Viet-Anh on Software Logo

What is: FiLM Module?

SourceWaveGrad: Estimating Gradients for Waveform Generation
Year2000
Data SourceCC BY-SA - https://paperswithcode.com

The Feature-wise linear modulation (FiLM) module combines information from both noisy waveform and input mel-spectrogram. It is used in the WaveGrad model. The authors also added iteration index nn which indicates the noise level of the input waveform by using the Transformer sinusoidal positional embedding. To condition on the noise level directly, nn is replaced by αˉ\sqrt{\bar{\alpha}} and a linear scale C=5000C = 5000 is applied. The FiLM module produces both scale and bias vectors given inputs, which are used in a UBlock for feature-wise affine transformation as:

γ(D,αˉ)U+ζ(D,αˉ)\gamma\left(D, \sqrt{\bar{\alpha}}\right) \odot U + \zeta\left(D, \sqrt{\bar{\alpha}}\right)

where γ\gamma and ζ\zeta correspond to the scaling and shift vectors from the FiLM module, DD is the output from corresponding DBlock, UU is an intermediate output in the UBlock.