Diffusion Models | PyTorch Implementation
Вставка
- Опубліковано 8 чер 2024
- Diffusion Models are generative models just like GANs. In recent times many state-of-the-art works have been released that build on top of diffusion models such as #dalle , #imagen or #stablediffusion . In this video I'm coding a PyTorch implementation of diffusion models in a very easy and straightforward way. At first I'm showing how to implement an unconditional version and subsequently train it. After that I'm explaining 2 popular improvements for diffusion models: classifier free guidance and exponential moving average. I'm also going to implement both updates and train a conditional model on CIFAR-10 and afterwards compare the different results.
Code: github.com/dome272/Diffusion-...
#diffusion #dalle2 #dalle #imagen #stablediffusion
00:00 Introduction
02:05 Recap
03:16 Diffusion Tools
07:22 UNet
13:07 Training Loop
15:44 Unconditional Results
16:05 Classifier Free Guidance
19:16 Exponential Moving Average
21:05 Conditional Results
21:51 Github Code & Outro
Further Reading:
1. Paper: arxiv.org/pdf/1503.03585.pdf
2. Paper: arxiv.org/pdf/2006.11239.pdf
3. Paper: arxiv.org/pdf/2102.09672.pdf
4. Paper: arxiv.org/pdf/2105.05233.pdf
5. CFG: arxiv.org/pdf/2207.12598.pdf
6. Timestep Embedding: machinelearningmastery.com/a-...
Follow me on instagram lol: / dome271 - Наука та технологія
Link to the code: github.com/dome272/Diffusion-Models-pytorch
21:56 The way you starred your own repo makes my day bro 🤣🤣 really appreciate your work, just keep going!!
@@bao-dai xd
@@outliier Thanks for sharing but how do you not get bored or tired of doing the same thing for so long and deal with all the math?
@@leif1075 I love to do it. I don’t get bored
After I saw your next video "Cross Attention | method and math explained", I would like to see ControlNet's openpose in PyTorch Implementation which control posing on image of a dogs. Or if it is too complicate, you may simplify it to control 2 - 3 branches shape of a tree.
Hello, this has become a great video once again. We didn't understand much, but it's still nice to watch. Greetings from home say Mam & Dad. ;-))))
Great, this video is finally out! Awesome coding explanation! 👏
This videos is crazy! I don't get tired of recommend it to anyone interesting in diffusion models. I have recently started to research with these type of models and I think your video as huge source of information and guidance in this topic. I find myself recurrently re-watching your video to revise some information. Incredible work, we need more people like you!
Thank you so much for the kind words!
This video is really timely and needed. Thanks for the implementation and keep up the good work!
thank you so much for your detailed explaination of the code. It helped me a lot on my way of learning diffusion model. Wish there are more youtubers like you!
Incredible. Very thorough and clear. Very, very well done.
This channel seems to be growing very fast. Thanks for this amazing tutorial.🤩
most informative and easy to understand video on diffusion models on youtube, Thanks Man
After my midterm week i wanna study diffusion models with your videos im so exited .thanks a lot for good explanation
We chose Diffusion Model as part of our course project, and your videos do save much of my time to understand the concepts and have more focus on implementing the main part. I am really grateful for your contribution.
Sincere gratitude for this tutorial, this has really helped me with my project. Please continue with such videos.
Thank you for sharing the implementation since authentic resources are rare
Amazing tutorial, very informative and clear, nice work!
The best video for diffusion! Very Clear
Congrats, This is a great channel!! hope to see more of these videos in the future.
Very helpful walk-through. Thank you!
Thank you. Best explanation with good DNN models
I was wating for so long i learnd about condicional difusion models
great tutorial! looking to seeing more of this! keep it up!
Dude, you're amazing! Thanks for uploading this!
Thanks, this implementation really helped clear things up.
this is the most underrated channel i've ever seen, amazing explanation !
thank you so much!
Thank you so much for this sharing, that was perfect!
Incredible explanation, thanks a lot!
Very nicely explained. Thanks.
Hi, @Outlier , thank you for the awesome explanation !
Just one observation, I believe in line 50 of your code (at 19:10) it should be:
uncond_predicted_noise = model(x,t,None)
😁
good catch thank you. (It's correct in the github code tho :))
Thank you very much, it has solved my urgent need
nice demonstration, thanks for sharing
It's definitely cool and helpful! Thanks!!!
thanks for your amazing efforts!
Very well done! Keep the great content!!
awesome implementation!
Looking forward for some video on Classifier Guidance as well. Thanks.
