SUBMIT A PR SUBMIT AN ISSUElast edit: Jul 04, 2025

Subnet Metagraph

This page documents the Bittensor subnet metagraph.

The metagraph is a core data structure in Bittensor that represents the complete state of a subnet at any given block. It contains comprehensive information about all neurons (miners and validators) participating in a subnet, their emissions, bonds, and trust, as well as subnet metrics.

source code

The metagraph is implemented in the Bittensor blockchain (Subtensor) as a Rust data structure. The source code is located in the Subtensor repository.

Accessing the Metagraph

You can access metagraph data through multiple interfaces:

Bittensor CLI (btcli)

The btcli command-line interface provides access to a subset of metagraph information (corresponding to "lite" mode in the SDK). For full metagraph data including weights and bonds, use the Python SDK with lite=False.

# Dump metagraph subset to file (lite mode)
btcli subnets metagraph --netuid 14 --network finney \
    --json-output > sn14_metagraph.json

# View abridged metagraph
btcli subnets metagraph --netuid 14 --network finney

                                                    Subnet 14: TAOHash
                                                     Network: finney

 UID ┃ Stake (ξ) ┃ Alpha (ξ) ┃   Tao (τ) ┃ Dividends ┃ Incentive ┃ Emissions (ξ) ┃ Hotk… ┃ Coldkey ┃ Identity
━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━╇━━━━━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━
 29  │ 271.67k ξ │ 254.83k ξ │  τ 16.84k │ 0.129183  │ 0.000000  │  19.122456 ξ  │ 5Cf4… │ 5CKhH8  │ Owner14 (*Owner)
  3  │ 387.08k ξ │  61.46k ξ │ τ 325.62k │ 0.184314  │ 0.000000  │  27.280861 ξ  │ 5C59… │ 5GZSAg  │ 

...

btcli Limitations

The btcli output shows a subset of metagraph data (lite mode). For complete data including ranks, trust scores, weights, and bonds, use the Python SDK with lite=False.

Python SDK

The Bittensor Python SDK Metagraph module provides programmatic access to metagraph data:

from bittensor.core.metagraph import Metagraph

# Initialize metagraph for subnet 14 (lite mode - excludes weights/bonds)
m = Metagraph(netuid=14, network="finney", sync=True)

# Initialize metagraph with full data including weights and bonds
m = Metagraph(netuid=14, network="finney", lite=False, sync=True)

Smart Contract Access (Metagraph Precompile)

For smart contract integration, you can access metagraph data through the Metagraph Precompile at address 0x0000000000000000000000000000000000000802. This provides read-only access to individual neuron metrics and network information.

Smart Contract Integration

For detailed smart contract examples and complete ABI, see the Metagraph Precompile documentation.

RPC Functions

The blockchain provides several RPC functions for accessing metagraph data:

get_metagraph(netuid) - Returns complete metagraph for a subnet
get_all_metagraphs() - Returns metagraphs for all subnets
get_selective_metagraph(netuid, indexes) - Returns partial metagraph data

See Subtensor:Metagraph RPC source code

Performance Considerations

Lite vs Full Sync

Lite Mode (lite=True): Faster sync, excludes weights and bonds (corresponds to btcli output)
Full Mode (lite=False): Complete data including weight matrices and bond matrices

Caching

The metagraph supports local caching to persistent files:

# Save metagraph for later use
metagraph.save()

# Load cached metagraph
metagraph.load()

# Custom save directory
metagraph.save(root_dir=['/custom', 'path'])

Cache Location

Metagraph files are saved to ~/.bittensor/metagraphs/network-{network}/netuid-{netuid}/block-{block_number}.pt by default. The files are persistent and not temporary.

Source: bittensor/bittensor/core/metagraph.py:96-115

Data Structures

Metagraph Object

In the Bittensor Python SDK, the Metagraph class encapsulates the following information about a particular subnet.

