AttributeError: 'TensorVMSplit_TimeEmbedding' object has no attribute 'module'

[Kane-LAU]

THANKS for your contribution.but I have some issues when I run your code on my machine(UBUNTU 20.04 ,RTX 4090)

python train.py --config configs/Nvidia.txt

When I was training, I noticed that it was very slow, taking about 1 hour to train only 5%, and it also threw an error at 5%. The error message was as follows:

how can i fix it？

ps：I want to know if there are any acceleration techniques available for Nerf, such as Nerfacc, or if there are any additional packages that need to be installed

[balazon]

Same thing here with a solution https://github.com/facebookresearch/robust-dynrf/issues/5#issuecomment-1614292291

i delete the module in renderer.py in line 488 like this and it works, But I don't know if this will affect the results scene_flow_f, scene_flow_b = tensorf.get_forward_backward_scene_flow( pts_ref, ts_chunk )

[Kane-LAU]

Same thing here with a solution #5 (comment)

It worked!!! thanks a lot ,but i came another issue , I suceeded to pass 5% but fail in 50% 😂

[balazon]

I've never gotten that far because of memory problems :) But that should be easy to fix, some if condition missing, like args.optimize_poses (and the same for optimize_focal_length) https://github.com/facebookresearch/robust-dynrf/blob/c56be705940c488cdfeac73d51c332d75101d60c/train.py#L2592C1-L2595C68