MVTLoader and impact of mvt.layers on array size

[alaws-USGS]

Hello, I am trying to better understand what the MVTLoader does under the hood when an array of layer names is passed to mvt.layers.
Say for example the decoded format of my MVT tiles are as follows:

[{
"water": {...all the water layer info...},
"land": {...all the land layer info...},
"sky": {...all the land layer info...}
}]

And I passed ["water"] to mvt.layers, would the MVTLoader still load all three layers into the GPU or would it just load the "water" layer into the GPU (or at the very least, pass this data on to a library like deck.gl to handle loading). I'm trying to understand the memory footprint for constructing my own MVT tiles. Thanks!