Metagraph class specification, SDK reference

Metagraph Properties

Name	Description
`netuid`	The subnet's unique identifier within the Bittensor network
`network`	Name of the Bittensor network, i.e. mainnet ('finney'), test, or a locally deployed chain
`version`	Bittensor version number
`n`	Total number of neurons registered on the subnet
`block`	Block number when the metagraph record was retrieved
`total_stake`	Total stake weight (α + τ × 0.18) across all neurons
Stake / `S`	Total stake weight (α + τ × 0.18) of each neuron, determining consensus power and emissions
Alpha Stake / `AS`	Alpha token stake (α) for each neuron
Tao Stake / `TS`	TAO token stake (τ) for each neuron
Ranks / `R`	Final performance scores after consensus weight clipping - stake-weighted sum of clipped weights that directly determine emissions to miners
Trust / `T`	Consensus alignment ratio (final rank / pre-rank) - measures how much consensus clipping affected the rank, where 1.0 indicates perfect consensus alignment
Validator Trust / `Tv`	Validator trust - sum of clipped weights set by each validator, measuring validator influence in consensus
Consensus / `C`	Consensus score - stake-weighted median of weights per neuron, serving as consensus threshold for weight clipping
Incentive / `I`	Normalized ranks representing incentive allocation for miners based on performance
Emission / `E`	Token emission amounts in RAO (10^-9 TAO) per block
Dividends / `D`	Bond-based rewards for validators from their investments in miners
Bonds / `B`	Inter-neuronal bond matrix representing validator investments in miners, used to calculate validator emissions
Weights / `W`	Weight matrix (validator → miner assignments) formed from validator weight vectors, input for Yuma Consensus
`uids`	Unique UID identifiers for each neuron
`hotkeys`	Neuron hotkey addresses
`coldkeys`	Neuron coldkey addresses
`addresses`	Network IP addresses
`axons`	Network connection details for axon servers
`neurons`	Complete neuron objects with all metadata
`active`	Neuron activity status within the `activity_cutoff` window
`last_update`	Last update block numbers for staleness detection
`validator_permit`	Boolean array indicating whether each neuron has validator permits to set weights and participate in consensus
`name`	Subnet name
`symbol`	Subnet token symbol
`network_registered_at`	Registration block when subnet was created
`num_uids`	Current number of neurons
`max_uids`	Maximum allowed neurons (typically 256)
`identities`	List of chain identities
`identity`	Subnet identity information
`pruning_score`	List of pruning scores based on emissions, used for deregistration when subnet is full
`block_at_registration`	List of registration blocks for each neuron, used for immunity period calculations
`tao_dividends_per_hotkey`	TAO dividends by hotkey
`alpha_dividends_per_hotkey`	Alpha dividends by hotkey
`last_step`	Last step block number
`tempo`	Tempo - block interval for updates (360 blocks = 72 minutes)
`blocks_since_last_step`	Blocks since last step
`owner_coldkey`	Subnet owner coldkey
`owner_hotkey`	Subnet owner hotkey
`hparams`	Subnet hyperparameters (`MetagraphInfoParams`)
`pool`	Liquidity pool information (`MetagraphInfoPool`)
`emissions`	Emission configuration (`MetagraphInfoEmissions`)

Neuron Info

A neuron represents any registered participant on the subnet, whether a miner or a validator.

See also:

Neuron Info Properties

Name	Description
`uid`	Unique UID identifier for the neuron within the subnet
`hotkey`	Hotkey address for network operations and signing
`coldkey`	Coldkey address for secure storage and high-risk operations
`stake`	Total stake weight (α + τ × 0.18) determining consensus power and emissions
`rank`	Final performance rank after consensus weight clipping, directly determining emissions
`trust`	Consensus alignment ratio (final rank / pre-rank) measuring impact of consensus filtering
`consensus`	Consensus score - stake-weighted median of weights serving as clipping threshold
`incentive`	Normalized incentive score representing reward allocation for miners
`emission`	Token emission rate in RAO per block
`dividends`	Bond-based dividend amount for validators
`validator_trust`	Validator trust measuring validator influence in consensus
`active`	Activity status within the `activity_cutoff` window
`last_update`	Last update block number for staleness detection
`validator_permit`	Boolean indicating validator permit status for weight setting and consensus participation
`weights`	Weight vector assignments from this neuron to others
`bonds`	Bond investments from this neuron to others
`axon_info`	Network connection details for the axon server

Axons

An axon represents a server run by a registered miner, capable of answering requests by validators.

See also:

Axon Properties

Name	Description
`hotkey`	Neuron hotkey address
`coldkey`	Neuron coldkey address
`ip`	IP address for the axon server
`port`	Port number for the axon server
`ip_type`	IP type (IPv4/IPv6)
`version`	Protocol version for axon-dendrite communication
`placeholder1`	Reserved field for future use
`placeholder2`	Reserved field for future use

MetagraphInfoParams

Represents the hyperparameters of a subnet.

