Any suggestion to generate 1-D vectors using diffusion model about denoising-diffusion-pytorch HOT 7 CLOSED

For previous diffusion models with 1D data, such as spatio-temporal sequences and 1D vector graphs, I think you need to at least modify the channel of Unet.

from denoising-diffusion-pytorch.

spatial-temporal sequences should be 2D vectors instead of 1D vectors, and there is no time dimension or size dimension in the 1D vectors I am talking here

from denoising-diffusion-pytorch.

Hi, have you figured out a way to do 1-D vector training and generation?

from denoising-diffusion-pytorch.

Hi, have you figured out a way to do 1-D vector training and generation?

Is there any solution for this problem?
In my case , each data is a 240-dim vector, [x1,y1,x2,y2, ..., x_N,y_N]

from denoising-diffusion-pytorch.

it is here, in the readme https://github.com/lucidrains/denoising-diffusion-pytorch#1d-sequence

from denoising-diffusion-pytorch.

it is here, in the readme https://github.com/lucidrains/denoising-diffusion-pytorch#1d-sequence

However, I can't understand some number in the code.
training_seq = torch.rand(64, 32, 128)

model = Unet1D(
    dim = 64,
    dim_mults = (1, 2, 4, 8),
    channels = 32
)

I have 1w samples, each of which is a 240-dim vector, [x1,y1,x2,y2, ..., x_N,y_N].

I don't know how to modify the the channels of Unet1d (maybe set channels=1?) and have no idea about the specific meaning of "64, 32, 128" for training_seq.

from denoising-diffusion-pytorch.

@fikry102 i recommend joining a beginner machine learning community online

from denoising-diffusion-pytorch.