This folder contains the student template of project 2
We have prepared an environment on the GPU cluster ran by the IT support group of the computer science department. See the project information sheet (and the links therein) for more information.
For those who are running the project with Google Colab, we prepared the interface to run the code in the Instructions_GoogleColab.ipynb.
Please check out the installation guide on Moodle for this.
Make sure you upload the whole project2 folder (including the uncompressed datasets) to your Google Drive and follow the instructions in the Instructions_GoogleColab.ipynb to run the code.
If you have a computer with a GPU, you might want to run the project locally. In this case, please set up an Anaconda environment running python3.10. Please check out the installation guide, e.g. from Informatik II for this.
If you are using Windows, we recommend to use either the VS code terminal or the Anaconda terminal, which is installed with Anaconda.
Please activate your project 2 environment by using:
conda activate <environment_name>
Then navigate to the folder containing the project files and run:
pip install --upgrade pip
pip install -r requirements.txt
If you require any additional packages, run:
pip install <package_name>
Make sure to extract all the data in the ./datasets folder.
Please note that the script train.py takes arguments when you run them. These arguments are used when the script is carried out. The arguments to a python script can be specified in the following manner:
python train.py --<argument_1_name> <argument_1_value> --<argument_2_name> <argument_2_value>train.py takes two arguments, namely the path to the datasets folder and where the training loop should save your model checkpoints.
For more information on the available arguments to these scripts, please run the following command:
python train.py -hSee the instructions on the project document and the example in the file my_job.sh
Check out the Instructions_GoogleColab.ipynb for instructions on how to run the code in Google Colab.
To run your solution locally, first make sure you have activated your conda environment. Then open a terminal and run the following command with your arguments to train the model:
python train.py <your_arguments_here>