BitMind-AI · dylanuys · May 14, 2025 · May 2, 2025 · May 8, 2025 · May 8, 2025
diff --git a/.env.miner.template b/.env.miner.template
@@ -0,0 +1,21 @@
+# ======= Miner Configuration (FILL IN) =======
+# Wallet
+WALLET_NAME=
+WALLET_HOTKEY=
+
+# Network
+CHAIN_ENDPOINT=wss://entrypoint-finney.opentensor.ai:443
+# OTF public finney endpoint:  wss://entrypoint-finney.opentensor.ai:443
+# OTF public testnet endpoint: wss://test.finney.opentensor.ai:443/
+
+# Axon port and (optionally) ip
+AXON_PORT=8091
+AXON_EXTERNAL_IP=[::]
+
+FORCE_VPERMIT=true
+
+# Device for detection models
+DEVICE=cpu
+
+# Logging
+LOGLEVEL=trace
diff --git a/.env.validator.template b/.env.validator.template
@@ -0,0 +1,29 @@
+# ======= Validator Configuration (FILL IN) =======
+# Wallet
+WALLET_NAME=
+WALLET_HOTKEY=
+
+# API Keys
+WANDB_API_KEY=
+HUGGING_FACE_TOKEN=
+
+# Network
+CHAIN_ENDPOINT=wss://entrypoint-finney.opentensor.ai:443
+# OTF public finney endpoint:  wss://entrypoint-finney.opentensor.ai:443
+# OTF public testnet endpoint: wss://test.finney.opentensor.ai:443/
+
+# Validator Proxy
+PROXY_PORT=10913
+PROXY_EXTERNAL_PORT=10913
+
+# Cache config
+SN34_CACHE_DIR=~/.cache/sn34
+HEARTBEAT=true
+
+# Generator config
+GENERATION_BATCH_SIZE=3
+DEVICE=cuda
+
+# Other
+LOGLEVEL=info
+AUTO_UPDATE=false
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
diff --git a/.gitmodules b/.gitmodules
diff --git a/LICENSE b/LICENSE
@@ -1,21 +1,17 @@
-MIT License
+The MIT License (MIT)
+Copyright © 2023 Yuma Rao
+Copyright © 2025 BitMind
 
-Copyright (c) 2023 Opentensor
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated
+documentation files (the "Software"), to deal in the Software without restriction, including without limitation
+the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software,
+and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
 
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of
+the Software.
 
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO
+THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+DEALINGS IN THE SOFTWARE.
diff --git a/README.md b/README.md
@@ -1,105 +1,87 @@
 <p align="center">
-  <img src="static/Bitmind-Logo.png" alt="BitMind Logo" width="150"/>
+  <img src="docs/static/Bitmind-Logo.png" alt="BitMind Logo" width="150"/>
 </p>
-<h1 align="center">BitMind Subnet<br><small>Bittensor Subnet 34 | Deepfake Detection</small></h1>
-
-![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)
-
-The BitMind Subnet is **the world's first decentralized AI-generated content detection network**. Built on Bittensor, our incentive mechanism rewards the most accurate detection algorithms, creating an adaptive defense against synthetic media.
-
-<table style="border: none !important; width: 100% !important; border-collapse: collapse !important; margin: 0 auto !important;">
-  <tbody>
-    <tr>
-      <td><b>Docs</b></td>
-      <td><b>Resources</b></td>
-    </tr>
-    <tr style="vertical-align: top !important">
-      <td>
-        ⛏️ <a href="docs/Mining.md">Mining Guide</a><br>
-        🔧 <a href="docs/Validating.md">Validator Guide</a><br>
-        🏗️ <a href="#Subnet-Architecture">Architecture Diagrams</a><br>
-        📈 <a href="docs/Incentive.md">Incentive Mechanism</a><br>
-        🤝 <a href="docs/Contributor_Guide.md">Contributor Guide</a></td>
-      <td>
-        🚀 <a href="<https://www.bitmind.ai/apps>">SN34-Powered Applications</a><br>
-        🤗 <a href="https://huggingface.co/bitmind">BitMind Huggingface</a><br>
-        📊 <a href="https://wandb.ai/bitmindai/bitmind-subnet">Mainnet 34 W&B</a> | <a href="https://wandb.ai/bitmindai/bitmind">Testnet 168 W&B</a><br>
-        📖 <a href="docs/Glossary.md">Project Structure and Terminology</a><br>
-        <a href="https://docs.bittensor.com/learn/bittensor-building-blocks">🧠 Bittensor Introduction</a><br> 
-      </td>
-    </tr>
-  </tbody>
-</table>
+<h1 align="center">SN34<br><small>Deepfake Detection</small></h1>
 