See also:

MetagraphInfoParams (Hyperparams) Properties

Name	Description
`activity_cutoff`	Activity cutoff threshold
`adjustment_alpha`	Adjustment alpha parameter
`adjustment_interval`	Adjustment interval
`alpha_high`	Alpha high threshold
`alpha_low`	Alpha low threshold
`bonds_moving_avg`	Bonds moving average
`burn`	Burn amount
`commit_reveal_period`	Commit reveal period
`commit_reveal_weights_enabled`	Commit reveal weights enabled
`difficulty`	Network difficulty
`immunity_period`	Immunity period
`kappa`	Kappa parameter
`liquid_alpha_enabled`	Liquid alpha enabled
`max_burn`	Maximum burn
`max_difficulty`	Maximum difficulty
`max_regs_per_block`	Max registrations per block
`max_validators`	Maximum validators
`max_weights_limit`	Maximum weights limit
`min_allowed_weights`	Minimum allowed weights
`min_burn`	Minimum burn
`min_difficulty`	Minimum difficulty
`pow_registration_allowed`	POW registration allowed
`registration_allowed`	Registration allowed
`rho`	Rho parameter
`serving_rate_limit`	Serving rate limit
`target_regs_per_interval`	Target registrations per interval
`tempo`	Tempo - block interval for updates (360 blocks = 72 minutes)
`weights_rate_limit`	Weights rate limit for submissions
`weights_version`	Weights version for protocol compatibility

MetagraphInfoPool

Contains information about the subnet's liquidity pool

See also:

MetagraphInfoPool properties

Name	Description
`alpha_out`	Alpha token quanitity bound for emission to subnet participants
`alpha_in`	Alpha token quanitity emitted to the liquidity pool
`tao_in`	Tao token emission to the liquidity pool
`subnet_volume`	Total trading volume in the subnet's liquidity pool
`moving_price`	Moving average price of the subnet token

MetagraphInfoEmissions

Contains detailed information about the subnet's emissions.

See also:

MetagraphInfoEmissions properties

Name	Description
`alpha_out_emission`	Alpha token outflow emission rate
`alpha_in_emission`	Alpha token inflow emission rate
`subnet_emission`	Subnet emission rate to participants
`tao_in_emission`	TAO token inflow emission rate
`pending_alpha_emission`	Pending alpha token emission amount
`pending_root_emission`	Pending root network emission amount

Python Code Examples

This section provides practical examples of working with the Bittensor metagraph using the Python SDK. Each example demonstrates different aspects of metagraph analysis and data extraction.

Code examples can be found here

Prerequisites:

Bittensor Python SDK installed (pip install bittensor)
Network connection to access Bittensor blockchain
Python 3.7+ environment

Basic Metagraph Information

This example shows how to access basic metagraph metadata and subnet information:

#!/usr/bin/env python3

from bittensor.core.metagraph import Metagraph

def main():
    # Initialize metagraph for subnet 1
    print("Initializing metagraph for subnet 1...")
    metagraph = Metagraph(netuid=1, network="finney", sync=True)
    
    # Get basic metagraph metadata
    print("\n=== Basic Metagraph Metadata ===")
    print(f"Network: {metagraph.network}")
    print(f"Subnet UID: {metagraph.netuid}")
    print(f"Total neurons: {metagraph.n.item()}")
    print(f"Current block: {metagraph.block.item()}")
    print(f"Version: {metagraph.version.item()}")

    # Get subnet information
    print("\n=== Subnet Information ===")
    print(f"Subnet name: {metagraph.name}")
    print(f"Subnet symbol: {metagraph.symbol}")
    print(f"Registered at block: {metagraph.network_registered_at}")
    print(f"Max UIDs: {metagraph.max_uids}")
    print(f"Owner: {metagraph.owner_coldkey}")

if __name__ == "__main__":
    main() 

Neuron Metrics Analysis

This example demonstrates stake distribution and neuron metrics analysis:

#!/usr/bin/env python3

from bittensor.core.metagraph import Metagraph

def main():
    # Initialize metagraph for subnet 1
    print("Initializing metagraph for subnet 1...")
    metagraph = Metagraph(netuid=1, network="finney", sync=True)
    
