BFloat16

For conceptual model usage and type mapping, see:

• Model

class pyrogue.BFloat16(bitSize)

Model class for 16-bit Brain Float (BFloat16) numbers.

Parameters:

• bitSize (int) – Number of bits being represented. Must be 16.

• args (Any)

• kwargs (Any)

Return type:

Any

Notes

Format: 1 sign bit, 8 exponent bits, 7 mantissa bits (same exponent as float32). Bias = 127. Supports infinity and NaN. Maximum representable finite value is approximately 3.39e38 (same range as float32). Supported by NVIDIA Ampere (A100), Hopper (H100), and Blackwell GPUs.

pytype

alias of float

toBytes(value)

Convert float to 2-byte BFloat16 encoding.

BFloat16 is the upper 16 bits of the float32 bit pattern. All special values (NaN, infinity, zero, subnormals) are preserved.

Return type:

bytearray

Parameters:

value (float)

fromBytes(ba)

Decode 2-byte BFloat16 encoding to float.

Reconstructs float32 by shifting the 16-bit pattern left by 16.

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 finite BFloat16 value (~-3.39e38).

Return type:

float

maxValue()

Return the maximum representable finite BFloat16 value (~3.39e38).

Return type:

float

class pyrogue.BFloat16BE(bitSize)

Model class for BFloat16 floats stored as big endian.

Parameters:

• args (Any)

• kwargs (Any)

Return type:

Any