Add a runtime factory for optimizers
Add a runtime factory for optimizers: a task in MiMo-V2.6-RL-harbor-code: MiMo-V2.6-RL Code (Harbor) (Harbor dataset). SINGA's C++ model library provides several gradient-descent optimizers (vanilla SGD with momentum, Nesterov, Adagrad and RMSProp). Today the only way to obtain one is to name the…
The task
SINGA's C++ model library provides several gradient-descent optimizers (vanilla SGD with momentum, Nesterov, Adagrad and RMSProp). Today the only way to obtain one is to name the concrete class at compile time. Code that drives training from a configuration or from a binding layer needs to pick the optimizer at…
Part of FineEnvs/MiMo-V2.6-RL-harbor-code.