Table 2. Encoder architecture parameters... | Emerald Publishing

Table 2.

Encoder architecture parameters

Component	Configuration
Embedding layers
Character embedding dim.	192
TQL embedding dim.	192
Positional encoding dim.	192
TQL encoder	2-layer MLP with SiLU
Positional encoding
Type	Sinusoidal (frozen)
Scaling coefficient $α$	Learnable (INIT: 1.0)
Transformer encoder
Number of layers	6
Self-attention hidden dim.	192
Feed-forward hidden dim.	768

Component	Configuration
Embedding layers
Character embedding dim.	192
TQL embedding dim.	192
Positional encoding dim.	192
TQL encoder	2-layer MLP with SiLU
Positional encoding
Type	Sinusoidal (frozen)
Scaling coefficient $α$	Learnable (INIT: 1.0)
Transformer encoder
Number of layers	6
Self-attention hidden dim.	192
Feed-forward hidden dim.	768

Source(s): Authors’ own work