Function dequantize_4bit

Dequantizes FP4 blockwise quantized values. Dequantizes the tensor A with maximum absolute values absmax in blocks of size blocksize. Parameters ---------- A : torch.Tensor The input tensor (packed 4-bit values). quant_state : QuantState object with quantis

(
    A: Tensor,
    quant_state: Optional[QuantState] = None,
    absmax: Optional[torch.Tensor] = None,
    out: Optional[torch.Tensor] = None,
    blocksize: int = 64,
    quant_type="fp4",
)