Conditional Cross Attention Network for Multi-Space Embedding without Entanglement in Only a SINGLE Network