Fantastic video!
This is my first few days of trying to understand diffusion models. Coding was kinda fun on this one. I will take a break for 1-2 months and study something related like GANs or VAE, or even energy-based models. Then comeback with more general understanding :) Thanks !
And transformers for the attention mechanisms + positional encoding
I got that snatched in the past 2 months. Gotta learn the math, what is actually a distribution etc.@@zenchiassassin283
it is very helpful!! You are a genius.. :) thank you!!
Amazing stuff!
Thank you for sharing!
Great video!
This is GOLD
great video, you got one new subscriber
Thank you so much for this amazing video! You mention that changing the original DDPM to a conditional model should be as simple as adding in the condition at some point during training. I was just wondering if you had any experience with using DDPM to denoise images? I was planning on conditioning the model on the input noisy data by concatenating it to yt during training. I am going to try and play around with your github code and see if I can get something to work with denoising. Wish me luck!
This video is priceless.
Thank you!!
Great video!! You make coding seem like playing super mario 😂😂
Nice tutorial
great walkthrough. but where would i implement dynamic or static thresholding as described in the imagen paper? the static thresholding clips all values larger then 1 but my model regularly outputs numbers as high as 5. but it creates images and loss decreases to 0.016 with SmoothL1Loss.
Could you please explain the paper "High Resolution Image Synthesis With Latent Diffusion Models" and its implementations? Your explanations are exceptionally crystal.
Super cool
best diffusion youtube
Thank you so much for this amazing video! In mention that the first DDPM paper show no necessary of lower bound formulation, could you tell me the specific place in the paper? thanks!
terrific!
Great Video,
On what Data did you train your model again?
The best
Thank you for the video.
How can we use diffusion model for inpainting?
Can you do one for tensorflow too btw very good explaination
Great video, thanks for making it. I started working with diffusion models very recently and I used you implementation as base for my model. I am currently facing a problem that the MSE loss starts very close to 1 and continues like that but varying between 1.0002 and 1.0004, for this reason the model is not training properly. Did you face any issue like this one? I am using the MNIST dataset to train the network, I wanted to first test it with some less complex dataset.
I am facing similar problems. I did the experiment on CIFAR10 dataset. The mse loss starts descresing normally but at some points the loss increse to 1 and never descrese again.
Like your channel, please make more videos
Thank you for the review. So, what is the key to make a step from text description to image? Can you please pinpoint where it is explained?
Your videos are a blessing. Thank you very much!!! Have you tried using DDIM to accelerate predictions? Or any other idea to decrease the number of steps needed?
I have not tried any speedups in any way. But feel free to try it out and tell me / us what works best. In the repo I do linked a fork which implements a couple additions which make the training etc. faster. You can check that out too here: github.com/tcapelle/Diffusion-Models-pytorch
@@outliier Thank you! I will try it for sure.
Awesome! How did you type Ɛ in code?
Great videos on diffusion models, very understandable explanations! For how many hours did you train it? I tried adjusting your conditional model and train with a different dataset, but it seems to take forever :D
Yea it took quite long. On the 3090 it trained a couple days (2-4 days I believe)
@@outliier Thanks for the feedback. Ok seems like I didn't do a mistake, but only need more patience!
@@maybritt-sch Yea. Let me know how it goes or if you need help
Very cool. How would DDIM models be different? Do they use a deterministic denoising sampler?
yes indeed
Can you please explain how to use Woodfisher technique to approximate second-order gradients? Thanks
Thx Mr gigachad
Great video! How long did it take to train the models?
About 3-4 days on an rtx 3090.
Wonderful video! I notice that at 18:50, the equation for the new noise seems to differ from Eq. 6 in the CFG paper, as if the unconditioned and conditioned epsilons are reversed. Can you comment on that?
having hard time to understand the mathematical and code aspect of diffusion model although i have a good high level understanding...any good resource i can go through? id appreciate it
pog!!!
Hello, thanks for your a lot contribution ! But a bit confused, At 06:04, just sampling from N(0, 1) totally randomly would not have any "trace" of an image. How come the model infer the image from the totally random noise ?
Hey there, that is sort of the "magic" of diffusion models which is hard to grasp your mind around. But since the model is trained to always see noise between 0% and 100% it sees full noise during training for which it is then trained to denoise it. And usually when you provide conditioning to the model such as class labels or text information, the model has more information than just random noise. But still, unconditional training still works.
Hi , I want to use a single underwater image dataset what changes do i have to implement on the code?
