Results
13
comments of
Zebin Xing
@fengjiang5 Hi, you can refer to [issue29](https://github.com/YvanYin/GoalFlow/issues/29) for details.
@smilebetterworld We found that applying **force_dropout** during training improves the robustness of the model. However, during flow sampling, it can lead to overly divergent trajectories, which in turn degrade the...
In training