skainet-io-gguf/sk.ainet.io.gguf.llama/QuantizedTensorFactory/toQ8_0

toQ8_0

fun toQ8_0(rawTensor: Tensor<Int8, Byte>, logicalShape: Shape): Q8_0TensorData(source)

Convert a raw byte tensor to Q8_0TensorData.

Return

Q8_0TensorData ready for quantized matmul

Parameters

rawTensor

Tensor containing raw Q8_0 bytes (loaded with RAW_BYTES policy)

logicalShape

The logical shape in elements (not bytes/blocks)

fun toQ8_0(rawTensor: Tensor<Int8, Byte>): Q8_0TensorData(source)

Convert a raw byte tensor to Q8_0TensorData using the tensor's shape.

Note: For Q8_0, the raw tensor shape should be the logical element shape, not the packed byte count.