search

sail-sg / Consistent3D

The official PyTorch implementation of Consistent3D (CVPR 2024)
MIT License
62 stars 0 forks source link

Additional Loss term compared with paper #4

Closed zcai0612 closed 3 months ago

zcai0612 commented 3 months ago

May I ask about the first term of the loss function? (sigma_t / (sigma_t_next + 1e-9)) * (pred_epsilon - noise_pred) It seems that it always equals zero.

https://github.com/sail-sg/Consistent3D/blob/9118a5cc8ba4d16aa1c092301aedffb68c31f18e/threestudio/models/guidance/stable_diffusion_guidance.py#L394

nickwzk commented 3 months ago

It is borrowed from Common Diffusion Noise Schedules and Sample Steps Are Flawed. You can optionally activate it by setting _"recon_stdrescale" to non-zero integers. However, we found it not suitable in our work, so we set this rescale to zero as default.