    # Get all neuron UIDs
    uids = metagraph.uids
    print(f"\nNeuron UIDs: {uids.tolist()}")

    # Analyze stake distribution
    stakes = metagraph.S  # Total stake
    alpha_stakes = metagraph.AS  # Alpha token stake
    tao_stakes = metagraph.TS  # TAO token stake

    print(f"\n=== Stake Analysis ===")
    print(f"Total stake across all neurons: {stakes.sum().item():.2f}")
    print(f"Average stake per neuron: {stakes.mean().item():.2f}")
    print(f"Highest stake: {stakes.max().item():.2f}")
    print(f"Lowest stake: {stakes.min().item():.2f}")

    # Find top staked neurons
    top_staked_indices = stakes.argsort()[::-1][:10]
    print("\nTop 10 staked neurons:")
    for i, idx in enumerate(top_staked_indices):
        uid = uids[idx].item()
        stake = stakes[idx].item()
        print(f"  {i+1}. UID {uid}: {stake:.2f} τ")

if __name__ == "__main__":
    main() 

Performance and Ranking Analysis

This example shows how to analyze neuron performance, ranks, and trust scores:

#!/usr/bin/env python3

from bittensor.core.metagraph import Metagraph

def main():
    # Initialize metagraph for subnet 1
    print("Initializing metagraph for subnet 1...")
    metagraph = Metagraph(netuid=1, network="finney", sync=True)
    
    # Get performance metrics
    ranks = metagraph.R  # Performance ranks
    trust = metagraph.T  # Trust scores
    consensus = metagraph.C  # Consensus scores
    validator_trust = metagraph.Tv  # Validator trust
    uids = metagraph.uids

    # Find top performing neurons
    top_ranked_indices = ranks.argsort()[::-1][:10]
    print("\n=== Top 10 Ranked Neurons ===")
    for i, idx in enumerate(top_ranked_indices):
        uid = uids[idx].item()
        rank = ranks[idx].item()
        trust_score = trust[idx].item()
        consensus_score = consensus[idx].item()
        print(f"  {i+1}. UID {uid}: Rank={rank:.4f}, Trust={trust_score:.4f}, Consensus={consensus_score:.4f}")

    # Analyze trust distribution
    print(f"\n=== Trust Analysis ===")
    print(f"Average trust score: {trust.mean().item():.4f}")
    print(f"Trust score std dev: {trust.std().item():.4f}")
    print(f"Highest trust: {trust.max().item():.4f}")
    print(f"Lowest trust: {trust.min().item():.4f}")

    # Find most trusted validators
    validator_indices = metagraph.validator_permit.nonzero()[0]
    if len(validator_indices) > 0:
        validator_trust_scores = validator_trust[validator_indices]
        top_validators = validator_indices[validator_trust_scores.argsort()[::-1][:5]]
        print("\n=== Top 5 Trusted Validators ===")
        for i, idx in enumerate(top_validators):
            uid = uids[idx].item()
            vtrust = validator_trust[idx].item()
            print(f"  {i+1}. UID {uid}: Validator Trust={vtrust:.4f}")
    else:
        print("\nNo validators found in this subnet.")

if __name__ == "__main__":
    main() 

Economic Analysis

This example demonstrates analysis of economic metrics like incentives, emissions, and dividends:

#!/usr/bin/env python3

from bittensor.core.metagraph import Metagraph

def main():
    # Initialize metagraph for subnet 1
    print("Initializing metagraph for subnet 1...")
    metagraph = Metagraph(netuid=1, network="finney", sync=True)
    
    # Get economic metrics
    incentives = metagraph.I  # Incentive scores
    emissions = metagraph.E  # Emission rates
    dividends = metagraph.D  # Dividend distributions
    uids = metagraph.uids

    # Analyze incentive distribution
    print(f"\n=== Incentive Analysis ===")
    print(f"Total incentives: {incentives.sum().item():.4f}")
    print(f"Average incentive: {incentives.mean().item():.4f}")
    print(f"Highest incentive: {incentives.max().item():.4f}")

    # Find highest incentivized neurons
    top_incentive_indices = incentives.argsort()[::-1][:10]
    print("\n=== Top 10 Incentivized Neurons ===")
    for i, idx in enumerate(top_incentive_indices):
        uid = uids[idx].item()
        incentive = incentives[idx].item()
        emission = emissions[idx].item()
        dividend = dividends[idx].item()
        print(f"  {i+1}. UID {uid}: Incentive={incentive:.4f}, Emission={emission:.4f}, Dividend={dividend:.4f}")

