Reference

dnadiffusion

dnadiffusion.utils.train_util

data

dnadiffusion.data.dataloader

SequenceDataset

Bases: Dataset

__getitem__(index)

Generates one sample of data

__init__(seqs, c, transform=T.Compose([T.ToTensor()]))

Initialization

__len__()

Denotes the total number of samples

models

dnadiffusion.models.layers

LearnedSinusoidalPosEmb

Bases: Module

following @crowsonkb 's lead with learned sinusoidal pos emb

ResBlock

Bases: Module

Iniialize a residual block with two convolutions followed by batchnorm layers

dnadiffusion.models.diffusion

dnadiffusion.models.unet