search

Nightmare-n / GD-MAE

GD-MAE: Generative Decoder for MAE Pre-training on LiDAR Point Clouds (CVPR 2023)
Apache License 2.0
114 stars 6 forks source link

about self-training log #8

Closed ShuojieMo closed 1 year ago

ShuojieMo commented 1 year ago

Could you provide the log file for GDMAE self-supervised training? What is the Chamfer Distance loss approximately when it reaches convergence?

Nightmare-n commented 1 year ago

train_loss Here is the self-supervised training loss, where the loss seems to converge to 0.3 (30 epochs).