pub fn gather_qmm<'b>(
x: impl AsRef<Array>,
w: impl AsRef<Array>,
scales: impl AsRef<Array>,
biases: impl Into<Option<&'b Array>>,
options: GatherQmmOptions<'_>,
) -> Result<Array, QuantizationError>Expand description
Multiply selected input and packed-weight batches.
Pass affine biases or None for floating-point modes. Quantization options
must match the packed weights; indices select the respective batch dimensions.