installation issue

[ZhuHouYi]

I have some installation problems：

Before installing the nilm-contrib package, my virtual environment is as follows:
The python version：3.8
cuda version：11.6
cudnn version：8.6.0
nilmtk：0.4.0 (However, in the conda environment, it will be displayed as: nilmtk-0.4.0.dev1+git.303d45b)
pandas：0.25.3, numpy：1.13.3, Keras：>=2.4, tensorflow-gpu：2.4.0

There is a package conflict problem when installing nilm-contrib. I have to download the package offline from anaconda.org and install it with the command conda install nilmtk-contrib-0.1.1-py_0.tar.bz2 (for me it is only normal through this method Install, but some compatibility errors will still be reported after the installation is complete)

The official recommended command: conda install -c conda-forge -c nilmtk nilmtk-contrib does not solve the problem, even if I recreated a virtual environment and the environment only contains python3.8 and some basic packages. This process will take a long time and eventually report the following error message, but it does not show which modules are conflicting. I've been stuck with this problem for a week now, hope it gets fixed

[ZhuHouYi]

The environmental problem was finally solved.
I reinstalled a virtual environment using conda, and the python version is 3.7 (the 3.8 I used before). Then install the older version of tensorflow first, then install nilmtk, and finally nilm-contrib. In addition to installing nilm-contrib, the error about Keras is reported during the whole process (but it seems to have no effect). Then check if the GPU is available (False). Then the WindowsGRU library was run again, but the CPU was used, and the code was executed for a long time, but fortunately, the environmental problem was solved. The next step may be the problem of cuda and cudnn, I will update it later

My environment configuration is as follows:：

python：3.7
tensorflow-gpu：1.14.0
Keras：2.3.1
pandas：0.25.3
numpy：1.19.5
nilmtk：0.4.0.dev1+git.303d45b
nilmtk-contrib：0.1.1

[ZhuHouYi]

It took a lot of time to finally install all the environments. The installation process is as follows😀😀:

1. First create the virtual environment of python3.7 in conda
2. Use conda install tensorflow-gpu=1.14.0 to install the tensorflow package
3. Use conda install -c nilmtk nilmtk to install the nilmtk package (it is recommended to add conda-forge to conda channels)
4. Download the nilm-contrib 0.0.1 version package from the anaconda.org official website, and use conda install --offline nilmtk-contrib-0.1.1-py_0.tar.bz2 to install the package offline after activating the python environment
5. If you do not have the corresponding cuda and cudnn environment, first download and install the corresponding version of the installation package (python3.7 and tensorflow-gpu1.14.0 require corresponding versions of cuda10.0 and cudnn7.4.2)
6. Run the deep learning algorithm corresponding to nilm-contrib. If you find that the saving.py file in the Keras package will report an error, open the file and delete all the codes with the suffix .decode() (this problem may be due to python2.x Caused by incompatibility with python3.x)

[ccruzalegui]

Hi, I am trying to work with this libraries in Google Colab. Did you by any chance manage to get them to work there? Can you advice on how to do that?

I managed to install and use the NILMTK and NILM_Contrib following the standard installation procedures in their repos (with independent environments) in my computer, but my experiments take too long and want to try the GPUs in cloud.

I would appreciate any support.

[Raphael164]

Hi, I am trying to work with this libraries in Google Colab. Did you by any chance manage to get them to work there? Can you advice on how to do that?

I managed to install and use the NILMTK and NILM_Contrib following the standard installation procedures in their repos (with independent environments) in my computer, but my experiments take too long and want to try the GPUs in cloud.

I would appreciate any support.

How did you manage the Installation without Google Collab? I would most certainly appreciate any help

[ccruzalegui]

Hi, I am trying to work with this libraries in Google Colab. Did you by any chance manage to get them to work there? Can you advice on how to do that?
I managed to install and use the NILMTK and NILM_Contrib following the standard installation procedures in their repos (with independent environments) in my computer, but my experiments take too long and want to try the GPUs in cloud.
I would appreciate any support.

How did you manage the Installation without Google Collab? I would most certainly appreciate any help

I did not do anything sophisticated. Just followed the instructions in the repos for both environments (nilmtk & nilm-contrib). I am trying to install contrib in another machine to run some experiments, I will let you know if I encounter any issues now.

[Raphael164]

Hi, I am trying to work with this libraries in Google Colab. Did you by any chance manage to get them to work there? Can you advice on how to do that?
I managed to install and use the NILMTK and NILM_Contrib following the standard installation procedures in their repos (with independent environments) in my computer, but my experiments take too long and want to try the GPUs in cloud.
I would appreciate any support.

How did you manage the Installation without Google Collab? I would most certainly appreciate any help

I did not do anything sophisticated. Just followed the instructions in the repos for both environments (nilmtk & nilm-contrib). I am trying to install contrib in another machine to run some experiments, I will let you know if I encounter any issues now.

Would you mind telling me exactly what you did for installing the two repos? Did you install them via conda-forge or pip install git+url?

Thanks in advance

[ccruzalegui]

Hi, I am trying to work with this libraries in Google Colab. Did you by any chance manage to get them to work there? Can you advice on how to do that?
I managed to install and use the NILMTK and NILM_Contrib following the standard installation procedures in their repos (with independent environments) in my computer, but my experiments take too long and want to try the GPUs in cloud.
I would appreciate any support.

How did you manage the Installation without Google Collab? I would most certainly appreciate any help

I did not do anything sophisticated. Just followed the instructions in the repos for both environments (nilmtk & nilm-contrib). I am trying to install contrib in another machine to run some experiments, I will let you know if I encounter any issues now.

Would you mind telling me exactly what you did for installing the two repos? Did you install them via conda-forge or pip install git+url?

Thanks in advance

I can try to summarize it like this:
For NILMTK:

• Install a fresh new anaconda
• Open anaconda prompt
• Create a new environment: conda create --name nilmtk-env
• Install conda: conda config --add channels conda-forge
• Update conda version if needed (this was necessary in my case)
• Activate the environment you just created: conda activate nilmtk-env
• Install NILMTK: conda install -c nilmtk nilmtk (I used this line because the line provided in the repo did not worked, you can try different commands from here if needed: link. However, that line worked smoothly for me on 2 different machines already
• The installation ran for about 6-8 hours in my case.
• Run your IDE and test it. I am using Jupyter notebooks and it worked.

For the NILM-Contrib I did the same, just changed the line to create the a dedicated environment using the anaconda prompt as adviced in the repo. However, I am trying to install it in a different computer and is not working so far.

Hope I am able to help you. Good luck!