DDPM-Pytorch

DDPM-Pytorch copied to clipboard

Question on the training processo of a Diffusion Model

4

Hi Sir, I'm working on a LDM for my thesis and your video was very helpful in figuring out how the DDPM works. I only have a doubt in the...

danielemolino

what changes would we need to do if we used our own dataset?

89

Thanks for the awesome explanation. Could you tell me which changes we need before training the model on our data?

awais00012

In Diffusion Models: Why does not use 'Positional Encoding' in self-attention layers?

1

Thx for nice practicing about DM. Actually, I'm really curious about why does not use 'Positional Encoding' (which was used in ViT or VanillaTransformer.. etc..) in self-attention layers? Is that...

lionking6792

what parameter changes would I need to make sure it runs on our dataset?

12

I am running this code on set of images but getting thisu error " CUDA out of memory. Tried to allocate 150.06 GiB (GPU 0; 15.89 GiB total capacity; 720.18...

Rushi117108

Sorry to bother the author again and again, I would like to ask a few additional questions: 1. How much video memory should I use with your model for my...

T02097

Deepa511

Hi author, when I trained my model using coloured datasets, I found that the memory required for training was too large, after adding multiple GPUs I switched to a data-parallel...

T02097

`self.t_emb_layers = nn.ModuleList([ nn.Sequential( nn.SiLU(), nn.Linear(t_emb_dim, out_channels) ) for _ in range(num_layers) ])` shouldn't it be following instead `self.t_emb_layers = nn.ModuleList([ nn.Sequential( nn.Linear(t_emb_dim, out_channels), nn.SiLU(), nn.Linear(out_channels, out_channels) ) for _...

ishan-modi

``` noise = torch.randn_like(im).to(device) t = torch.full((im.shape[0],), diffusion_config['num_timesteps']-1, device=device) #t = torch.randint(0, diffusion_config['num_timesteps'], (im.shape[0],)).to(device) xt = scheduler.add_noise(im, noise, t) for i in tqdm(reversed(range(diffusion_config['num_timesteps']))): # Get prediction of noise noise_pred =...

2000lf

RuntimeError:CUDA out of memory

4

Can I remove the attention layer for high resolution img?

2000lf