flashinfer.quantization.NVFP44Over6ErrMode

class flashinfer.quantization.NVFP44Over6ErrMode(value)

Error metric that picks between the two 4over6 block-scale candidates.

Each 16-element block is quantized with both the standard amax / 6 scale and the tighter amax / 4 scale, dequantized, and compared with the source. MAE sums absolute errors, MSE sums squared errors. The tighter candidate clips the block’s largest magnitudes, which MSE penalizes harder, so MSE selects it less often.

__init__()

Methods

conjugate	Returns self, the complex conjugate of any int.
bit_length()	Number of bits necessary to represent self in binary.
bit_count()	Number of ones in the binary representation of the absolute value of self.
to_bytes(length, byteorder, *[, signed])	Return an array of bytes representing an integer.
from_bytes(byteorder, *[, signed])	Return the integer represented by the given array of bytes.
as_integer_ratio()	Return integer ratio.

Attributes

real	the real part of a complex number
imag	the imaginary part of a complex number
numerator	the numerator of a rational number in lowest terms
denominator	the denominator of a rational number in lowest terms
MAE
MSE