+<div align="center">
+ <a href="https://www.bitmind.ai/apps">Applications</a>
+</div>
+
+<div align="center">
+ <a href="docs/Mining.md">Mining Guide</a> ·
+ <a href="docs/Validating.md">Validator Guide</a> ·
+  <a href="docs/Incentive.md">Incentive Mechanism Overview</a>
+</div>
+
+<div align="center">
+ <a href="https://huggingface.co/bitmind">HuggingFace</a> ·
+ <a href="https://wandb.ai/bitmindai/bitmind-subnet">Mainnet 34 W&B</a> ·
+ <a href="https://wandb.ai/bitmindai/bitmind-subnet">Testnet 168 W&B</a>
+</div>
+
+<div align="center">
+ <a href="https://app.bitmind.ai/statistics">Leaderboard</a>
+</div>
 
 ## Decentralized Detection of AI Generated Content
-The explosive growth of generative AI technology has unleashed an unprecedented wave of synthetic media creation. These AI-generated images, videos, and other content have become remarkably sophisticated, making them virtually indistinguishable from authentic media. This development presents a critical challenge to information integrity and societal trust in the digital age, as the line between real and synthetic content continues to blur.
+The explosive growth of generative AI technology has unleashed an unprecedented wave of synthetic media creation. AI-generated audiovisual content has become remarkably sophisticated, oftentimes indistinguishable from authentic media. This development presents a critical challenge to information integrity and societal trust in the digital age, as the line between real and synthetic content continues to blur.
 
 To address this growing challenge, SN34 aims to create the most accurate fully-generalized detection system. Here, fully-generalized means that the system is capable of detecting both synthetic and semi-synthetic media with high degrees of accuracy regardless of their content or what model generated them. Our incentive mechanism evolves alongside state-of-the-art generative AI, rewarding miners whose detection algorithms best adapt to new forms of synthetic content.
 
 
 ## Core Components
 
-> This documentation assumes basic familiarity with Bittensor concepts. For an introduction, please check out the docs: https://docs.bittensor.com/learn/bittensor-building-blocks. 
+> This documentation assumes basic familiarity with [Bittensor concepts](https://docs.bittensor.com/learn/bittensor-building-blocks). 
+
+<b><a href="docs/Mining.md">Miners</a></b>
 
-**Miners** 
 - Miners are tasked with running binary classifiers that discern between genuine and AI-generated content, and are rewarded based on their accuracy. 
-- Miners predict a float value in [0., 1.], with values greater than 0.5 indicating the image or video is AI generated. 
+- For each challenge, a miner is presented an image or video and is required to respond with a multiclass prediction [$p_{real}$, $p_{synthetic}$, $p_{semisynthetic}$] indicating whether the media is real, fully generated, or partially modified by AI.
 
 
-**Validators** 
+<b><a href="docs/Validating.md">Validators</a></b>
 - Validators challenge miners with a balanced mix of real and synthetic media drawn from a diverse pool of sources.
-- We continually add new datasets and generative models to our validators in order to maximize coverage of the types of diverse data. Models and datasets are defined in  `bitmind/validator/config.py`.
+- We continually add new datasets and generative models to our validators in order to evolve the subnet's detection capabilities alongside advances in generative AI. 
 
