# Jupyter with Environment File on Cluster

A summary on how to launch a Jupyter notebook with specific Conda environment file. Before you start, make sure to download and install [Anaconda](https://www.anaconda.com/). Also, you have to download our project from the GitHub page. The project contains the environment files we use.

Two environments (TensorFlow 1/ TensorFlow 2) we are using for this class:
- TensorFlow 2 ([Mac](https://github.com/AOS551/AOS551_PINN_tf2/blob/main/environment-osx.yml), [Win](https://github.com/AOS551/AOS551_PINN_tf2/blob/main/environment-win.yml))
- TensorFlow 1 ([Mac](https://github.com/AOS551/AOS551_PINN_tf1.14/blob/main/environment-osx.yml), [Win](https://github.com/AOS551/AOS551_PINN_tf1.14/blob/main/environment-win.yml))

## Using Jupyter Locally
- Launch Terminal
- For Windows, you must use [**Anaconda Prompt**](https://hackmd.io/_uploads/Skt-qJLmt.png)
- For Mac, use the regular terminal
- Create the environment by executing `conda env create -f <the .yml file>`
- It will take few minutes to download and install the environment.
- Activate environment by executing `conda activate <My Environment>`
- `<My Environment>` should be `tf24`, `tf25`, or `tf114` created in the previous step.
- Launch Jupyter by executing `jupyter notebook`, and it should open Jupyter using a Web browser.

## Using Jupyter On Cluster (Della/Adroit)
1. Download the code from [our class Gituhb](https://github.com/AOS551/AOS551_Curve-fitting) to your local machine. Remember the directory of the code. For example `Downloads/AOS551/AOS551_Curve-fitting-main/`

2. Open terminal on your local machine. Use the following command to copy the folder of codes into your Adroit account:
`scp -r <directory_of_your_code> <username>@adroit.princeton.edu:/scratch/network/<username>`

3. Make directories writeable
- Log into Adroit with `ssh <username>@adroit.princeton.edu`
- Run this on the command line `chmod -R u+w /scratch/network/<username>/`

4. Create environment: (Skip this step if you had already done it before)
- Log into Adroit with `ssh <username>@adroit.princeton.edu`
- Change directory into your scratch directory with
`cd ../../scratch/network/<username>/`
- Creating the environment while under the /scratch directory will keep more free space on your home directory (which only has 10GB on Adroit)
- Load Anaconda with `module load anaconda3/2020.7`
- Create the environment by executing `conda env create -f <the .yml file>`
- We use [**Mac environment file**](https://github.com/AOS551/AOS551_Curve-fitting/blob/main/environment-osx.yml) on the cluster
For example, because your .yml file is preiviously saved along with the code under the home directory using scp, you will use `conda env create -f AOS551_Curve-fitting-main/environment-osx.yml`
- It will take few minutes to download and install the environment.

5. Launch the Web interface of the cluster ([Della](https://mydella.princeton.edu/), [Adroit]( https://myadroit.princeton.edu))
- Select `Interactive Apps` -> `Jupyter for Classes`
- Input parameters for the Jupyter
- For class materials, `1 hour`, `1 core`, `8 (GB) Memory`, and `AOS551` would work.
- Click `Launch` (it takes time) -> `Connect to Jupyter`
- Click the target Jupyter notebook file
- Click `File` -> `Trust Notebook`
- **Switch the kernel to the target environment** by click `Kernel` -> `Change kernel`
6. Remember to delete the session after use