RL Enzyme Engineering

Run on fly.io

fly.io offers simple access to GPU nodes.

To launch an instance of this algorithm on fly.io follow those steps:

1. Create an account on fly.io.
   Beware: to get access to GPU nodes you may need to write them a short mail to ask for access incl. your acounts id, as they are still rolling out their offer.

2. Install the fly.io CLI flyctl

3. Clone the repostiory to your local system:

   git clone [email protected]:li-il-li/rl-enzyme-engineering.git

4. Create a new 'app' on fly.io:

   cd flyio
   fly launch
   # Would you like to allocate dedicated ipv4 and ipv6 addresses now? y -> copy IP

5. Add you SSH public key:

   fly secrets set "AUTHORIZED_KEYS=$(cat ~/.ssh/id_rsa.pub)"

   If you want to work on the machine I recommend also setting:

   fly secrets set "GIT_NAME=<name>"
   fly secrets set "GIT_EMAIL=<email>"
   fly secrets set "GIT_PASSWORD=<password>"

6. Deploy the app:

   fly deploy

7. Recommendation: Use VSCode to connect to the instance via SSH or just:

   ssh root@<copied_IP>

8. Configure your wildtype_AA_seq and ligand_smile in conf.yaml

9. Run algorithm via:

   uv run src/main.py

10. Important: When you are done, shutdown the instance:

    fly scale count 0

    This will scale down the instance but keep the volume (costs!).
    To delete the whole app:

    fly apps destroy rl-enzyme-engineering