kobzaond/RLVRAMBench
RLVRAMBench Which language-model training configurations can I use with the memory I have, and how much testing does that decision require? RLVRAMBench is a measurement dataset with open evaluation tasks for a specific language-model training system. It measures memory feasibility when response generation and reinforcement-learning updates share the same graphics processors. It provides measured outcomes, fixed prediction tasks, a budgeted decision replay, reference methods, and… See the full description on the dataset page: https://huggingface.co/datasets/kobzaond/RLVRAMBench.
0223
1version https://git-lfs.github.com/spec/v12oid sha256:9b28d833742a1befddc4b55b67968fb449d469e95b57f3451adeb509a59098653size 3984864 