    # Analyze dividend distribution
    print(f"\n=== Dividend Analysis ===")
    print(f"Total dividends: {dividends.sum().item():.4f}")
    print(f"Average dividend: {dividends.mean().item():.4f}")
    print(f"Dividend std dev: {dividends.std().item():.4f}")

if __name__ == "__main__":
    main() 

Network Connectivity Analysis

This example shows how to analyze network addresses and axon information:

#!/usr/bin/env python3

from bittensor.core.metagraph import Metagraph

def main():
    # Initialize metagraph for subnet 1
    print("Initializing metagraph for subnet 1...")
    metagraph = Metagraph(netuid=1, network="finney", sync=True)
    
    # Get network information
    axons = metagraph.axons
    uids = metagraph.uids

    # Analyze network addresses
    addresses = [axon.ip for axon in axons]
    unique_addresses = set(addresses)
    unique_hotkeys = set(metagraph.hotkeys)
    unique_coldkeys = set(metagraph.coldkeys)

    print(f"\n=== Network Address Analysis ===")
    print(f"Total unique addresses: {len(unique_addresses)}")
    print(f"Total unique hotkeys: {len(unique_hotkeys)}")
    print(f"Total unique coldkeys: {len(unique_coldkeys)}")

    # Find neurons sharing addresses
    address_to_uids = {}
    for i, address in enumerate(addresses):
        if address not in address_to_uids:
            address_to_uids[address] = []
        address_to_uids[address].append(uids[i].item())

    print(f"\n=== Neurons Sharing Addresses ===")
    for address, uids_list in address_to_uids.items():
        if len(uids_list) > 1:
            print(f"  Address {address}: UIDs {uids_list}")

    # Show axon details for first few neurons
    print(f"\n=== Axon Details (First 5 Neurons) ===")
    for i in range(min(5, len(axons))):
        axon = axons[i]
        uid = uids[i].item()
        hotkey = metagraph.hotkeys[i][:10] + "..."
        print(f"  UID {uid}: IP={axon.ip}, Port={axon.port}, Hotkey={hotkey}")

if __name__ == "__main__":
    main() 

Weight Matrix Analysis

This example demonstrates weight matrix analysis (requires lite=False):

#!/usr/bin/env python3

from bittensor.core.metagraph import Metagraph

def main():
    # Initialize metagraph for subnet 1 with full sync (not lite)
    print("Initializing metagraph for subnet 1 (full sync)...")
    metagraph = Metagraph(netuid=1, network="finney", lite=False, sync=True)
    
    uids = metagraph.uids
    
    # Get weight matrix (requires lite=False)
    if not metagraph.lite and hasattr(metagraph, 'weights') and metagraph.weights.size > 0:
        weights = metagraph.W  # Weight matrix
        
        print(f"\n=== Weight Matrix Analysis ===")
        print(f"Weight matrix shape: {weights.shape}")
        print(f"Total weights: {weights.sum().item():.4f}")
        print(f"Average weight: {weights.mean().item():.4f}")
        print(f"Max weight: {weights.max().item():.4f}")
        
        # Find miners receiving most weights
        weight_received = weights.sum(axis=0)  # Sum of incoming weights
        top_receivers = weight_received.argsort()[::-1][:10]
        print("\n=== Miners Receiving Most Frequent Weights ===")
        for i, idx in enumerate(top_receivers):
            uid = uids[idx].item()
            total_weight = weight_received[idx].item()
            print(f"  {i+1}. UID {uid}: {total_weight:.4f}")
        
        # Find validators sending most weights
        weight_sent = weights.sum(axis=1)  # Sum of outgoing weights
        top_senders = weight_sent.argsort()[::-1][:10]
        print("\n=== Validators Setting Weights Most Frequently ===")
        for i, idx in enumerate(top_senders):
            uid = uids[idx].item()
            total_weight = weight_sent[idx].item()
            print(f"  {i+1}. UID {uid}: {total_weight:.4f}")
        
        # Find highest set weight
        max_weight_idx = weights.argmax()
        sender_idx = max_weight_idx // weights.shape[1]
        receiver_idx = max_weight_idx % weights.shape[1]
        max_weight = weights.max().item()
        print(f"\n=== Highest Single Set Weight ===")
        print(f"UID {uids[sender_idx].item()} -> UID {uids[receiver_idx].item()}: {max_weight:.4f}")
    else:
        print("Weights not available. Make sure to use lite=False when initializing the metagraph.")

