We will use Pipfile as a symlink from the Pipfile.cpu or Pipfile.gpu depending on the training

• Install Pipenv
• Navigate into the main directory
• Run pipenv install to install all the required packages needed for the experimentation
• In order to load the local env just run pipenv shell

One can use the CLI to preprocess the original dataset or train a model based on some settings.

In order to check the CLI command run the following:

python -m similarity_learning.cli --help

To preprocess the original dataset, one may run the following:

• Check the parameters python -m similarity_learning.cli dataset --help

• Run the default parameters python -m similarity_learning.cli dataset

• Alternatively, in order to create all datasets for all the variations you may run the following: python -m similarity_learning.scripts.handle_raw_dataset

Note!!! that the aforementioned execution will take some time.

To run an experiment you can use the following:

• Check the config files in order to use one of the .yml files or create one of your own.
• Run using the following: python -m similarity_learning.cli train --exp_name <some-exp-name> --settings <config-file>.yml

To evaluate an experiment you can use the following:

• Check the config files (in the test section) in order to use one of the .yml files or create one of your own.
• Run using the following: python -m similarity_learning.cli evaluate --settings <config-file>.yml

For example: python -m similarity_learning.cli evaluate --settings test_similarity.yml