elasticai.creator.nn.quantized_grads.fixed_point.two_complement_fixed_point_config

Module Contents

Classes

FixedPointConfigV2

This class behaves almost like a frozen instance, but calculates and sets minimum_as_rational_tensor and maximum_as_rational_tensor just once in post__init. After the post_init the setattr and getattr methods block the change of variables.

API

class elasticai.creator.nn.quantized_grads.fixed_point.two_complement_fixed_point_config.FixedPointConfigV2

This class behaves almost like a frozen instance, but calculates and sets minimum_as_rational_tensor and maximum_as_rational_tensor just once in post__init. After the post_init the setattr and getattr methods block the change of variables.

total_bits: int

None

frac_bits: int

None

__post_init__()

This python API call for dataclass is executed after init. This way we check if the FixedPointConfigV2 is valid while keeping the config immutable.

property minimum_as_rational

property maximum_as_rational

property resolution_per_int