@outliier Do you think there is a way to run the code with a 3060 GPU on personal desktop? I get the error message: CUDA out of memory.
last self attention layer (64, 64) changes my training type from 5 minutes to hours per epoch, do you know why?
training on a single rtx 3060 TI gpu
Your content is the best one in the field of AI! I love it. I tried to use this code on my own dataset and find out that even after 100 epochs the model generates some block images and I don't understand why? is this kind of a bug or it's a normal thing and it gets fixed with longer training? I am thinking that the output of the model, ie the predicted noise, is not always in the same range as the noise (which seams to be 0,1 as we are returning something with torch.rand_like) and this might end up with some values outside of the 0, 255 final image range. I am thinking to add sigmoid activation for the last layer to clip the data to 0, 1. Do you have any thoughts about that.
I also tried to use cosine schedule, tho the results of epoch 80 were still not that complex, and it looked worst than the one generated in epoch 40 by the linear schedule. I noticed that the loss was also higher so my though is that this task is a more difficult one and needs a lot more train. This is also the reason why you chose the linear one? (about the code, I just replaced the prepare_noise_schedule function with the one provided by OpenAI, not 100% if this approch is correct).
Hey there. First of all thank you so much for the nice words. Can I ask you to open an issue on GitHub and upload some images of what you are experiencing? Ill take a look then and maybe also others :)
`
x = torch.randn((n, 3, self.img_size, self.img_size)).to(self.device)
predicted_noise = model(x, t)
`
in the deffusion class why you create an noise and pass that noise into the model to predict noise ... please explain
Can you please tell me how much time was need to train this 3000 image for 500 Epoch?
With this training method, wouldn't there be a possibility of some timesteps not being trained in an epoch? wouldn't it be better to shuffle the whole list of timesteps and then sample sequentially with every batch?
can you do a text to image in small dataset similar to SD from scratch?
So the process of adding noise and removing it happens in a loop
Really nice video! I also enjoyed your explanation video - great work in general :)
However, I noticed at around 5:38, you are defining sample_timesteps with low=1. I am pretty sure that this is wrong, as Python indexes at 0 meaning you skip the first noising step every time you access alpha, alpha_cumprod etc. Correct me if I am wrong but all the other implementations also utilise zero-indexing.
this function sample the timesteps of the denoising step. selecting time=0 is the original image itself. there is no point in taking 0 timestep.
How can i increase the img size to 128 pixels square?
Thank you very much for this very easy-to-understand implementation. I have one question: I don't understand the function def noise_images.
Assume that we have img_{0}, img_{1}, ..., img_{T}, which are obtained from adding the noise iteratively. I understand that img{t} is given by the formula "sqrt_alpha_hat * img_{0} + sqrt_one_minus_alpha_hat * Ɛ".
However, I don't understand the function "def noise_images(self, x, t)" in [ddpm.py].
It return Ɛ, where Ɛ = torch.randn_like(x). So, this is just a noise signal draw directly from the normal distribution. I suppose this random noise is not related to the input image? It is becasue randn_like() returns a tensor with the same size as input x that is filled with random numbers from a normal distribution with mean 0 and variance 1
In training, the predicted noise is compared to this Ɛ (line 80 in [ddpm.py]).
Why we are predicting this random noise? Shouldn't we predict the noise added at time t, i.e. "img_{t} - img_{t-1}"?
I had the same misconception before. It was actually explained by "AI Coffee Break with Letitia" channel in a video titled "How does Stable Diffusion work? - Latent Diffusion Models EXPLAINED".
Basically, the model tries to predict the WHOLE noise added to the image to go from noised image to a fully denoised image in ONE STEP. Because it's a hard task to do, the model does not excel at that so at inference we denoise it iteratively, each time subtracting only a small fraction of the noise predicted by the model. In this way, the model produces much better quality samples. At least that's how I understood it :P
@@Laszer271 While I understand it predicts the "whole noise", this "whole noise" is newly generated and I suppose the ground truth is (img_{t} - img_{0)).. still can't wrap my head around it.
You do not use any LR scheduler. Is this intentional? My understanding is that EMA is a functional equivalent of LR scheduler, but then I do not see any comparison between EMA vs e.g. cosine LR scheduler. Can you elaborate more on that?
Thank you...U just made diffusion so easy to understand... I would like to ask ; What changes do I need to make in order to give an image as condition rather than a label as condition. I mean how to load ground Truth from GT repository as label (y).
Depends on your task. Could you specify what you want to achieve? Super resolution? Img2Img?
