A lightweight convolution block as described in ... https://arxiv.org/abs/1901.10430 Reference PyTorch Implementation (L587): https://github.com/pytorch/fairseq/blob/v0.6.2/fairseq/models/lightconv.py Args: name: name of the params input_dim: Input dimension. ker
(self,
name,
input_dim,
kernel_size,
activation='RELU',
is_causal=False,
glu_with_tanh=False,
residual_dropout_prob=0)
source not stored for this graph (policy: none)
nothing calls this directly
no test coverage detected