For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo function

dequant_mxfp4

def dequant_mxfp4[*, SF_VECTOR_SIZE: Int = Int(32)](ctx: DeviceContext, output: TileTensor[Storage=output.Storage, address_space=output.address_space, linear_idx_type=output.linear_idx_type], input: TileTensor[Storage=input.Storage, address_space=input.address_space, linear_idx_type=input.linear_idx_type], scales: TileTensor[Storage=scales.Storage, address_space=scales.address_space, linear_idx_type=scales.linear_idx_type], num_rows: Int, num_cols: Int, pdl_level: PDLLevel = PDLLevel())

Dequantize MXFP4 packed weights to FP8 or BF16.

Parameters:

SF_VECTOR_SIZE (Int): Number of consecutive elements each E8M0 block scale covers (defaults to 32).

Args:

ctx (DeviceContext): Device context for kernel launch.
output (TileTensor[Storage=output.Storage, address_space=output.address_space, linear_idx_type=output.linear_idx_type]): Output tensor [num_rows, num_cols] of float8_e4m3fn or bfloat16.
input (TileTensor[Storage=input.Storage, address_space=input.address_space, linear_idx_type=input.linear_idx_type]): Input tensor [num_rows, num_cols // 2] of uint8 (packed FP4).
scales (TileTensor[Storage=scales.Storage, address_space=scales.address_space, linear_idx_type=scales.linear_idx_type]): Scale tensor [num_rows, num_cols // SF_VECTOR_SIZE] of float8_e8m0fnu.
num_rows (Int): Number of rows (N dimension for weights).
num_cols (Int): Number of columns (K dimension, unpacked).
pdl_level (PDLLevel): PDL optimization level for kernel launch.