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

sigmoid

def sigmoid[dtype: DType, simd_width: Int, accum: DType = get_accum_type[dtype]()](x: SIMD[dtype, simd_width]) -> SIMD[dtype, simd_width]

Compute the sigmoid activation using the equation $1 / (1 + e^{-x})$ .

The computation is performed in a higher-precision accumulation type (see get_accum_type) for low-precision inputs and cast back to dtype, to match the numerics of the graph-level ops.sigmoid implementation.

Constraints:

Type must be a floating point Dtype.

Parameters:

dtype (DType): DType used for the computation.
simd_width (Int): SIMD width used for the computation.
accum (DType): Higher-precision accumulation dtype used internally; defaults to get_accum_type[dtype]().

Args:

x (SIMD[dtype, simd_width]): The value to compute the sigmoid operation on.

Returns:

SIMD[dtype, simd_width]: The result of the sigmoid operation.