Questions about training with large-scale data and 3DGS rendering

[yqytomorrow]

Thank you very much for open-sourcing this excellent work.

I would like to kindly ask a few questions regarding training and rendering. I am interested in training the model on my own dataset, which is relatively large: around 400 images with a resolution of 4056×3040. Could you please share some guidance on the recommended hardware configuration (e.g., GPU memory, number of GPUs) to train such a dataset smoothly?

In addition, I would like to ask whether the rendered 3DGS results (e.g., the exported .ply files) are compatible with the Real-Time Viewer or if there are any specific requirements or conversion steps needed to visualize them correctly.

Thank you very much for your time and for making this great work publicly available. I would really appreciate any advice you could share.