Neural electric bass guitar synthesis framework enabling attack-sustain-representation-based technique control

EURASIP Journal on Audio, Speech, and Music Processing

Table 3 The configuration of the alignment model

Encoder layer	3
Multi-head attention heads	2
Multi-head attention hidden units	64
Feed-forward network hidden units	256
Feed-forward network kernel sizes	9, 1
Max sequence	1000
Max time	96
Max duration	96