วีดีโอ

This is great, I learned a lot. Thank you very much for sharing this!

Intermediate steps for coding could be each of the commits in turn that leads to an accepted PR. Often there are even multiple PRs in open projects and one is accepted but not another. Usually if this is the case there would be cross links between the issues.

Fresher should try ml?

banger as always, thanks for doing those king

Take a shot every time I mix up 'policy', 'process' and 'preference' :D

Loved seeing the thought process of how you iterate quickly on research ideas. I did wonder what effect it would have if "CITY" was replaced by a dummy word. In that case, it should initially have very low similarity with "Paris", but the trained model should see it converge. Additionally, I wonder if LoRA had a larger effect on the values. What would happen inf you tried LoRA on the smaller models?

why he moves screen every 3 seconds, there's no need to do this

thank you very informative

GNN model merge please

great explanation!

I love these paper deep dives, can we also do more toy projects here :)

great video can you cover the other sampling method that yannic already talked about? maybe also with the code implementation and side by side comparisons

this is not working on updated stable diffusion webui and forge

You’re my favourite TH-camr!

thanks man

keep doing it

Thank you for highlighting the latest top papers

welcome back keep em coming

thank you for this video. I hope to one day be as capable as you in describing these papers. these are fascinating pieces of technology

Where is this? "GS website (with links to paper):"

Could you please explain the loss function in a bit more detail? Thanks

@datasciencecastnet Loving these videos! Is there going to be coverage of units 3 and 4 at any point?

Can you please share the git repo for all your code, it would be great to follow along and see the results on my end.

Thank you for this nice getting started video. Could learn a lot from it. One question, did you write the function json schema or you used any function to generate the schema.

the video is laggy :(

THat was an interesting video about steganographic techniques. The first thing I thought of was the movie Sneakers, where they used a cassette tape as the steganographic device. The movie is about 20+

Hi, really good video. Could you share the jupyter notebook you showed in this video? It would be so grateful!

very interesting, thank you! Can you also share the notebook code of your on this video?

isnt open ai api a paid feature?

Yeah it valueable. I've just added one to my read list. I don't think anyone goes over papers in a quick-fire method as you've just done where similar (also dis-similar) papers are compared at a high level on why you picked them and some institution of how they work.

Yaaay Johno!

Nice work haha i don't clean up zotero often enough

This is a really interesting idea.. With regards to guidance, do you know if anyone has tried to train models that try and predict those noise deltas you would get from pushing the gradient backwards, but doing it in a forward direction? For example, you could train a model to take the features from the "up" layers in Stable Diffusion, and then predict a secondary noise delta to try and correct the regular Stable Diffusions noise in the right direction. Like estimate what the delta of the back propagation noise would need to be. I'm not sure if that would actually save in compute power during inference, compared to having to do a backwards pass and holding the whole graph while you generate images, because I assume that model would need to be reasonable in size. But it might also allow larger steps in prediction than you might get from a single gradient backwards pass. And it would remove the need for an internal RGB image stage at all, because you would only need to use that model during training.. Although it would likely break the awesome part of this method that it requires very few samples to get good results, at the cost of shifting work to the inference stage.

Thank you very much for such a great work and explanation. Hatt's off

thanks a lot

hello i have a confusion regarding one topic in diffusions, can i get your contact info Jonathan, like your mail ID or any other contact info please, i am working on my last year project anything would help

very great introduction! I can see a lot of efforts have been put into this video! It helps a lot to understand the paper! thank you for sharing!!

Oh my god, man you don't understand how happy I am for your storytelling about how things went in the timeline of developing on the idea of model merging up to this point, where it started how it went, that and how they were thinking about reasons why it works, etc.etc.. I want to get into this so that I understand the main ideas and be able to start working on these as well, but it's so hard to get to the root of things, it requires a huge amount of time to read and digest everything and slowly being able to put the pieces together, so boy do I mean it when I say thank you!

Can you share the code please?

Hi Johno, at the beginning you said you're somewhat skeptical of model merging. Iiuc, your criticism is only about iterative merging for a given goal, which leads to overfitting. Or are you skeptical of the general concept of model merging? Thanks!

That was an excellent overview of not just Sakana's evolutionary methods to identify good merge candidates, but also the popular techniques TIES, DARE and Passthrough/Frankenmerge. Appreciate it as usual, Johno!

Deep dive: model merging m.th-cam.com/video/cvOpX75Kz4M/w-d-xo.html

very helpful!

It would've been cool if you could visualize which point in the scene you are showing the spherical harmonics for

Thanks for hosting this

when approaching a manifold, what would happen if the approach was aligned the norm of the manifold surface?

Can you share this code you have made for this video ?

Practical and useful, thank you!

Subscribed, rare to see someone tackle a new category of competition on Kaggle, nice work

enjoying it :)