A Cooperative Lightweight Translation Algorithm Combined with Sparse-ReLU

Table 2

CNN submodel size.

Description	Substructure	Layer name	Size
Pretreatment	Embedding and location coding	Embedding	(10000, 512)
CNN	Convolution submodel	LayerNorm	eps = 1e-05
		Conv2d	in_ch = 1, out_ch = 512, kernel = (5, 512), stride = (1, 1), pad = (2, 0)
		Linear	(512, 512)
	Activation function	Sparse-ReLU	Sparse-ReLU ：a = 0.25,b = 1,c = 0.1,d = 0.4