Newer
Older
The notebook for the exercise in Chapter 10 is available from here:
[Notebook for mini U-net](https://github.com/vedranaa/teaching-notebooks/blob/main/02506_week10_MiniUnet.ipynb)
Training the network may take long time, especially without GPU. We therefore sketch several ways of running the notebook with GPU support.
## Run the notebook on Google Colab
You can open the notebook in Google Colab from here, and enable GPU :
[](https://colab.research.google.com/github/vedranaa/teaching-notebooks/blob/main/02506_week10_MiniUnet.ipynb)
## Run the notebook on DTU Gbar
You can run the notebook on Gbar on an interactive GPU node. First get to a terminal on the cluster either ssh, ThinLinc client, or browser-based ThinLinc [thinlinc.gbar.dtu.dk](thinlinc.gbar.dtu.dk). Open the terminal (command line).
Log onto an interactive GPU node such as `voltash` by running the following command on the command line:
To set up a Python environment, you can use the script `env02506.sh` which we provided for you. You should place the file `env02506.sh` in a folder on the Gbar and then run the command:
`source env02506.sh`
The first time you run this, the script will create a new python environment called `env02506` and activate it.
Then you should install the packages needed which include:
`pip install torch torchvision`
`pip install Pillow`
`pip install notebook`
Now you are good to go. You can navigate to the folder that you wish to store the code, download the notebook from the link above, and open a jupyter notebook by typing the following in the command line:
`jupyter-notebook`
## Use QIM platform
We are working on establishing an easier way of using Gbar functionality which does the steps described above automatically. This is still experimental.
Go to [https://platform.qim.dk/](https://platform.qim.dk/). Log on. Under Tools start Jupyter launcher. Type in your DTU credentials (it still needs to be done twice). Under basic config QIM environment choose `qim3d` (it has the largest number of packages installed). Under advanced config choose HPC que called `gpuqim`. Tick the checkbox Reset SSH tunnel. Click the button `Start jupyter server`.
The platform shold launch your job and a button with `Open Jupyter` should appear, leading you to a jupyter server running in your home directory on the Gbar.