@@outliier I want to generate thermal IR images conditioned on their respective RGB images . I know that in order to achieve this task i.e ; Image (RGB) to Image (Thermal IR) translation, I have to concat the input to U-net (which of course is thermal noise image ) with its corresponding RGB (condition image) and give this concatenated output as final input to the unet ; but problem is that I am not able to put this all together in the code (especially concatenating each RGB image (condition) from RGB image folder with its corresponding Thermal noise images so that I can pass the concatenated resultant image as final input to Unet as my aim is to generate RGB conditioned Thermal image using Diffusion.
lol training this on my ONLY "RTX 3090" :D
How can i increase the size of the generated image here?
is anyone find the DDPM Unet architecture figure, I can't find it
Hey, I am getting an error when i try to use one channel
"RuntimeError: Given groups=1, weight of size [64, 1, 3, 3], expected input[4, 3, 64, 64] to have 1 channels, but got 3 channels instead"
What can I do?
You need to change the input and output channels in the unet code
8:38 in the UNet section, how do you decide on the number of channels to set in both input and output to the Down and Up classes. Why just 64,128, etc. ?
People just go with powers of 2 usually. And usually you go to more channels in the deeper layers of the network.
@@outliier oh okay got it. Thank you so much for clearing that and for the video! I had seen so many videos / read articles for diffusion but yours were the best and explained every thing which others considered prerequisites!! Separating the paper explanation and implementation was really helpful.
People in Earth Observation know that images from Synthetic Aperture Radar have random speckles. People have tried removing the speckles using wavelets. I wonder how Denoising Diffusion would fare. The difficulty that I see is the need for x0 the un-noised image.
What do you think?
Sorry if I am misunderstanding, but at 19:10, shouldn't the code be:
"uncond_predicted_noise = model(x, t, None)" instead of "uncond_predicted_noise = model(x, labels, None)"
Also, according to the CFG paper's formula, shouldn't the next line be: "predicted_noise = torch.lerp(predicted_noise, uncond_predicted_noise, -cfg_scale)" under the definition of lerp?
One last question: have you tried using L1Loss instead of MSELoss? On my implementation, L1 Loss performs much better (although my implementation is different than yours). I know the ELBO term expands to essentially an MSE term wrt predicted noise, so I am confused as to why L1 Loss performs better for my model.
Thank you for your time.
Great videos by the way
Ah, I see you already fixed the first question in the codebase
thanks for the easiest implementation. could you plz tell us how to find FID and IS score for these images?
I think you would just sample 10-50k images from the trained model and then take 10-50k images from the original dataset and then calculate the FID and IS
@@outliier thanks
6:57 Why the formula is ... + torch.sqrt(beta) instead of calculated posterior variance like in paper?
Which paper are you referring to? In the first paper, you would just set the variance to beta and since you add the std * noise you take the sqrt(beta)
Hi! Can you please explain why the output is getting two stitched images?
What do you mean with two stitched images?
I think your code bugs when adjust image_size?
great video. please can you list the creators of the other helpful videos at 00:52? thanks
There are from Yannick Kilcher (on the right side), the one in the lower left is from AICoffeeBreak, the one in the top right corner is the first video that comes when you google „diffusion models explained“ and I forgot the middle one sorry. But shouldnt be hard to find
Hi, Thank you for the Video!
Can you please explain the test part:
n = 4
device = "cpu"
model = UNet_conditional(num_classes=4).to(device)
ckpt = torch.load(r"C:\Users
oueft\Downloads\Diffusion-Models-pytorch-V7\models\DDPM_conditional\ckpt.pt", map_location=torch.device('cpu'))
model.load_state_dict(ckpt)
diffusion = Diffusion(img_size=64, device=device)
y = torch.Tensor([6] * n).long().to(device)
x = diffusion.sample(model, n, y)
plot_images(x)
What is n, and why did the following error come up when I ran it?
ddpm_conditional.py", line 81, in sample
n = len(labels)
TypeError: object of type 'int' has no len()
Hi Sir, Good afternoon. i wanna run the ddpm_conditional for my ultrasound images dataset having 5 classes and all the images have equal sizes 256*256 and also images are greyscale images. but i am encountering this error. " RuntimeError: Given groups=1, weight of size [256, 1, 3, 3], expected input[4, 3, 256, 256] to have 1 channels, but got 3 channels instead". i already had a changing regarding the channel and the size
Hey, can you post your code on github and give the error?
How do you use this models to generate text to image?
You would need to train it on text-image pairs instead of label-image pairs as in the video. And you would need to scale up the model and dataset size to get some nice results