Merging Decision Transformers: Weight Averaging for Forming Multi-Task Policies
This work addresses the challenge of democratizing multi-task policy formation for robotics, though it is incremental as a preliminary step.
The paper tackles the problem of creating multi-task policies without centralized training by merging task-specific Decision Transformers through weight averaging on MuJoCo locomotion tasks, achieving functional multi-task models with methodological insights like using pre-trained initializations and Fisher information.
Recent work has shown the promise of creating generalist, transformer-based, models for language, vision, and sequential decision-making problems. To create such models, we generally require centralized training objectives, data, and compute. It is of interest if we can more flexibly create generalist policies by merging together multiple, task-specific, individually trained policies. In this work, we take a preliminary step in this direction through merging, or averaging, subsets of Decision Transformers in parameter space trained on different MuJoCo locomotion problems, forming multi-task models without centralized training. We also demonstrate the importance of various methodological choices when merging policies, such as utilizing common pre-trained initializations, increasing model capacity, and utilizing Fisher information for weighting parameter importance. In general, we believe research in this direction could help democratize and distribute the process that forms multi-task robotics policies. Our implementation is available at https://github.com/daniellawson9999/merging-decision-transformers.