AttributeError: module 'mxnet' has no attribute 'random' about datawig

[tdhd]

Which version of mxnet do you have installed?

[AeroClassics]

Hi Phillip, 1.4 is what I downloaded and built. Best Regards / Mit freundlichen Grüßen, Doug

[tdhd]

I guess this has to do with how you have installed mxnet. I just did a clean install of mxnet 1.4, via pip, into a new conda environment and the module was found. How have you installed mxnet?

[AeroClassics]

I suppose I can believe that. I was following this tutorial. The main difference from the tutorial are:
OpenCV 3.4.5 (local build, rather than global)
Cuda 10
cuDNN for Cuda 10
TensorRT
Python 3.6
and I used one additional compiler flag USE_NCCL=1
gcc-6 & g++-6 (Does it matter whether it is gcc-6 or gcc-7?)

I can try again and see what happens.

[AeroClassics]

Same error
Traceback (most recent call last):
File "test.py", line 8, in
mx.random.seed(1234)
AttributeError: module 'mxnet' has no attribute 'random'

I pulled a fresh copy using:
git clone --recursive https://github.com/apache/incubator-mxnet mxnet
cd mxnet
The I built it using:
make -j12 USE_OPENCV=1 USE_BLAS=openblas USE_CUDA=1 USE_CUDA_PATH=/usr/local/cuda USE_CUDNN=1 USE_NCCL=1 USE_NCCL_PATH=/usr/lib/x86_64-linux-gnu
cd ~/.virtualenvs/mxnet/lib/python3.6/site-packages/
ln -s ~/devel/software/mxnet mxnet
Did a workon mxnet
and python3 test.py

With that failing I tried to do the Python install as outlined here. Su:
cd python
pip install -e .
And added my OpenCV lib folder to the LD_LIBRARY_PATH

do python3 test.py
Get spit out:
ModuleNotFoundError: No module named 'numpy.core._multiarray_umath'
ModuleNotFoundError: No module named 'numpy.core._multiarray_umath'
And it appeared to hang. Next time I ran the program I got:
ModuleNotFoundError: No module named 'numpy.core._multiarray_umath'
ModuleNotFoundError: No module named 'numpy.core._multiarray_umath'
INFO:root:downloaded https://s3.us-east-2.amazonaws.com/mxnet-public/letter_recognition/letter-recognition.data into letter-recognition.data successfully
Epoch 0, Training ('accuracy', 0.434625)
Epoch 1, Training ('accuracy', 0.6516875)
Epoch 2, Training ('accuracy', 0.6968125)
Epoch 3, Training ('accuracy', 0.72875)
Epoch 4, Training ('accuracy', 0.7565)

Is this what I should be getting?

Thanks!

[felixbiessmann]

We fixed some issues with dependencies and tested the installation again. Following the instructions in the README should work, hence closing this issue, feel free to re-open if the problem persists.