Notebook fails on colab

[alasdairmotiv]

Thanks for sharing your work.
I tried running on colab with the T4 Gpu and it failed with the following error message while running "from pipeline import run_treelearn_pipeline":

In file included from /usr/local/cuda/include/cuda_bf16.h:4082,
from /usr/local/lib/python3.10/dist-packages/cumm/include/tensorview/dtypes.h:21,
from /usr/local/lib/python3.10/dist-packages/cumm/include/tensorview/tensorview.h:28,
from /usr/local/lib/python3.10/dist-packages/cumm/include/tensorview/cuda/driverops.h:16,
from /usr/local/lib/python3.10/dist-packages/cumm/include/tensorview/contexts/core.h:20,
from /usr/local/lib/python3.10/dist-packages/cumm/include/tensorview/context.h:2,
from /usr/local/lib/python3.10/dist-packages/cumm/include/tensorview/tensor.h:94,
from /usr/local/lib/python3.10/dist-packages/cumm/include/tensorview/pybind.h:16,
from /usr/local/lib/python3.10/dist-packages/cumm/include/tensorview/pybind_utils.h:2,
from /usr/local/lib/python3.10/dist-packages/cumm/build/core_cc/include/tensorview_bind/TensorViewBind.h:2,
from /usr/local/lib/python3.10/dist-packages/cumm/build/core_cc/src/tensorview_bind/PyBindTensorViewBind/PyBindTensorViewBind_bind_TensorViewBind.cc:2:
/usr/local/cuda/include/cuda_bf16.hpp:238:26: note: candidate: ‘constexpr __nv_bfloat16& __nv_bfloat16::operator=(const __nv_bfloat16&)’
238 | struct CUDA_ALIGN(2) __nv_bfloat16 {
| ^~~~~~~~~~~~~
/usr/local/cuda/include/cuda_bf16.hpp:238:26: note: candidate: ‘constexpr __nv_bfloat16& __nv_bfloat16::operator=(__nv_bfloat16&&)’
[2/39] [GCC][c++/pch]/usr/local/lib/python3.10/dist-packages/cumm/build/core_cc/include/csrc/arrayref/ArrayPtr.h.gch
/usr/local/lib/python3.10/dist-packages/cumm/build/core_cc/include/csrc/arrayref/ArrayPtr.h:1:9: warning: #pragma once in main file
1 | #pragma once
| ^~~~
[3/39] [GCC][c++/pch]/usr/local/lib/python3.10/dist-packages/cumm/build/core_cc/include/tensorview_bind/TensorViewBind.h.gch
/usr/local/lib/python3.10/dist-packages/cumm/build/core_cc/include/tensorview_bind/TensorViewBind.h:1:9: warning: #pragma once in main file
1 | #pragma once
| ^~~~
ninja: build stopped: subcommand failed.

CalledProcessError Traceback (most recent call last)
in <cell line: 3>()
1 import sys
2 sys.path.append("/content/TreeLearn/tools/pipeline")
----> 3 from pipeline import run_treelearn_pipeline
4 import argparse, pprint
5 from tree_learn.util import get_config

9 frames
/usr/local/lib/python3.10/dist-packages/ccimport/buildtools/writer.py in build_simple_ninja(target, build_dir, sources, build_meta, target_filename, suffix_to_compiler, out_root, verbose, shared, pch_to_sources, pch_to_include, objects_folder, compiler_to_path, linker_to_path, source_meta, linker)
1106 proc.wait()
1107 if proc.returncode:
-> 1108 raise subprocess.CalledProcessError(proc.returncode, cmds)
1109
1110 no_work_to_do = False

CalledProcessError: Command '['ninja']' returned non-zero exit status 1.

[JanvDelden]

Hello alasdairmotiv,

thanks for pointing this out! It seems that google colab changed their default environment, which means that we also have to change the setup part of the notebook. I was not immediately able to find a working solution, but will give it another try soon.

[rupert-br]

Hi @JanvDelden, any idea how the fix would look like for running the jupyter notebook on google colab? Great work on the research btw.

[JanvDelden]

Hi rupert,
thanks for asking! The solution would be to find a combination of packages, including spconv, that work directly with the google colab environment. Unfortunately, I did not find a configuration that works due to the cuda version on google colab. I think it would be best to use an alternative service, where you have more control over the environment than google colab. A candidate is Hugging Face spaces. However, we will probably not implement this ourselves due to time constraints.

[rupert-br]

Thank you for the tip with the hugging face space. I'll try this for now.