You can quickly set up a Cortex project by running the following command:

cerebrium init-cortex-project <<your_initialisation_directory>>

This will create a Cortex project in the specified directory with the following structure:

your_project/
├── main.py
├── requirements.txt
├── pkglist.txt
├── conda_pkglist.txt
└── config.yaml

​
Configuring your Cortex Project

The initial files are initialised with some default values, and you can add to your requirements.txt, pkglist.txt and conda_pkglist.txt as you would normally.
Alternatively, you can use the --requirements-list, --pkg-list and --conda-pkglist commands to customise the contents of these files.

For example, you can add a few packages to your requirements.txt as follows:

--requirements-list "['transformers', 'datasets', 'sentencepiece==0.1.97']"

The config.yaml file contains the parameters for your deployment. It is initialised with the following default values:

%YAML 1.2
---
hardware: GPU
include: "[./*, main.py, requirements.txt, pkglist.txt, conda_pkglist.txt]"
exclude: "[./.*, ./__*]"
log_level: INFO
disable_animation: false

However, you can change each of these values by using their respective flags as you would with a normal cortex deployment. For example, to change the hardware to CPU, you can run: See the documentation here for more info on config files.

Model StatusPersistent Storage for Caching