Update README.md

README.md

@@ -31,35 +31,64 @@ ssh $USER@igrida-oar-frontend
 In order to launch jobs on IGRIDA, you need to submit jobs using the `oarsub` command:
 
 ```bash
-# create the job
-oarsub -S -t timesharing -l /nodes=1/core=4/gpu_device=1,walltime=01:30:0 scripts/run_igrida_staller.sh
-# connect to it using the OAR_JOB_ID, you can list your jobs with oarstat -u
+# create the job (add -t timesharing if you do this tutorial with lots of people in parallel)
+# the option gpu_device=1 ask the grid scheduler for a node with 1 gpu
+# walltime is the maximum duration of the job, here we ask for 40 hours.
+oarsub -I -l gpu_device=1,walltime=40:00:0
+```
+
+This will launch an interactive job, on which you will be connected immediately.
+If you loose your connection or close your terminal, the job will be stopped immediately.
+That is why I prefer to launch a job using a staller script, on which you can connect and disconnect freely.
+
+```bash
+# first clone this project on your /udd/$USER space
+cd $HOME  # or wherever you want to clone this repo
+git clone https://gitlab.inria.fr/kchoi/Deep_Learning_tutorial.git
+cd Deep_Learning_tutorial
+# create the job (add -t timesharing if you do this tutorial with lots of people in parallel)
+oarsub -S -l gpu_device=1,walltime=01:30:0 scripts/run_igrida_staller.sh
+# you can check the status of your jobs with the oarstat command
+oarstat -u
+# connect to it using the Job id
 oarsub -C JOB_ID
-oarsub -I -t besteffort -p "host = 'igrida-abacus.irisa.fr'"  -l /nodes=1/core=4/gpu_device=1,walltime=40:00:0
 ```
 
-You can find the script `scripts/run_igrida_staller.sh` [here](scripts/run_igrida_staller.sh).
+You can find the script  [`scripts/run_igrida_staller.sh`](scripts/run_igrida_staller.sh).
 
 More info at http://igrida.gforge.inria.fr/
 
 ## Clone the project
 
-You can now clone the project on igrida with git:
+If you have not done it already, you can now clone the project on igrida with git:
 
 ```bash
 cd where_ever_you_want
 git clone https://gitlab.inria.fr/kchoi/Deep_Learning_tutorial.git
 ```
 
-## Load python environment and the jupyter notebook
+## Install miniconda
 
-You can source the python environment using the following command:
+[Miniconda](https://docs.conda.io/en/latest/miniconda.html#) is the prefered way of installing a python environment for deep learning.
 
 ```bash
-# replace gpu by cpu if on an cpu only node
-source /temp_dd/igrida-fs1/kchoi/miniconda3/bin/activate gpu
+# get the install script
+wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
+# launch the script
+# by default miniconda will install in your $HOME/miniconda3
+# you can choose another directory, but do not install on /temp_dd/igrida-fs1/$USER !!
+bash Miniconda3-latest-Linux-x86_64.sh
+# if not already done, activate the miniconda environment
+source $HOME/miniconda3/bin/activate
+# make sure that your python is really the miniconda one
+which python
+> /home/$USER/miniconda3/bin/python
+# install keras, jupyter and matplotlib
+conda install keras-gpu jupyter matplotlib  # use keras if no gpu
 ```
 
+## Load the jupyter notebook
+
 Next, run the jupyter notebook from which we will be doing the tutorial:
 
 ```bash
@@ -82,6 +111,6 @@ jupyter notebook --ip=* --no-browser
 Follow the output of the command and connect to the link `http://igrida-abacus:8888/?token=ad046861594b923...`.
 Remember to replace `localhost` to the server you actually are running a job on.
 
-Once you have loaded the jupyter notebook on your browser, you can start the tutorial by the opening the first [notebook]().
+Once you have loaded the jupyter notebook on your browser, you can start the tutorial by the opening the first [notebook](1. Introduction to Deep Learning.ipynb).
 
 That's all for now, happy tutorial!
\ No newline at end of file