Instructions I found helpful to solve problems or install various packages and softwares on my Manjaro linux.

Sights of Experience.md

Sights of Experience!

Instructions I found helpful to solve problems or install various packages and softwares on my Manjaro linux.

Sights of Experience2.md

Linux Issues

tty not found!

checkout: Boot livedisk + chroot + update + mkinitcpio

I take it that this means you cannot boot into prompt mode at all. Which means you have to chroot into the installed partition and make repairs to it. But the situation does not look promising.

[1] - I rarely use chroot and these are the steps I used. There are several (okay,2) simpler ways to chroot but I have never used them. Perhaps if others step in on chroot, use their procedure. It’s simpler. Here are my steps:

mount /dev/sda3 /mnt
cd /mnt
mount -t proc proc /mnt/proc
mount -t sysfs sys /mnt/sys
mount -o bind /dev /mnt/dev
mount -t devpts pts /mnt/dev/pts/
chroot /mnt

I used /dev/sda3 as an example, Use the correct Manjaro partition in your case.

Chroot [1] from a live cd and perform these operations at chroot

sudo pacman-mirrors -g
sudo pacman -Syyu
sudo mkinitcpio -P

When finished with chroot, type

exit

And then don’t do anything else, just in case chroot is still in operations. Reboot. Good luck.

Sights of Experience3.md

Machine Learning Installations

Opencv

summary : Install opencv-python from pip

after installing conda, I tried to install opencv using command:

  conda install -c conda-forge opencv

the problem was it didn't have the GUI libs necessary, so it would error something like:

The function is not implemented. Rebuild the library with Windows, GTK+ 2.x ... install libgtk2.0-dev and pkg-config, then re-run cmake ...

so I searched the web, and tried mapno channel's opencv from conda, and also installing gtk2 and pkg-config packages didn't help. the solution for me was to install opencv-python package from pip:

sudo pip install opencv-python

Tensorflow GPU

summary : Install tensorflow-gpu from conda + read here and here

Manjaro linux repositories have Tensorflow-cuda and python-tensorflow-cuda packages, unfortunately none of them worked for me cause my manjaro had python 3.7 and also installed Anaconda package had the same version and as of now Tensorflow is not supporting the 3.7 version. Installing tensorflow gpu from pip would prompt an error like so :

Could not find a version that satisfies the requirement tensorflow-gpu (from versions: )
No matching distribution found for tensorflow-gpu

This is how I finnaly installed it:

First I created a new environment unsing conda with python 3.6:

conda create -n myenv python=3.6

Install GPU drivers:

sudo mhwd -a pci nonfree 0300

I installed Cuda and Cudnn using:

sudo pacman -S cuda cudnn

then I installed tensorflow-gpu package from the manjaro repos (probebly won't help anything but thats what I did).

finnaly I tried to install tensorflow-gpu from conda using:

conda install tensorflow-gpu

and It solved it!

for Adding the virtual environment to supported kernels of jupyter notebook use:

python -m ipykernel install --user --name [deep-learning] --display-name "Python (deep-learning)"

+ more info

after running a python notebook and testing, tensorflow would import and work fine, but it wasn't using the GPU. in order to solve that I followed this Instruction. luckily it should start your GPU:

sudo optirun -b none nvidia-settings -c :8

Sonddevice

summary : Install portaudio python-pyaudio using your package manager

Installing sounddevice from pip didn't work and trough this error:

 OSError: PortAudio library not found

Solution (oddly I didn't see it elsewhere) was to install PortAudio library and its python wrapper:

pacman -Syu portaudio python-pyaudio 

Converting ipython notebooks to scripts

summary : use jupyter nbconvert

jupyter nbconvert --to python *.ipynb --TemplateExporter.exclude_markdown=True --TemplateExporter.exclude_input_prompt=True