Open neauniko opened 7 months ago
Hello, I have a question about the regression loss. Is it feasible to constrain the offsets of the start and end offsets using L1 loss?
We have not tried L1 loss yet, but from the code perspective, it seems relatively easy to implement, as the model ultimately obtains predicted offset values and target offset values.
Hello, I have a question about the regression loss. Is it feasible to constrain the offsets of the start and end offsets using L1 loss?