Gensyn Nodes Guide by EZ Labs with Autoscript

Requirement
16GB RAM
CUDA Devices RTX 3090, RTX 4090, A100, H100
Python Version >= 3.10

Note: You can run the node without a GPU using CPU-only mode.

⚙️Installation of the Node:

Step 1: Install Dependencies:

wget -qO gensyndepedencies.sh https://raw.githubusercontent.com/ezlabs-nodes/gensyn/refs/heads/main/gensyndepedencies.sh && chmod +x gensyndepedencies.sh && ./gensyndepedencies.sh

Execute the below command:

sudo reboot

Step 2: Create a screen:

Open a screen to run it in the background

screen -S gensyn

Step 3: Clone the Repo & execute the rlswarm.sh:

cd $HOME && rm -rf rlswarm.sh && wget https://raw.githubusercontent.com/ezlabs-nodes/gensyn/refs/heads/main/rlswarm.sh && chmod +x rlswarm.sh && ./rlswarm.sh

• Official doc guide to select the parameters: https://github.com/gensyn-ai/rl-swarm

• For GPU → Big model (7B, 32B or 72B) + Math Hard (DAPO-Math 17K dataset)

Required: A100 (80GB) or H100 (80GB) Optional GPU (for small models) → Qwen 0.5B or 1.5B + GSM8K

• Suggested: Any GPU with >8GB vRAM (e.g., RTX 3090, 4090, A100, H100)

The script will install some Python packages. If prompted for Hugging Face credentials, type ’N’ to skip.

Step 4: Login:

Now, it will print the website URL as shown in the screenshot. Just open it, use the password displayed in the terminal, enter the password, and submit.

• Then, log in using your Google credentials. Once logged in.

  

• After that, the script will prompt you to enter your Hugging Face credentials. Type’N’ as shown in the screenshot below.

• If you see this , Done for run and go to home screen your VPS CTRL+A+D

Check your Log:

If you want to view your log, just reattach to the screen using the command below.

screen -r gensyn

• Check Wins: @gensyntrackbot on telegram

Enter your Peer-ID, you will find that in your log

• Official dashboard: https://dashboard.gensyn.ai/

You can search your Node name in the dashboard after a while when you have done your first training completed

• Backup your Userdata & API Key:

  cat ~/rl-swarm/modal-login/temp-data/userData.json && cat ~/rl-swarm/modal-login/temp-data/userApiKey.json