Skip to content

RSE-Cambridge/RSECon25-Dawn-Workshop

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

29 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

RSECon25-Dawn-Workshop

1. Dawn

Dawn is a supercomputer hosted at the University of Cambridge, and is part of the AI Resource Research (AIRR). It makes available Intel Data Centre GPU Max 1550 GPUs. To install and run the workshop software on Dawn, you will need to have been granted access. For further information, see: Access to Dawn.

2. Access to Dawn via the login-web interface

If you haven't done this already, and have an account, enable access to Dawn via the login-web interface.

3. Connect to Jupyter server

Login at: https://login-web.hpc.cam.ac.uk/

At the top of the dashboard page presented after login, select:

Interactive Apps -> Jupyter Notebook

On the Jupyter Notebook form enter:

- Project account: training-dawn-gpu
- Partition: pvc9
- Reservation [leave blank]:
- Number of hours: 2
- Number of cores: 1
- Number of GPUs: 2
- Modules: intel-oneapi-mkl intel-oneapi-compilers jupyterlab
- Number of nodes: 1

The above request the resources needed for this workshop, for a period of 2 hours. The number of hours may be decreased or increased as needed.

Click the Launch button.

Your request for a Jupyter Notebook will progress through the states: Queued, Starting, Running. Once the Running state is reached, click the Connect to Jupyter button.

Once connected to the Jupyter server, you will initially be shown the File Browser, in the Jupyter Home tab.

4. Download examples and set up environment

On the Jupyter page, select:

File -> New -> Terminal

The Terminal window that opens will be in the home directory of your account on Dawn.

Clone this repository to your home directory:

git clone https://github.com/RSE-Cambridge/RSECon25-Dawn-Workshop

Set up the environement:

source ~/RSECon25-Dawn-Workshop/scripts/workshop_setup.sh

This will clone another repository with examples (practical-ml-with-pytorch) to your home directory, will create account-specific setup files in ~/RSECon25-Dawn-Workshop/install and will create kernel-definition files in ~/.local/shared/jupyter/kernels.

5. Submit batch jobs

Move to the workshop examples directory, and submit the two jobs that perform multi-node processing:

cd ~/RSECon25-Dawn-Workshop/examples
sbatch --account=training-dawn-gpu run_mnist_classify_ddp.sh
sbatch --account=training-dawn-gpu run_lightning_toy_example.sh

6. Run Jupyter notebooks

On the Jupyter page, select:

File -> View -> Open JupyterLab

6.1 Check devices

In the left panel, navigate to RSECon25-Dawn-Workshop/examples. Open, and experiment with: check_devices.ipynb. Before running the notebook, set the kernel to ai:

Kernel -> Change Kernel... -> ai

6.2 Practical ml with PyTorch

In the left panel, navigate to practical-ml-with-pytorch/worked-solutions. Open and experiment with the four notebooks. Before running each notebook, set the kernel to practical-ml-with-pytorch:

Kernel -> Change Kernel... -> practical-ml-with-pytorch

About

Resources for Dawn workshop at RSECon25

Resources

License

Stars

2 stars

Watchers

1 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages