lamdas of loss have ever changed or adjust with training?

[diamondcm]

Loss

lambda_3d_velocity: 20.0
lambda_scale: 0.5
lambda_lv: 0.0
lambda_lg: 0.0
lambda_a: 0.0
lambda_av: 0.0

Hi again, My result got a little bit jitter, comparing to yours.
So those lambdas stay the same in your training?

[SoroushMehraban]

@diamondcm
Yes, we didn't play with these loss coefficients and it's the same as what MotionBERT used.