Start Validator Node

Before starting your validator node, ensure the port chosen for the node is open for communicating with peers. The Decentralized LLM Standard lets other peers know your IP address. If this is a concern, we recommend using a VPN or other means to hide it. Like a blockchain, other peers cannot make you run custom code, only what is in the registered RPC logic for model training and inference.

LibP2P has a detailed security analysis.

  • Model name: Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2

    • The path to download the model. The subnet automatically assumes it's a Hugging Face repository unless specified otherwise.

  • Public IP: The public IP of the server.

  • Port: The port opened for communications amongst peers in the subnet.

  • Identity Path: The path to the ed25519 private key file previously generated in the first step.

    • This defaults to private_key.key.

python -m subnet.cli.subnet.run_server_validator Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2 --public_ip {IP} --port {PORT} --identity_path private_key.key

Parameters:

Model Name: The path to download the model. The subnet automatically assumes it's a Hugging Face repository unless specified otherwise.

--public_ip: The public IP of the server.

--port: The port opened for communications amongst peers in the subnet

--identity_path: The path to the ed25519 private key file previously generated in the generate keypair step.

  • This defaults to private_key.key

--num_blocks: (Optional) Specify the number of served transformer blocks.

--local: (Optional) Argument for local testing to use the LOCAL_RPC environment variables.

See the list of arguments in the clidirectory.

This will:

  • Load the model transformer blocks

  • Start consensus integration automatically once the node is activated.

PreviousAddNextActivate

Last updated