Fine-tuning in practice

LoRA with rank $r = 16$ is applied to the gate, up, and down projections of every layer in a synthetic model with $d = 4096$, $d_{ff} = 11008$, and 32 layers. Attention is not targeted. How many trainable parameters?