Skip to content

Latest commit

 

History

History
62 lines (45 loc) · 2.85 KB

README.md

File metadata and controls

62 lines (45 loc) · 2.85 KB

EMBL Advanced Deep Learning for Image Analysis 2025

Schedule:

  • Webinar 1: Monday 13th January 14:00 - 15:00 CET
  • Webinar 2: Monday 20th January 14:00 - 15:00 CET
  • Webinar 3: Monday 3rd February 14:00 - 15:00 CET

Set-up:

We will use on-premise EMBL resources (BARD) for the course. You can also use google colab to run the U-Net notebooks if you prefer. However, it might make sense to familiarise yourself with BARD as this is what you will be using during the course.

BARD

You will all be given a login to a BARD account, working within this VM you can clone the git repo to your home folder and complete the exercises. More details will be given in precourse webinar 1 and can be found on the ecampus page or youcan always ask if you have other questions.

Running the exercises with BARD:

  1. Login into your BARD account (The first login on a node may be slow)
  2. Start VSCode by clicking its icon in the dock
  3. Make sure vscode is in "Trust" mode, not "Restricted"
  4. Install Python and Jupyter extensions
  5. Go to Terminal-> new Terminal to start a terminal in vscode IMPORTANT: It is best to only use the vscode terminal as it is inside a docker container that has access to git, conda and everything you need. Other terminals on BARD are in separate docker containers that don’t have access to everything. TLDR: just use the vscode terminal 🙂
  6. In the terminal, git clone https://github.com/kreshuklab/ADL4IA-2025
  7. cd into U-Net_exercise/pytorch
  8. Open a .ipynb file in vscode
  9. From the top right of vscode, ‘select kernel’ /opt/conda/envs/pytorch/bin/python

Note on wandb logging:

We are using wandb logging inside the notebooks as it is, in our opinion, nicer than tensorboard. Follow the instructions given by wandb when you run it to view your logs in a web browser.

You will need to login in wandb in order to start logging and will need to register if you don’t already have an account. The login is prompted in a cell at the beginning of every notebook although it will remember you after the first time. You can also login via the vscode terminal

  1. register with wandb and get a api key
  2. in the terminal within vscode source /opt/conda/etc/profile.d/conda.sh conda activate pytorch wandb login paste wandb api key and enter Run the notebook

Conda environments:

A few premade conda environments already exist within the docker container. These are read only. The pytorch environment will let you run all the pre-course exercises The kernels/conda envs available are

/opt/conda/envs/pytorch/bin/python

/opt/conda/envs/keras/bin/python

/opt/conda/envs/sam/bin/python

These kernels/envs are read only.

======== To create your own conda env

  1. start a terminal in vscode cd to your home directory
  2. source /opt/conda/etc/profile.d/conda.sh
  3. use conda as usual to create new environments