Can we continue training on the basis of 105000.ckpt?

cvlab-columbia / zero123

Zero-1-to-3: Zero-shot One Image to 3D Object (ICCV 2023)

https://zero123.cs.columbia.edu/

MIT License

2.59k stars 188 forks source link

Open TimenoLong opened 4 months ago

TimenoLong commented 4 months ago

May I ask if you can publish the yaml file of 105000.ckpt to us so that we can train on the model you have already trained?

stvn037 commented 4 months ago

i concur--is there a way to finetune models? Is there any chance z123 will ever support LoRa's or some similar?