Model Architectures¶

All architectures are stateless nn.Module classes. Training logic lives in dnanet.modules.

U-Net¶

from dnanet.models.unet import UNet, DoubleConv, EncoderBlock, DecoderBlock

Autoencoders¶

from dnanet.models.autoencoder import (
    Conv1dAutoencoder,
    PerDyeConv1dAutoencoder,
    SharedWeightPerDyeConv1dAutoencoder,
    UNet2DAutoEncoder,
    FourierAutoencoder,
)

Peak Classifier¶

from dnanet.models.peak_classifier import PeakClassificationModel

Combined Classifiers (PeakNet)¶

from dnanet.models.peaknet import (
    CombinedClassifier,
    PeakOnlyClassifier,
    MLPCombiner,
    FiLMCombiner,
    CrossAttentionCombiner,
)

Loss Functions¶

from dnanet.models.loss import DiceLoss, FocalLoss