing. I can create the files, but I’m not sure if it is better for Jeremy to own the files…
Yep we would very much appreciate help on the math lesson 
Thank you so much!
Actually I’m just re-rendering that bit now. It’ll add a few minutes to all the timestamps after 45mins, since I’ve recording a new section. Will be done in an hour or two.
This is a great initiative and adds a lot of value! As a fellow member of the community I say thank you!
Got it. It’d help me greatly if you could tell me at what time you added/re-recorded a new segment and how long the duration was.
The new section was inserted at 53:04. The new section is 14:53 long. I’ve uploaded it the new video YouTube but it’s processing very slowly. I’ll post here when it’s done.
Lesson 9B
0:00 - Introduction
2:19 - Data distribution
6:38 - Math behind lesson 9’s “Magic API”
18:50 - CLIP (Contrastive Language–Image Pre-training)
27:04 - Forward diffusion (markov process with gaussian transitions)
36:11 - Likelihood vs log likelihood
42:16 - Denoising diffusion probabilistic model (DDPM)
48:04 - Conclusion
Additional Links:
Deep Unsupervised Learning using Nonequilibrium Thermodynamics - https://arxiv.org/abs/1503.03585
Denoising Diffusion Probabilistic Models - https://arxiv.org/abs/2006.11239
The new video is now available
I’ve added 14:53 to the times after 53:04 now in your post.
Perfect! That saves me the effort. I believe that someone earlier in the thread also asked about transcriptions. Do you want to put them in a Google Doc again for us to edit?
Also, I forgot a timestamp for Ch9
2:07:04 Why don’t we do this all in one step?
Lesson 10
0:00 - Introduction
0:35 - Showing student’s work over the past week.
6:04 - Recap Lesson 9
12:55 - Explaining the papers “Progressive Distillation for Fast Sampling of Diffusion Models” and “On Distillation of Guided Diffusion Models”
26:53 - Explaining the paper “Imagic: Text-Based Real Image Editing with Diffusion Models”
33:53 - Stable diffusion pipeline code walkthrough
41:19 - Scaling random noise to ensure variance
50:21 - Recommended homework for the week
53:42 - What are the foundations of stable diffusion? Notebook deep dive
1:06:30 - Numpy arrays and PyTorch Tensors from scratch
1:28:28 - History of tensor programming
1:37:00 - Random numbers from scratch
1:42:41 - Important tip on random numbers via process forking
Additional Links:
Progressive Distillation for Fast Sampling of Diffusion Models - https://arxiv.org/abs/2202.00512
On Distillation of Guided Diffusion Models - https://arxiv.org/abs/2210.03142
Imagic: Text-Based Real Image Editing with Diffusion Models - https://arxiv.org/abs/2210.09276
@jeremy the video is publicly listed as well lesson 9. Is it correct? Or should they be available only for part 2 attendees until the public release?
Nevermind, I just read the posts about the preview
Awesome work! Lesson 10 chapter markers are now done as well. The YouTube video can now be updated with all the timestamps Jeremy.
