pyngp, ImportError on pyngp.cpython-37m-x86_64-linux-gnu.so: undefined symbol

[smtabatabaie]

I'm trying to run it on Google Colab, and I'm following this guide:
https://github.com/NVlabs/instant-ngp/blob/master/notebooks/instant_ngp.ipynb
Everything builds without any errors, when I try to run the run.py code (which is the 8th step of the mentioned link) I get this error:

Traceback (most recent call last):
  File "scripts/run.py", line 29, in <module>
    import pyngp as ngp # noqa
ImportError: /content/instant-ngp/build/pyngp.cpython-37m-x86_64-linux-gnu.so: undefined symbol: _ZNK3ngp7Testbed16crop_box_cornersEb

I saw this issue in which the compiler couldn't find pyngp and there was a suggestion to specify the build folder #43 , so I also tried adding the lines to specify the build folder in run.py, but still no luck.

pyngp_path = '\build'
sys.path.append(pyngp_path)
import pyngp as ngp  # this line should already be there near the top of the code

I'd appreciate it if someone could help me with this.

[domef]

I'm having the same problem.

[domef]

I just found out that instant-ngp/build/testbed works.

[smtabatabaie]

I just found out that instant-ngp/build/testbed works.

Thanks @domef , you mean I should set my path to build/testbed in run.py?

[domef]

No, I mean that you can directly run instant-ngp/build/testbed, it is the equivalent of run.py but compiled from c++. (it should have about the same options). With testbed I'm able to train nerf on colab.

[domef]

I see now that testbed has many less options than run.py, but at the moment I think it's a good fallback.

Edit: there are few options in testbed , so actually it's pretty hard to work in headless mode. I can't even save a checkpoint or set the number of steps.

[smtabatabaie]

That's unfortunate, If @andreykramer who's wrote the guide could help us it would be great.

[domef]

I think I will try to compile some older versions of instant-ngp and see if they work.

[domef]

I managed to run run.py and train on the fox dataset in headless mode by checking out the commit add notebook with guide for execution in colab : git checkout 50ed131f0e296801d4b3d09ff58295add2af42fa.

[smtabatabaie]

T

I managed to run run.py and train on the fox dataset in headless mode by checking out the commit add notebook with guide for execution in colab : git checkout 50ed131f0e296801d4b3d09ff58295add2af42fa.

Thanks very much, saved my day!

[taruntiwar]

Can you please explain to me how you managed to run run.py in colab notebook
Thank you

[domef]

I followed the instructions in https://github.com/NVlabs/instant-ngp/blob/master/notebooks/instant_ngp.ipynb with the following branch 50ed131.

[domef]

I just verified that the last working commit is 25dec33c253f035485bb2e1f8563e12ef3134e8b.

[andreykramer]

I just verified that the last working commit is 25dec33c253f035485bb2e1f8563e12ef3134e8b.

Thanks, this sure will be helpful to find out what's wrong.

That's unfortunate, If @andreykramer who's wrote the guide could help us it would be great.

Unfortunately I don't know this repo deeply enough to have a clue of what may have broken the compilation nor how to fix it in the guide.

[sumanttyagi]

i am also facing the same issue please guide me how to solve this problem ? to train the model ?

[domef]

I followed the instructions in https://github.com/NVlabs/instant-ngp/blob/master/notebooks/instant_ngp.ipynb with the following branch 50ed131.

Follow the instructions specified in the notebook but use the same commit I used.

[saisurbehera]

I get the same error with python with headless even with the branch

[akbazuka]

Likewise, just tried to run it today and got issues with both of the suggested branches.

[SamDM]

Hi, I made a PR which fixes this issue, in the meantime, see here: #991