Lumino Contracts Client

A Python client for interacting with the Lumino protocol, a decentralized computing platform for machine learning jobs.

Overview

The Lumino Contracts Client provides two main executables:

• lumino-node: For node operators to participate in the Lumino network
• lumino-client: For users to submit and monitor jobs on the network

Installation

Requirements

• Python 3.11
• An Ethereum wallet address
• Lumino tokens
• ETH for gas fees
• RPC endpoint for the blockchain

Installing via pip

pip install lumino-contracts-client
export PATH=$HOME/.local/bin:$PATH  # Add Python bin to PATH

Ubuntu-specific setup

If Python 3.11 is not already installed:

sudo apt upgrade -y python3
sudo update-alternatives --config python  # (select version 3.11)
python --version  # (confirm version == 3.11)
python -m pip install lumino-contracts-client && export PATH=/home/$USER/.local/bin:$PATH

Configuration

Both clients create a $HOME/.lumino directory for storing state:

• Node artifacts and configuration
• Client state and job information
• Logs and temporary files

Node Client (lumino-node)

Features

• Interacts with smart contracts for leader election, job assignment, and incentives
• Executes machine learning jobs upon assignment
• Automatically manages node's stake on the node escrow (i.e. top-up)
• Registers node to the Lumino network

Requirements

• A whitelisted Ethereum address
• Lumino tokens transferred to the address
• The lumino artifacts password
• Some ETH for gas costs
• Compatible GPU hardware

Supported GPU Hardware

• 1, 2, 4, or 8x NVIDIA A100 (40GB or 80GB)
• 8x NVIDIA H100 (80GB)

Running the Node

lumino-node

User Client (lumino-client)

Features

• Submits machine learning jobs to the Lumino network
• Monitors job execution and retrieves results
• Manages user's balance on the job escrow contract

Basic Usage

Top up your balance for paying for jobs:

lumino-client topup

Submitting Jobs

Example dummy job:

lumino-client create-job --args '{"shuffle": true, "use_lora": true, "use_qlora": false, "batch_size": 4, "dataset_id": "gs://lum-dev-pipeline-zen-datasets/0ca98b07-9366-4a31-8c83-569961c90294/2024-12-17_21-57-21_text2sql.jsonl", "num_epochs": 1, "job_config_name": "llm_dummy"}' --model llm_dummy --ft_type "LORA" --monitor

Example LLaMA 3.2 1B LoRA job:

lumino-client create-job --args '{"shuffle": true, "use_lora": true, "use_qlora": false, "batch_size": 4, "dataset_id": "gs://lum-dev-pipeline-zen-datasets/0ca98b07-9366-4a31-8c83-569961c90294/2024-12-17_21-57-21_text2sql.jsonl", "num_epochs": 1, "job_config_name": "llm_llama3_2_1b"}' --model llm_llama3_2_1b --ft_type "LORA"

Monitoring Jobs

Monitor all submitted jobs:

lumino-client monitor-all

Development

For development setup and contributing guidelines, see DEVEL.md.