Training details for NeuMan dataset

[lxxue]

Hi,

Thanks for the great work and for sharing the well-written code. I want to train a model on my own video with similar configurations as done in the paper. I wonder if you use the default configuration (e.g. 4096 rays/batch, 128 samples/ray, 1M iterations) for training all the sequences? It seems that it takes longer than 2 days to train the background model with such configurations on an RTX 3090.