Skip to main content
yangyu12
Projects
motion_lm
Reports
MotionLM复现实验
Log in
Sign up
Share
Comment
Star
Share
Comment
Star
MotionLM复现实验
Yu Yang
Created on February 7
|
Last edited on February 20
Comment
v0.2 (大decoder) 比 v0.3（小decoder）在相同训练步数下的性能更好
train/traj_acc
train/traj_acc
2k
4k
6k
8k
Step
0
0.1
0.2
0.3
0.4
0.5
0.6
train/token_acc
train/token_acc
2k
4k
6k
8k
Step
0.4
0.5
0.6
0.7
train/loss
train/loss
2k
4k
6k
8k
Step
0.6
0.8
1
1.2
1.4
1.6
1.8
2
waymo_pred/min_fde
waymo_pred/min_fde
200k
400k
600k
800k
trainer/global_step
3.2
3.4
3.6
3.8
waymo_pred/min_ade
waymo_pred/min_ade
200k
400k
600k
800k
trainer/global_step
1.3
1.35
1.4
1.45
1.5
Run set
3
Add a comment