torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 76.40 GiB (GPU 0; 8.00 GiB total capacity; 365.94 MiB already allocated; 6.55 GiB free; 400.00 MiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

[wsAndy]

in windows, I use docker image and render the trained ply by the following command:

set TrainOutputPath="E:\docker\data\tandt_db\tandt\train"
set ViewerPath="./"
docker run --name  viewer_97f459dee22843aaa728ded6999cf6ab -it --rm --gpus all -p 8080:8080 -v %ViewerPath%:/root/code -v %TrainOutputPath%:/root/data wsandy95/cuda:11.1-devel-ubuntu2204-nerfviewer-ffmpeg  /bin/bash  -c "source ~/miniconda3/bin/activate && conda init bash && conda activate gspl  && cd /root/code && python viewer.py  /root/data/output" 

In Nvidia RTX 2080（6G），I get the error:

Traceback (most recent call last):
  File "/root/code/internal/viewer/client.py", line 121, in run
    self.render_and_send()
  File "/root/code/internal/viewer/client.py", line 92, in render_and_send
    image = self.renderer.get_outputs(camera, scaling_modifier=self.viewer.scaling_modifier.value)
  File "/root/code/internal/viewer/renderer.py", line 18, in get_outputs
    return self.renderer(
  File "/root/miniconda3/envs/gspl/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
    return forward_call(*args, **kwargs)
  File "/root/code/internal/renderers/vanilla_renderer.py", line 102, in forward
    rendered_image, radii = rasterizer(
  File "/root/miniconda3/envs/gspl/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
    return forward_call(*args, **kwargs)
  File "/root/miniconda3/envs/gspl/lib/python3.9/site-packages/diff_gaussian_rasterization/__init__.py", line 210, in forward
    return rasterize_gaussians(
  File "/root/miniconda3/envs/gspl/lib/python3.9/site-packages/diff_gaussian_rasterization/__init__.py", line 32, in rasterize_gaussians
    return _RasterizeGaussians.apply(
  File "/root/miniconda3/envs/gspl/lib/python3.9/site-packages/torch/autograd/function.py", line 506, in apply
    return super().apply(*args, **kwargs)  # type: ignore[misc]
  File "/root/miniconda3/envs/gspl/lib/python3.9/site-packages/diff_gaussian_rasterization/__init__.py", line 92, in forward
    num_rendered, color, radii, geomBuffer, binningBuffer, imgBuffer = _C.rasterize_gaussians(*args)
torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 76.40 GiB (GPU 0; 8.00 GiB total capacity; 365.94 MiB already allocated; 6.55 GiB free; 400.00 MiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

But it could work in Nvidia RTX 4090 (24GB)

[yzslab]

May be 6GB is not enough for your scene. Try to smaller the initial resolution of the web viewer manually here:

https://github.com/yzslab/gaussian-splatting-lightning/blob/8c673b1cdd6df6a3e8231316c6fb47731b9a9b89/viewer.py#L326

https://github.com/yzslab/gaussian-splatting-lightning/blob/8c673b1cdd6df6a3e8231316c6fb47731b9a9b89/viewer.py#L343

[svsambandam]

i also get the same error when running even when setting initial_value as 256: main.py fit --config configs/deformable_blender.yaml --data.path /deformable-data/dnerf/lego I am running on a RTX 3090 with 24GB. Do you know where I might be going wrong :) @yzslab