Basic-setting

Basic settings about conda, jupyter, tensorflow, or other settings.

Anaconda

Following scripts are applicable on Anaconda Distribution Version 5.3 | Release Date: September 28, 2018. conda 4.5.11 comes with the Anaconda distribution.

Conda

Conda Manual

See environments list
- conda info --envs
Create new environments
- Python 3.6, conda create --name py36 python=3.6 [other packages]
- Python 2.7, conda create --name py27 python=2.7 [other packages]
Create new environments with ipykernel (Recommended)
- conda create -n py35 python=3.5 ipykernel
- Consider following packages when above command is used:
  - (Basic)
    - conda install numpy
    - conda install pandas
    - conda install scikit-learn (scipy will be installed)
    - conda install -c conda-forge matplotlib
  - (Deep learning)
    - tensorflow See install guide
    - pytorch See install guide
  - (Network) networkx

[Note] Since jupyter lab is launched, nb_conda_kernel should be installed to load the kernels in other environments. Do the following command in the base environment.

[Deprecated]conda install -c conda-forge nb_conda_kernels
conda install nb_conda_kernels

If ipykernel is installed in other environments, we can load the ipykernel in main jupyter-notebook. It is unnecessary to install jupyter-notebook or other packages for jupyter-notebook.

[Note] If a following error appears,

undefined symbol: zmq_curve_public

Reinstall pyzmq and zmq.

$ pip install pyzmq
$ pip install pyzmp --upgrade

Then, it should be able to load other kernels. See issue.

Switch environments
- source activate py36
- source deactivate
Path of kernels
- ~/anaconda2/envs/[env_name]/share/jupyter/kernels/[kernel-name]

[Note] If a package is NOT installed in an active environment but the conda-root, the package is not imported. For example, numpy is only installed in the conda-root. (not installed in py35)

source activate py36
python
import numpy
ImportError: No module named 'numpy'

SSH login with (authentication) key

Objectives:

Set up password-less SSH login

Benefits:

Easier to login

Let's begin.

On local machine, create an authentication key-pair if you don't have one. Don't enter the passphrase if you don't want to use it
```
ssh-keygen -t rsa -b 2048 -f id_rsa
```
Add your key to local authentication agent
```
ssh-add ~/.ssh/id_rsa
```
Copy your key from local machine to the server
```
scp ~/.ssh/id_rsa.pub {**SERVER**}:~
```

On the server, Add your key to the list of authorized keys

cd ~/.ssh
cat ~/id_rsa.pub >> authorized_keys2
rm ~/id_rsa.pub

sungyongs/Basic-setting

Basic-setting

Anaconda

Conda

SSH login with (authentication) key