if __name__ == "__main__":
    main() 

Bond Analysis

This example shows bond matrix analysis (requires lite=False):

#!/usr/bin/env python3

from bittensor.core.metagraph import Metagraph

def main():
    # Initialize metagraph for subnet 1 with full sync (not lite)
    print("Initializing metagraph for subnet 1 (full sync)...")
    metagraph = Metagraph(netuid=1, network="finney", lite=False, sync=True)
    
    uids = metagraph.uids
    
    # Get bond matrix (requires lite=False)
    if not metagraph.lite and hasattr(metagraph, 'bonds') and metagraph.bonds.size > 0:
        bonds = metagraph.B  # Bond matrix
        
        print(f"\n=== Bond Matrix Analysis ===")
        print(f"Bond matrix shape: {bonds.shape}")
        print(f"Total bonds: {bonds.sum().item():.4f}")
        print(f"Average bond: {bonds.mean().item():.4f}")
        
        # Find miners with most bonds
        bonds_received = bonds.sum(axis=0)  # Sum of incoming bonds
        top_bonded = bonds_received.argsort()[::-1][:10]
        print("\n=== Top 10 Most Bonded Miners ===")
        for i, idx in enumerate(top_bonded):
            uid = uids[idx].item()
            total_bonds = bonds_received[idx].item()
            print(f"  {i+1}. UID {uid}: {total_bonds:.4f}")
    else:
        print("Bonds not available. Make sure to use lite=False when initializing the metagraph.")

if __name__ == "__main__":
    main() 

Neuron Activity Analysis

This example demonstrates analyzing validator activity:

#!/usr/bin/env python3

from bittensor.core.metagraph import Metagraph

def main():
    # Initialize metagraph for subnet 1
    print("Initializing metagraph for subnet 1...")
    metagraph = Metagraph(netuid=1, network="finney", sync=True)
    
    # Get activity information
    active = metagraph.active  # Activity status
    last_update = metagraph.last_update  # Last update blocks
    validator_permit = metagraph.validator_permit  # Validator permissions
    uids = metagraph.uids
    stakes = metagraph.S
    ranks = metagraph.R

    # Analyze activity
    active_count = active.sum().item()
    total_count = len(active)
    print(f"\n=== Activity Analysis ===")
    print(f"Active validators: {active_count}/{total_count} ({active_count/total_count:.1%})")


    # Analyze validator distribution
    validator_count = validator_permit.sum().item()
    print(f"\n=== Validator Analysis ===")
    print(f"Validators: {validator_count}/{total_count} ({validator_count/total_count:.1%})")

    # Find validators
    validator_indices = validator_permit.nonzero()[0]
    if len(validator_indices) > 0:
        print("\n=== Validators ===")
        for idx in validator_indices:
            uid = uids[idx].item()
            stake = stakes[idx].item()
            rank = ranks[idx].item()
            print(f"  UID {uid}: Stake={stake:.2f}, Rank={rank:.4f}")
    else:
        print("\nNo validators found in this subnet.")

if __name__ == "__main__":
    main() 

Subnet Economics

This example shows how to access subnet hyperparameters, pool, and emissions:

#!/usr/bin/env python3

from bittensor.core.metagraph import Metagraph

def main():
    # Initialize metagraph for subnet 1
    print("Initializing metagraph for subnet 1...")
    metagraph = Metagraph(netuid=1, network="finney", sync=True)
    
    # Get subnet hyperparameters
    hparams = metagraph.hparams
    print(f"\n=== Subnet Hyperparameters ===")
    print(f"  Activity cutoff: {hparams.activity_cutoff}")
    print(f"  Adjustment alpha: {hparams.adjustment_alpha}")
    print(f"  Adjustment interval: {hparams.adjustment_interval}")
    print(f"  Alpha high: {hparams.alpha_high}")
    print(f"  Alpha low: {hparams.alpha_low}")
    print(f"  Burn rate: {hparams.burn_rate}")
    print(f"  Max burn: {hparams.max_burn}")
    print(f"  Min burn: {hparams.min_burn}")
    print(f"  Difficulty: {hparams.difficulty}")
    print(f"  Max difficulty: {hparams.max_difficulty}")
    print(f"  Min difficulty: {hparams.min_difficulty}")
    print(f"  Max validators: {hparams.max_validators}")
    print(f"  Tempo: {hparams.tempo}")
    print(f"  Weights version: {hparams.weights_version}")

