readValues
open override fun readValues(layer: Int, startPos: Int = 0, endPos: Int = currentSeqLen): FloatArray(source)
Read cached values for a layer, dequantized to float.
Returns the value cache for positions [startPos, endPos) as a contiguous float array shaped numHeads, (endPos - startPos), headDim.
Parameters
layer
Layer index
startPos
First token position (inclusive)
endPos
Last token position (exclusive), defaults to currentSeqLen