Loading
Loading
Neuron registration trends
Graph optimization problems—particularly the Traveling Salesman Problem (TSP) and its variants—are computationally expensive. Finding efficient routes across networks, depot locations, or constrained supply chains requires algorithms that balance solution quality with execution time. Traditional centralized solvers become bottlenecks when problem complexity grows or when real-time adjustments are needed across many simultaneous requests.
A decentralized network of specialized solvers can address this by distributing computation across multiple participants, each incentivized to develop and deploy high-performance algorithms. Validators generate diverse problem instances and evaluate solver submissions in real time, rewarding contributions based on solution quality relative to peers and established benchmarks. This creates competitive pressure for algorithmic innovation without centralized bottlenecks.
Graphite implements this on Bittensor by registering miners who solve TSP and multi-depot routing problems, metric TSP with constraints, and portfolio rebalancing tasks. Validators send problem instances via the Bittensor network, evaluate responses against greedy baselines and peer solutions, and assign scores that feed directly into stake-weighted rewards. Miners can use provided solvers (nearest neighbor, dynamic programming, beam search, pointer networks) or develop custom algorithms optimized for specific problem sizes and variants.
Use cases were auto-generated from the GitHub repository and may not reflect the latest changes.
Graphite is a decentralized subnet on Bittensor designed to solve graph optimization problems, particularly the Traveling Salesman Problem (TSP) and its variants. Miners compete to find optimal or near-optimal solutions to graph problems within time constraints, while validators evaluate solutions and distribute rewards based on relative performance against benchmarks and peer submissions.
Graphite operates as a decentralized network where validators generate graph optimization problems (both synthetic and organic) and send them to registered miners via the Bittensor network. Miners solve these problems using various algorithms (greedy, dynamic programming, beam search, pointer networks) and return solutions. Validators evaluate solutions against benchmarks and peer performance, calculating scores that feed into the incentive mechanism.
Validators can run synthetic problem generators or integrate with external backend systems for organic problems. The system supports multiple problem variants including Metric TSP, Multi-Depot TSP, constrained variants, and portfolio rebalancing tasks.
Rewards are distributed based on relative performance of miner solutions against peer solutions and a performance floor set by the greedy nearest-neighbor heuristic. Miners that fall below the benchmark are penalized. For portfolio rebalancing tasks, additional scoring dimensions include historical profit/loss, Sharpe ratio, trading volume, copy trader metrics, and maximum drawdown.
This information was auto-extracted from the GitHub repository and may be incomplete or out of date. Always refer to the subnet's GitHub repository for the latest instructions.
The subnet allocates 60% to graph problems (TSP variants), 20% to portfolio rebalancing, and 20% to organic validator operations.
The sections above were auto-generated from the GitHub repository and may not reflect the latest changes. View repository
FAQs were auto-generated from the GitHub repository and may not reflect the latest changes.
Holder & total-staked metrics as of 1h ago (hourly snapshot)
Minimum 4 cores at 2.5 GHz, recommended 8 cores at 3.0 GHz. Min 4GB swap, recommended 8GB swap. 50GB disk recommended.
Minimum 4 cores at 2.5 GHz, recommended 8 cores at 3.0 GHz. Min 4GB swap, recommended 8GB swap. SSD storage recommended with min 1000 IOPS, recommended 5000 IOPS. 100GB disk recommended.
Clone the Graphite subnet repository
git clone https://github.com/GraphiteAI/Graphite-Subnet.git
Navigate into the cloned repository directory
cd Graphite-Subnet
Create and activate a virtual environment (optional but recommended)(optional)
python3.10 -m venv <your_environment_name> && source <your_environment_name>/bin/activate
Disable pip cache to avoid installation issues
export PIP_NO_CACHE_DIR=1
Install Python package dependencies
pip install -r requirements.txt
Create a local editable installation of the Graphite package
pip install -e .
Create a new coldkey for your wallet
btcli wallet new_coldkey --wallet.name <your_wallet_name>
Create a new hotkey associated with your coldkey
btcli wallet new_hotkey --wallet.name <your_coldkey_name> --wallet.hotkey <your_hotkey_name>
Register your wallet on the mainnet subnet (netuid 43 for Graphite)
btcli subnet register --wallet.name <your_wallet_name> --wallet.hotkey <your_hotkey_name> --netuid 43
Run the miner neuron to begin mining on Graphite
python3 -m neurons.miner --netuid 43 --subtensor.network finney --wallet.name <NAME> --wallet.hotkey <HOTKEY> --logging.debug --axon.port <PORT> --blacklist.force_validator_permit True
Clone the Graphite subnet repository
git clone https://github.com/GraphiteAI/Graphite-Subnet.git
Navigate into the cloned repository directory
cd Graphite-Subnet
Create and activate a virtual environment (optional but recommended)(optional)
python3.10 -m venv <your_environment_name> && source <your_environment_name>/bin/activate
Disable pip cache to avoid installation issues
export PIP_NO_CACHE_DIR=1
Install Python package dependencies
pip install -r requirements.txt
Create a local editable installation of the Graphite package
pip install -e .
Create a new coldkey for your validator wallet
btcli wallet new_coldkey --wallet.name <your_wallet_name>
Create a new hotkey associated with your coldkey
btcli wallet new_hotkey --wallet.name <your_coldkey_name> --wallet.hotkey <your_hotkey_name>
Register your validator wallet on the mainnet subnet (netuid 43)
btcli subnet register --wallet.name <your_wallet_name> --wallet.hotkey <your_hotkey_name> --netuid 43
Create a W&B account for validator metrics tracking
Head to https://wandb.ai/site and create a Weights & Biases account
Set up W&B API key in environment configuration
Create a .env file with WANDB_API_KEY=YOUR_API_KEY
Run a synthetic validator that generates random graph problems
python3 neurons/validator.py --netuid 43 --subtensor.network finney --wallet.name <NAME> --wallet.hotkey <HOTKEY> --logging.trace --axon.port <PORT> --organic False