Use of unsupported NumPy function 'numpy.einsum'

[AinaraC]

Hi!. I'm trying to run Frustum-Pointpillars in a conda environment in my Jetson Orin NX with the following main packages:

Jetpack = 6.1
CUDA = 12.6
Torch = 2.5 (following instructions for jetson platform)
Torchvision = 0.20.0 (following instructions for jetson platform)
Python = 3.10.15
Fire = 0.7.0
Numba = 0.60.0
Numpy = 1.26.0
Pillow = 11.0.0
Protobuf = 3.20
Pybind11 = 2.13.6
TensorboardX = 2.6.2.2
shapely = 2.0.6
scikit-image = 0.24.0
Sparsehash = 2.0.4 installed with conda (channel confa-forge)
libboost-all-dev = 1.74.0.3

When a I execute the training command: python ./pytorch/train.py train --config_path=./configs/pointpillars/car/xyres_16.proto --model_dir=/path/to/model_dir

I get this error on the terminal:

numba.core.errors.TypingError: Failed in nopython mode pipeline (step: nopython frontend)
Failed in nopython mode pipeline (step: nopython frontend)
Use of unsupported NumPy function 'numpy.einsum' or unsupported use of the function

The whole output from the terminal is this:

The issue seems to come from the versions of numpy and numba, as numba doesn't support numpy function einsum(). I've tried to change both versions, but I can't find a version that suits my environment. Is there any other similar function I can use to solve this problem? Any other solution?

Any idea is welcome, thank you