HF RL Explorer

I want a hybrid.decomposers.SublatticeDecomposer(seed=None, runopts) runnable that I can construct once…

I want a hybrid.decomposers.SublatticeDecomposer(seed=None, runopts) runnable that I can construct once…: a task in MiMo-V2.6-RL-harbor-code: MiMo-V2.6-RL Code (Harbor) (Harbor dataset). For a concrete session, if the BQM has tuple variables (0,) , (1,) , (2,) , linear biases {(0,): 0, (1,): 1…

The task

For a concrete session, if the BQM has tuple variables `(0,)`, `(1,)`, `(2,)`, linear biases `{(0,): 0, (1,): 1, (2,): 2}`, quadratic biases `{((0,), (1,)): 3, ((1,), (2,)): 4}`, the best binary sample assigns `(0,) = 1`, and the state has `problem_dims=(3,)`, `geometric_offset=(1,)`, `origin_embedding_index=0`, and…

Part of FineEnvs/MiMo-V2.6-RL-harbor-code.