 
 ## Subnet Architecture
 
-> Overview of the validator neuron, miner neuron, and other components external to the subnet.
-
-![Subnet Architecture](static/Subnet-Arch.png)
-
-**Challenge Generation and Scoring (Pink Arrows)**
-
-For each challenge, the validator randomly samples a real or synthetic image/video from the cache, applies random augmentations to the sampled media, and distributes the augmented data to 50 randomly selected miners for classification. It then scores the miners responses, and logs comprehensive challenge results to [Weights and Biases](https://wandb.ai/bitmindai/bitmind-subnet), including the generated media, original prompt, miner responses and rewards, and other challenge metadata.
-
-**Data Generation and Downloads (Blue Arrows)**:
-
-The synthetic data generator coordinates a VLM and LLM to generate prompts for our suite of text-to-image, image-to-image, and text-to-video models. Each generated image/video is written to the cache along with the prompt, generation parameters, and other metadata.
-
-The real data fetcher performs partial dataset downloads, fetching random compressed chunks of datasets from HuggingFace and unpacking random portions of these chunks into the cache along with their metadata. Partial downloads avoid requiring TBs of space for large video datasets like OpenVid1M.
-
-**Organic Traffic (Green Arrows)**
+Overview of the validator neuron, miner neuron, and other components external to the subnet.
+
+![Subnet Architecture](docs/static/Subnet-Arch.png)
+
+<details>
+<summary><b>Challenge Generation and Scoring (Peach Arrows</span>)</b></summary>
+<ul>
+  <li>The validator first randomly samples an image or video from its local media cache.</li>
+  <li>The sampled media can be real, synthetic, or semisynthetic, and was either downloaded from an dataset on Huggingface or generated locally by one of many generative models.</li>
+  <li>The sampled media is then augmented by a pipeline of random transformations, adding to the challenge difficulty and mitigating incentive mechanism gaming via lookups.</li>
+  <li>The augmented media is then sent to miners for classification.</li>
+  <li>The validator scores the miners responses and logs comprehensive challenge results to <a href="https://wandb.ai/bitmindai/bitmind-subnet">Weights and Biases</a>, including the generated media, original prompt, miner responses and rewards, and other challenge metadata.</li>
+</ul>
+</details>
+
+<details>
+<summary><b>Data Generation and Downloads (Blue Arrows)</b></summary>
+The blue arrows show how the validator media cache is maintained by two parallel tracks:
+<ul>
+<li>The synthetic data generator coordinates a VLM and LLM to generate prompts for our suite of text-to-image, image-to-image, and text-to-video models. Each generated image/video is written to the cache along with the prompt, generation parameters, and other metadata.</li>
+<li>The real data fetcher performs partial dataset downloads, fetching random compressed chunks of datasets from HuggingFace and unpacking random portions of these chunks into the cache along with their metadata. Partial downloads avoid requiring TBs of space for large video datasets like OpenVid1M.</li>
+</ul>
+</details>
+
+<details>
+<summary><b>Organic Traffic (Green Arrows)</b></summary>
 
 <a href="https://www.bitmind.ai/apps">Application</a> requests are distributed to validators by an API server and load balancer in BitMind's cloud. A vector database caches subnet responses to avoid uncessary repetitive calls coming from salient images on the internet. 
+</details>
+
 
 
 ## Community
 
 <p align="left">
   <a href="https://discord.gg/kKQR98CrUn">
-    <img src="static/Join-BitMind-Discord.png" alt="Join us on Discord" width="60%">
+    <img src="docs/static/Join-BitMind-Discord.png" alt="Join us on Discord" width="60%">
   </a>
 </p>
-
-For real-time discussions, community support, and regular updates, <a href="https://discord.gg/kKQR98CrUn">join our Discord server</a>. Connect with developers, researchers, and users to get the most out of BitMind Subnet.
-
-## License
-This repository is licensed under the MIT License.
-```text
-# The MIT License (MIT)
-# Copyright © 2023 Yuma Rao
-
-# Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated
-# documentation files (the “Software”), to deal in the Software without restriction, including without limitation
-# the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software,
-# and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
-
-# The above copyright notice and this permission notice shall be included in all copies or substantial portions of
-# the Software.
-
-# THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO
-# THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
-# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-# OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
-# DEALINGS IN THE SOFTWARE.
-```
diff --git a/VERSION b/VERSION
@@ -0,0 +1 @@
+3.0.0
diff --git a/autoupdate_miner_steps.sh b/autoupdate_miner_steps.sh
diff --git a/autoupdate_validator_steps.sh b/autoupdate_validator_steps.sh
diff --git a/base_miner/DFB/README.md b/base_miner/DFB/README.md
diff --git a/base_miner/DFB/config/__init__.py b/base_miner/DFB/config/__init__.py
diff --git a/base_miner/DFB/config/constants.py b/base_miner/DFB/config/constants.py