pub fn dequantize_tensor(int8: &[i8], params: &QuantParams) -> Vec<f32>
Dequantize INT8 tensor to FP32