To　satisfy　the　differentiated　service　requirements　of　delay-sensitive　and　computing-intensive　tasks　in　unmanned　aerial　vehicle　(UAV)　networks,　it　is　urgent　to　efficiently　allocate　limited　network　resources　to　improve　network　performance.　In　this　paper,　we　propose　an　intelligent　task　offloading　scheme　to　optimize　resource　allocation　in　UAV　networks　with　content　caching.　Specifically,　we　formulate　the　joint　optimization　of　task　offloading　and　resource　allocation　as　a　latency　minimization　model　for　the　caching-assisted　UAV　system.　Then,　a　new　deep　reinforcement　learning　(DRL)　algorithm　is　designed　to　make　offloading　and　resource　allocation　decisions　based　on　current　network　state　information,　significantly　improving　resource　utilization.　Numerical　results　indicate　that　the　model　significantly　reduces　network　latency　in　comparison　to　its　existing　benchmarks　in　caching-assisted　UAV　networks.