search

syguan96 / NeuroFluid

[ICML 2022] NeuroFluid: Fluid Dynamics Grounding with Particle-Driven Neural Radiance Fields
83 stars 10 forks source link

environment issue #8

Open taehoon-yoon opened 1 year ago

taehoon-yoon commented 1 year ago

Hi, Thanks for your project. I carefully followed installation guide which you've provided, after that I tried to evaluate Neuro fluid using python eval_e2e.py --resume_from $MODEL_PATH --dataset DATASET_NAME But I met error like following.

Traceback (most recent call last):
  File "/home/taehoon/Desktop/NeuroFluid/tmp.py", line 2, in <module>
    import open3d.ml.torch as ml3d
  File "/home/taehoon/anaconda3/envs/fluid-env/lib/python3.7/site-packages/open3d/ml/torch/__init__.py", line 42, in <module>
    _torch.__version__))
Exception: Version mismatch: Open3D needs PyTorch version 1.12.*, but version 1.13.1+cu116 is installed!

It seems like Open3d does not support pytorch version 1.13.1+cu116? So my question is, can I know your pytorch version used for this project?

Thanks!

syguan96 commented 1 year ago

Hi, the pytorch wheel we use is downloaded from open3d_downloads. See this page: https://github.com/isl-org/open3d_downloads/releases?page=2.

From this issue (https://github.com/isl-org/open3d_downloads/issues/2), it looks like that Open3d has moved to torch_1.12. But I have never tried this.

taehoon-yoon commented 1 year ago

Sorry for the late reply. I've managed to solve the environmental issue. Anyway I was walking through your code and found out at end2end training step, according to end2end.yaml, it seems that to train end2end it requires '/data/syguan/fluid_simulation/release/ckpts/trans/transv3_694999.pt' . According to your paper, you've used pretrained Deep Lagrangian Fluid model, so is transv3_694999.pt equal to pretrained_model_weights.pt?

Thanks again!