Hi, I want to realize optimizer transform(Adam --> SGD) method during training, so I need some state parameters such as exp_avg… From the document https://pytorch.org/cppdocs/api/classtorch_1_1optim_1_1_optimizer.html#_CPPv4N5torch5optim9OptimizerE , I found state parameter from ska::flat_hash_map<std::string, std::unique_ptr<OptimizerParamState>> &state(), but how can I access it ?
Please help me, thanks in advance!