Efficient Post-training Quantization with FP8 Formats