espnet2.layers.mixup_augmentation.MixupAugment

source

class espnet2.layers.mixup_augmentation.MixupAugment(mixup_probability: float)

Bases: Module

Mixup augmentation module for multi-label classification.

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(speech: Tensor, onehot: Tensor, speech_lengths: Tensor)

Applies mixup augmentation.

• Parameters:
  • speech – (Batch, Length)
  • onehot – (Batch, n_classes)
  • speech_lengths – (Batch,)
• Returns: (Batch, Length) onehot: (Batch, n_classes) speech_lengths: (Batch,): Minimum of the two lengths mixed.
• Return type: speech