fp8_quant_blockwise

paddle.incubate.nn.functional. fp8_quant_blockwise ( x: Tensor, epsilon: float = 0.0, input_transpose: bool = False, output_scale_transpose: bool = True, using_pow2_scale: bool = True, quant_method: str = '1x128', output_type: str = 'e4m3', name: Optional[str] = None ) [source]