MNIST classification with MLPs.
- 01-pytorch-mnist-mlp.ipynb
Image classification with CNNs.
- 02-pytorch-mnist-cnn.ipynb
Land segmentation with UNET.
- 03-train_model.py
- 03-train_model.sh
- 03-inference_and_evaluation.ipynb
Point cloud/object classification with GNNs.
- run_mahti.sh
- train_shape_geom.py
Fashion MNIST with BNNs.
- run_mahti.sh
- train_fashion_bayesian.py
During the course exercises are run on Mahti, which is a Finnish national supercomputer. Accessing Mahti requires a project with a budget. Finnish users get access to Mahti via CSC. For this course the course participants are added to the course project.
- Open https://www.mahti.csc.fi
- Log in with:
- HAKA, if you have (Finnish universities and some research institutes, e.g. FMI)
- CSC account, you need your CSC username and password
Open Login node shell
cd /scratch/project_2017263
mkdir $USER
cd $USER
git clone https://github.com/csc-training/lumi-aif-fmi.git
- Click "Jupyter" on dashboard
- Select following settings:
- Reservation: fmi-day1 or fmi-day2
- Project: project_2017263 during the course, own project later
- Partition: interactive
- CPU cores: 4
- Local disk: 0
- Time: 4:00:00 (or adjust to reasonable)
- Working directory: /scratch/project_2017263 during the course, own project's scratch later
- Python: pytorch
- Click launch and wait until granted resources
- Click "Connect to Jupyter"
- Open the cloned exercise folder under your
<your_username>in JupyterLab
Tip
If you see parts of the notebook disappearing when you scroll, this is unfortunately a known issue with newer versions of JupyterLab. A workaround is to set the Windowing mode to "defer" as follows:
- Open "Settings" menu (top bar)
- Open "Settings Editor"
- Search for "windowing mode"
- Set it to "defer", rather than the default "full"
Please acknowledge CSC in your publications, it is important for project continuation and funding reports. As an example, you can write "The authors wish to thank CSC - IT Center for Science, Finland (urn:nbn:fi:research-infras-2016072531) for computational resources and support".