Table 3.
Structure of the transformer model.
Model Layer | Input Vector | Output Vector | |
---|---|---|---|
Input layer | 1360 × 1 | 1360 × 1 | |
Token and position embedding | 1360 × 1 | 1360 × 512 | |
Transformer block | Multihead attention (head = 4) | 1360 × 512 | 1360 × 512 |
Feed forward network | 1360 × 512 | 1360 × 512 | |
Gloval average pooling | 1360 × 512 | 512 × 1 | |
Dense | 512 × 1 | 4 × 1 |