    # Get subnet pool information
    pool = metagraph.pool
    print(f"\n=== Subnet Pool ===")
    print(f"  Alpha out: {pool.alpha_out}")
    print(f"  Alpha in: {pool.alpha_in}")
    print(f"  TAO in: {pool.tao_in}")
    print(f"  Subnet volume: {pool.subnet_volume}")
    print(f"  Moving price: {pool.moving_price}")

    # Get subnet emissions
    emissions = metagraph.emissions
    print(f"\n=== Subnet Emissions ===")
    print(f"  Alpha out emission: {emissions.alpha_out_emission}")
    print(f"  Alpha in emission: {emissions.alpha_in_emission}")
    print(f"  Subnet emission: {emissions.subnet_emission}")
    print(f"  TAO in emission: {emissions.tao_in_emission}")
    print(f"  Pending alpha emission: {emissions.pending_alpha_emission}")
    print(f"  Pending root emission: {emissions.pending_root_emission}")

if __name__ == "__main__":
    main() 

Advanced Analysis Examples

This example demonstrates advanced analysis techniques including correlations and Gini coefficient of stake distribution.

#!/usr/bin/env python3

import numpy as np
from bittensor.core.metagraph import Metagraph

def main():
    # Initialize metagraph for subnet 1
    print("Initializing metagraph for subnet 1...")
    metagraph = Metagraph(netuid=1, network="finney", sync=True)
    
    # Get basic metrics
    stakes = metagraph.S
    ranks = metagraph.R
    trust = metagraph.T
    uids = metagraph.uids

    # Correlation analysis between metrics
    print("\n=== Metric Correlations ===")
    try:
        # Calculate correlations
        stake_rank_corr = np.corrcoef(stakes, ranks)[0, 1]
        stake_trust_corr = np.corrcoef(stakes, trust)[0, 1]
        rank_trust_corr = np.corrcoef(ranks, trust)[0, 1]

        print("Metric Correlations:")
        print(f"  Stake vs Rank: {stake_rank_corr:.4f}")
        print(f"  Stake vs Trust: {stake_trust_corr:.4f}")
        print(f"  Rank vs Trust: {rank_trust_corr:.4f}")
    except Exception as e:
        print(f"Could not calculate correlations: {e}")

    # Network efficiency analysis (if weights are available)
    if not metagraph.lite and hasattr(metagraph, 'weights') and metagraph.weights.size > 0:
        weights = metagraph.W
        
        print("\n=== Network Efficiency Analysis ===")
        # Calculate network efficiency (average path length)
        non_zero_weights = weights[weights > 0]
        if len(non_zero_weights) > 0:
            avg_weight = non_zero_weights.mean().item()
            weight_std = non_zero_weights.std().item()
            print(f"Network efficiency:")
            print(f"  Average non-zero weight: {avg_weight:.4f}")
            print(f"  Weight standard deviation: {weight_std:.4f}")
            print(f"  Weight distribution CV: {weight_std/avg_weight:.4f}")
    else:
        print("\nWeights not available for network efficiency analysis.")

    # Stake concentration analysis
    print("\n=== Stake Concentration Analysis ===")
    total_stake = stakes.sum().item()
    try:
        stake_percentiles = np.percentile(stakes, [25, 50, 75, 90, 95, 99])
        print("Stake distribution percentiles:")
        for p, val in zip([25, 50, 75, 90, 95, 99], stake_percentiles):
            print(f"  {p}th percentile: {val:.2f} τ")

        # Gini coefficient for stake inequality
        sorted_stakes = np.sort(stakes)
        n = len(sorted_stakes)
        cumulative_stakes = np.cumsum(sorted_stakes)
        gini = (n + 1 - 2 * np.sum(cumulative_stakes) / cumulative_stakes[-1]) / n
        print(f"Stake Gini coefficient: {gini:.4f}")
    except Exception as e:
        print(f"Could not calculate stake concentration metrics: {e}")

if __name__ == "__main__":
    main() 

Async Usage

This example demonstrates async metagraph usage:

#!/usr/bin/env python3

import asyncio
from bittensor.core.metagraph import async_metagraph

async def analyze_metagraph():
    # Create async metagraph
    print("Creating async metagraph...")
    metagraph = await async_metagraph(netuid=1, network="finney", lite=False)
    
