Finetuning TTT3R

[kennethyangg]

Hi, thanks for the great work on TTT3R.

I have two questions regarding fine-tuning:
1. Fine-tuning from CUT3R
Would it be possible to release the code or instructions for fine-tuning TTT3R from a CUT3R checkpoint?
2. Estimated resource requirements
Could you share the approximate training cost for fine-tuning TTT3R, such as:

•	GPU type / number of GPUs
•	batch size
•	approximate training time
•	peak GPU memory usage

Thanks a lot.