Float8

For conceptual model usage and type mapping, see:

• Model

class pyrogue.Float8(bitSize)

Model class for 8-bit E4M3 floating point numbers (NVIDIA FP8).

Parameters:

• bitSize (int) – Number of bits being represented. Must be 8.

• args (Any)

• kwargs (Any)

Return type:

Any

Notes

Format: 1 sign bit, 4 exponent bits, 3 mantissa bits (E4M3). Bias = 7. No infinity representation. NaN encoded as 0x7F. Maximum representable value is 448.0. Supported by NVIDIA Hopper (H100) and Blackwell GPUs.

pytype

alias of float

toBytes(value)

Convert float to 1-byte E4M3 encoding.

Return type:

bytearray

Parameters:

value (float)

fromBytes(ba)

Decode 1-byte E4M3 encoding to float.

Return type:

float

Parameters:

ba (bytes)

fromString(string)

Parse a string into a float value.

Return type:

float

Parameters:

string (str)

minValue()

Return the minimum representable value (-448.0).

Return type:

float

maxValue()

Return the maximum representable value (448.0).

Return type:

float

class pyrogue.Float8BE(bitSize)

Model class for 8-bit E4M3 floats stored as big endian.

Parameters:

• args (Any)

• kwargs (Any)

Return type:

Any