    # Perform analysis
    stakes = metagraph.S
    print(f"Total stake: {stakes.sum().item():.2f}")
    
    # Sync to latest block
    print("Syncing to latest block...")
    await metagraph.sync()
    print(f"Synced to block: {metagraph.block.item()}")

async def main():
    print("=== Async Metagraph Analysis ===")
    await analyze_metagraph()

if __name__ == "__main__":
    # Run async analysis
    asyncio.run(main()) 

Complete Neuron Information

This example shows how to access complete neuron object information:

#!/usr/bin/env python3

from bittensor.core.metagraph import Metagraph

def main():
    # Initialize metagraph for subnet 1
    print("Initializing metagraph for subnet 1...")
    metagraph = Metagraph(netuid=1, network="finney", sync=True)
    
    # Get complete neuron information for first 5 neurons
    print("=== Complete Neuron Information (First 5 Neurons) ===")
    
    for i in range(min(5, metagraph.n.item())):
        neuron = metagraph.neurons[i]
        print(f"\nNeuron {i}:")
        print(f"  UID: {neuron.uid}")
        print(f"  Hotkey: {neuron.hotkey}")
        print(f"  Coldkey: {neuron.coldkey}")
        print(f"  Stake: τ{neuron.stake:.9f}")
        print(f"  Rank: {neuron.rank}")
        print(f"  Trust: {neuron.trust}")
        print(f"  Consensus: {neuron.consensus}")
        print(f"  Incentive: {neuron.incentive}")
        print(f"  Emission: {neuron.emission}")
        print(f"  Dividends: {neuron.dividends}")
        print(f"  Active: {neuron.active}")
        print(f"  Last update: {neuron.last_update}")
        print(f"  Validator permit: {neuron.validator_permit}")
        print(f"  Validator trust: {neuron.validator_trust}")
        print(f"  Axon IP: {neuron.axon.ip}")
        print(f"  Axon port: {neuron.axon.port}")
        print(f"  ---")

if __name__ == "__main__":
    main() 

Source Code References

Core Implementation

Metagraph Class: bittensor/bittensor/core/metagraph.py
Chain Data: bittensor/bittensor/core/chain_data/metagraph_info.py
Subtensor RPC: subtensor/pallets/subtensor/src/rpc_info/metagraph.rs

Consensus Algorithm

Yuma Consensus: subtensor/pallets/subtensor/src/epoch/run_epoch.rs
Mathematical Operations: subtensor/pallets/subtensor/src/epoch/math.rs

Key Constants

TAO Stake Weight: bittensor/bittensor/core/settings.py:7 - ROOT_TAO_STAKE_WEIGHT = 0.18

Accessing the Metagraph​

Bittensor CLI (btcli)​

Python SDK​

Smart Contract Access (Metagraph Precompile)​

RPC Functions​

Performance Considerations​

Lite vs Full Sync​

Caching​

Data Structures​

Metagraph Object​

Neuron Info​

Axons​

MetagraphInfoParams​

MetagraphInfoPool​

MetagraphInfoEmissions​

Python Code Examples​

Basic Metagraph Information​

Neuron Metrics Analysis​

Performance and Ranking Analysis​

Economic Analysis​

Network Connectivity Analysis​

Weight Matrix Analysis​

Bond Analysis​

Neuron Activity Analysis​

Subnet Economics​

Advanced Analysis Examples​

Async Usage​

Complete Neuron Information​

Source Code References​

Core Implementation​

Consensus Algorithm​

Key Constants​

Accessing the Metagraph

Bittensor CLI (btcli)

Python SDK

Smart Contract Access (Metagraph Precompile)

RPC Functions

Performance Considerations

Lite vs Full Sync

Caching

Data Structures

Metagraph Object

Neuron Info

Axons

MetagraphInfoParams

MetagraphInfoPool

MetagraphInfoEmissions

Python Code Examples

Basic Metagraph Information

Neuron Metrics Analysis

Performance and Ranking Analysis

Economic Analysis

Network Connectivity Analysis

Weight Matrix Analysis

Bond Analysis

Neuron Activity Analysis

Subnet Economics

Advanced Analysis Examples

Async Usage

Complete Neuron Information

Source Code References

Core Implementation

Consensus Algorithm

Key Constants