The　goal　of　bridge　design　is　to　design　and　construct　an　elegant　and　safe　bridge　that　meets　all　functional　requirements　at　an　acceptable　cost.　Topology　optimization　has　emerged　as　a　powerful　tool　for　innovative　bridge　design.　However,　the　inherent　drawback　of　high　computational　cost　in　current　topology　optimization　remains　unresolved.　Moreover,　relevant　research　that　mainly　focuses　on　a　single　objective　fails　to　meet　the　safety　and　usability　requirements　of　the　bridge.　Therefore,　in　this　paper,　an　iteration-free　multi-objective　topology　optimization　(MOTO)　method　based　on　deep　learning　is　proposed　in　this　paper,　which　can　not　only　improve　the　computational　efficiency　of　MOTO,　but　also　meet　the　different　structural　demands.　Firstly,　a　MOTO　mathematical　model　based　on　constraint　programming　method　was　established,　and　the　dataset　with　stiffness,　strength　and　material　consumption　as　optimization　objectives　was　constructed.　Secondly,　based　on　the　self-attention　mechanism,　a　novel　network　is　constructed,　and　trained.　Finally,　the　performance　of　the　proposed　method　was　evaluated　in　terms　of　efficiency　and　accuracy.　Results　show　that　a　significant　reduction　in　computation　cost　of　MOTO　problem　was　achieved　with　little　sacrifice　on　the　performance　of　design　solutions.　The　proposed　method　can　provide　new　ideas　for　multi-objective　design　of　bridges　and　promote　the　improvement　of　bridge　design　level.　©　2024　The　Author(s).