Cannot find reference 'encoding' in 'nerf_helper'.py'.

[ZZy129326999]

Dear export, hello! I want to learn about the details of "Cannot find reference 'encoding' in 'nerf_helper'.py'. ". I saw the description of pybind11 from CUDA EXTENSION. But there is a reference information "Cannot find reference 'encoding' in 'nerf_helper'.py'. " in model.py when running. Thank you very much!

model.py
from nerf_helper import encoding

nerf_extension.cu
PYBIND11_MODULE (TORCH_EXTENSION_NAME, nerf_helper)
{
  	nerf_helper.def ("comservativeSampling", &cudaSampler, conservative_sampler_docs.c_str());
  	nerf_helper.def ("sampling", &easySampler, sampler_docs.c_str());
	nerf_helper.def ("encoding", &positionalEncode, pe_docs.c_str());  // This !! Thank you !!!
	nerf_helper.def ("invTransformSample", &inverseTransformSample, inv_tf_docs.c_str());
	nerf_helper.def ("invTransformSamplePt", &inverseTransformSamplePt, inv_tf_pt_docs.c_str());
	nerf_helper.def ("imageSampling", &imageSampler, image_sampler_doc.c_str());
	nerf_helper.def ("validSampling", &validSampler, valid_sampler_doc.c_str());
}

[Enigmatisms]

There is no nerf_helper.py. nerf_helper as you see, is defined by setup.py in folder cuda/ (line 10). The module is called nerf_helper and it is already compiled, dynamically linked and loaded by your executable script. Therefore there is no python definition for encoding. Function encoding is "declared" in python, and defined in CUDA c++ (plz refer to file cuda/pe.cu). The definition is linked to the name encoding.

So, if python interpreter can not fine encoding, it might be (if I got it right):

• You didn't correctly compile pytorch CUDA extension (follow README.md, running setup.py), so the interpreter had no where to find the definition of this function. Solution is to compile CUDA extension correctly (normally, successful compilation will result in .egg .dist(?maybe) files and it will tell you Finished ... (I don't remember) in the terminal).

[ZZy129326999]

Thank you dear export!! I could generate nerf_helper.py after running cuda.

nerf_helper.py
def __bootstrap__():
    global __bootstrap__, __loader__, __file__
    import sys, pkg_resources, importlib.util
    __file__ = pkg_resources.resource_filename(__name__, 'nerf_helper.cpython-39-x86_64-linux-gnu.so')
    __loader__ = None; del __bootstrap__, __loader__
    spec = importlib.util.spec_from_file_location(__name__,__file__)
    mod = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(mod)
__bootstrap__()

Is this not normal? Thank you very much!

[Enigmatisms]

I set alias python = python3. Therefore, make sure you use python3.
My env: python 3.8 + Ubuntu 20.04, CUDA 11.3. Compilation does not yield nerf_helper.py. Normally, I think it would produce a .so (dynamic lib) file. Here are my terminal outputs:

1. Trying to find nerf_helper related files (no nerf_helper.py):
   
2. python3 setup.py install --user output (last few lines):
   
3. after compilation: cuda/ folder structure:
   

If you need further help, plz provide more implementation details like: how you use nerf_helper, how you load this module, how you compile CUDA extension and the environment of your device.

[ZZy129326999]

(9 line) Thank you very much!!!!

[ZZy129326999]

I directly run setup.py with your method, I understand that cuda is not used in the project, but I am just curious why the compiled nerf_helper cannot retrieve the encoding. Thank you very much!

[Enigmatisms]

Well, it works for me even after checking out this older version and recompile:

Did you try importing this in a python shell? What does it output?

[ZZy129326999]

Wow! Does this mean it has!？

[Enigmatisms]

I think it does, but your IDE can never know its definition since it is loaded from .so. So if this does not occur during run-time and there is no further question, I will close this issue.

[ZZy129326999]

Yes! Your code will not generate any errors during runtime! They are as good as you are! ! Thank you very much! You are a real export!!!!!

[Enigmatisms]

Thank you. That's a little overstating. Hope u can learn the things u want.