The　existing　link　prediction　researches　of　information　networks　mainly　focus　on　the　dynamic　homogeneous　network　or　the　static　heterogeneous　network.　It　has　always　been　a　challenge　to　predict　future　relationships　between　nodes　while　learning　both　continuous-time　and　heterogeneous　information　simultaneously.　In　this　paper,　we　propose　a　Heterogeneous　and　Continuous-Time　Model　Based　on　Self-Attention　(HTAT)　to　complete　the　link　prediction　task　by　learning　temporal　evolution　and　heterogeneity　jointly.　The　HTAT　model　consists　of　the　base　layer　and　the　heterogeneous　layer.　The　base　layer　incorporates　a　functional　time　encoding　with　self-attention　mechanism　to　capture　continuous-time　evolution.　And　the　heterogeneous　layer　consists　of　multi-view　attention　to　learn　heterogeneous　information.　Experimental　results　show　that　HTAT　is　significantly　competitive　compared　with　four　state-of-the-art　baselines　on　three　real-world　datasets.