Skip to content

absmax_fp8

Classes

fastvideo.layers.quantization.absmax_fp8.AbsMaxFP8Config 

AbsMaxFP8Config()

Bases: QuantizationConfig

Config class for absmax float8_e4m3fn quantization. Currently only support per-tensor quantization.

Source code in fastvideo/layers/quantization/base_config.py 
64
65
66
67
def __init__(self) -> None:
    super().__init__()
    # mapping is updated by models as they initialize
    self.packed_modules_mapping: dict[str, list[str]] = dict()

fastvideo.layers.quantization.absmax_fp8.AbsMaxFP8LinearMethod

Bases: LinearMethodBase

Linear method with AbsMax FP8 quantization.

Functions: