@BittAudio

The Bittensor Music Subnet is BACK! 🎶

Feel the hype, feel the rhythm, and let the vibes flow!

🚨How can you help the Music Subnet grow within the Bittensor ecosystem?

💡It’s simple! If you have $TAO delegated to a Validator, here’s what you can do: 1⃣ Share this announcement with Validators, Opentensor Foundation and the Bittensor community 2⃣ Ask your Validator to set weights to Subnet 50 3⃣ Delegate weights yourself using Datura Validator or Miner's Union Validator

Subnet: 50

**LET THERE BE MUSIC! **

PS: We're still preparing some thing, we'll be up with the Github repository in a few hours.

@BittAudio Miners and Validators on Subnet 16,

To address the VRAM spike and crash issues on the miner's side, we have decided to implement the following changes:

Removal of Wandb We have removed Wandb from the miner side to enhance performance, as some processes were periodically getting stuck.

Modification of the Auto-Update Feature We have introduced a new auto-update script, specifically designed to automate the update process for miners and validators. This script will handle the initiation and updating of miners/validators using pm2, ensuring that your systems are consistently running the most current version without requiring manual intervention. When running the miner/validator script, pm2 will spawn two processes: one for the update and one for the main process.

Miners and Validators, please do a manual git pull and restart your miner/validator. There is no need to specify the --auto_update paramether in the running command anymore. git pull pip install -e .

These changes are aimed at improving the stability and performance of mining operations on Subnet 16. The registrations will remain closed for an additional 24-48h in order to monitor the latest changes.

We appreciate your understanding and cooperation as we continue to enhance our network.

@BittAudio For any Subnet 16 Validator experiencing issues with setting weights:

We've noticed issues are arising due to a Subtensor error:: ERROR | Subtensor related error. WebSocketConnectionClosedException: Connection to remote host was lost.

✅ To resolve this, please:

• Ensure you're using local subtensor and are on the latest version of Subtensor -> https://github.com/opentensor/subtensor
• Ensure you're running the latest version of the SN 16 code -> https://github.com/UncleTensor/AudioSubnet
• Restart your Validator

Thank you!

@BittAudio

We're thrilled to announce the launch of version 3.0.0 for the Audio Generation Subnetwork! This update introduces significant enhancements and exciting new features designed to elevate your audio generation journey. Here's what you can expect in this latest release:

###Text to Music Integration (TTM)

• Explore New Horizons in Music Creation: Engage with our latest Text to Music capabilities using facebook/musicgen-medium and facebook/musicgen-large models. These models empower you to transform textual prompts into full-fledged musical compositions, opening up new realms of creativity. https://huggingface.co/facebook/musicgen-medium https://huggingface.co/facebook/musicgen-large

###Fine-tuned Model Support

• Harness the Power of Customization: Improve your Text-to-Speech, Voice Cloning, and Text-to-Music outputs by adding your own fine-tuned models. Simply specify the path to your custom models and unlock personalized audio generation experiences.

###Request Optimization

• Streamlined Efficiency: We've optimized our querying process to ensure Text to Speech and Voice Cloning now efficiently query 5 axons, while Text to Music queries have been expanded to query 10 axons, guaranteeing higher quality outputs.

✅ Required Actions:

• Miners: Please perform a manual git pull and configure your setup to specify a TTS model, a VC model, and a TTM model to continue mining on SN16. All the miners need to run all 3 services. Instructions: https://github.com/UncleTensor/AudioSubnet/blob/main/docs/miner.md
• Validators: You're required to perform a manual git pull and restart your validator to apply the new updates. Instructions: https://github.com/UncleTensor/AudioSubnet/blob/main/docs/validator.md

Additional Info: *Miners and validators need to strictly follow installation steps in miner.md and validator.md *For both Miners and Validators, a GPU benchmarks guide is available here to assist in preparation and optimization efforts. https://github.com/UncleTensor/AudioSubnet/blob/main/docs/benchmark.md *Please note, registrations will be paused for the next 24 hours, allowing miners sufficient time to adjust and update their setups/running commands accordingly.

🎶 This release marks a milestone in our journey to provide a rich, seamless audio generation experience. We can't wait for you to explore these new features and enhancements!

What's next? -> API integration in order to allow everyone to harness the power of SN 16 and build their own web and mobile apps on top of it.

@BittAudio HuggingFace experienced a brief downtime, causing Miners and Validators to be stuck. Please restart your Miners and Validators. Thank you!

@BittAudio ✅ The changes presented above have been pushed to main. Miners and Validators, make sure you're using the latest version of our code in order to be able to receive prompts/set weights on the network.

@BittAudio Validators & Miners,

Please ensure that you are running the latest version of our code, version 2.0.8.

If you're not on the latest version and haven't been using Wandb, you'll need to update your system manually. You can git pull the latest version from our repository. git pull python -m pip install -e .

Additionally, it's important to install Wandb on your machines. To do this, use the following commands:

Install Wandb: pip install wandb Log in to Wandb: Run wandb login and enter your Wandb API key. You can obtain your Wandb API key from the following link: https://wandb.ai/

After these steps, please restart your Validator/Miner. The auto-update feature will handle everything else.

**🚨 Only 10 out of 23 Validators have their code updated and are also using Wandb at the moment. This affects the performance of both Miners and Validators.

We strongly encourage you all, Miners and Validators, to keep your code updated and to use Wandb, as this will allow us to debug things easier and also check the subnetwork's performance.**

Thank you!

@BittAudio Validators, Miners,

Please update your code manually by downloading the latest version of our repository and restart your Validator. From now on the auto-update will take care of the rest.

git clone https://github.com/UncleTensor/AudioSubnet.git cd AudioSubnet git checkout main pip install -r requirements.txt python -m pip install -e .

Make sure to install Wandb, run the login command and enter your API key. You can get your Wandb API key from here: https://wandb.ai/ pip install wandb wandb login

@BittAudio Validators and Miners,

We have updated the frequency of voice prompts from validators and miners to every 1,000 steps, which is approximately a 500-second delay. As a result, the total delay for prompts is now around 20 minutes. Please note that this is a temporary measure, and we are working to resolve this issue on a permanent basis.

If you're not running the auto-update, please do a manual git pull and restart your miner/validator.

In addition, we are preparing to release a script designed for the voice clone deletion operation. This script will facilitate the efficient removal of any voices generated through our voice cloning process, allowing you to manage your voice clones as needed.

Furthermore, we have reduced the dendrite per query to 3 for voice cloning, which will also result in fewer prompts for voice cloning. It's important to note that on-demand prompts from validators will not experience any delay, as this is an on-demand service and is not as extensive as the automatic prompts.

Please be aware that these adjustments are temporary solutions. We are committed to resolving these issues as swiftly as possible.

Important notes:

• Each miner is required to obtain an ElevenLabs API key. Please note that one API key is needed for each hotkey/miner.
• We strongly recommend opting for either the "Growing Business" plan or the "Enterprise" plan.
• We are currently in negotiations with the ElevenLabs team to secure more edit/remove operations per month for Voice Cloning, as well as a discount on the Enterprise package. This discount would be available to all miners registered through our referral link: http://elevenlabs.io/?from=partnerpalmer2912

Thank you for your patience and ongoing support.

Bittensor Subnet 16 Team

@BittAudio Validators & Miners, the issues have been fixed.

✅ Please git pull again. Please consider running your scripts with --auto_update yes, to automatically run the latest changes as they are pushed to the main branch.

Thank you!

@BittAudio The latest changes to the Audio Subnet (SN16) have been pushed to the main branch.

✅ Updates:

• Miners will need to operate both Text-to-Speech and Voice Cloning simultaneously.
• Added Voice Cloning (VC) capabilities for miners using the ElevenLabs API.
• Added Text-To-Speech (TTS) capabilities for miners using the ElevenLabs API.
• Introduced VC and TTS services for Validators to easily query the miners using custom prompts/voices.

Validators and Miners, please execute a manual 'git pull' to run the latest code changes and adapt your running scripts accordingly.

git clone https://github.com/UncleTensor/AudioSubnet.git cd AudioSubnet git checkout main pip install -e .

➡️ Validators:

For Voice Clone Service: To run the validator, replace {wallet_name}, {hotkey_name}, and {huggingface_access_token} with your wallet and hotkey name and HuggingFace access token, respectively. For custom Voice Cloning, place your audio files (for example, audio01.wav) and corresponding text files (for example, audio01.txt) in the vc_source folder. Once the files are processed, they will be moved to the vc_processed folder. The voice cloned output will be saved in the vc_target folder.

For TTS Service: To use prompts from HuggingFace, ensure your HuggingFace access token is correctly set. For custom prompts, place a CSV file named tts_prompts.csv in the tts_source directory. The TTS service will process these prompts and generate corresponding audio files. The generated audio files from TTS will be saved in the tts_target directory.

The supported languages for TTS include: Korean, Dutch, Turkish, Swedish, Indonesian, Vietnamese, Filipino, Ukrainian, Greek, Czech, Finnish, Romanian, Danish, Bulgarian, Malay, Hungarian, Norwegian, Slovak, Croatian, Classic Arabic, Tamil, English, Polish, German, Spanish, French, Italian, Hindi, and Portuguese.

*Note: Some languages perform very good while others underperform.

python neurons/validator.py --netuid 16 --wallet.name {wallet_name} --wallet.hotkey {hotkey_name} --logging.debug --hub_key {huggingface_access_token}

🟨 **How to use custom prompts/voices for VS and TTS as a Validator: **https://www.youtube.com/watch?v=LGlK8P_o87o

➡️ Miners:

To run the miner, replace {wallet_name}, {hotkey_name}, {elevenlabs_api_key}, and {machine_port} with your wallet name, hotkey name, ElevenLabs API key, and the port your machine uses, respectively. The miner will use the ElevenLabs API for generating TTS and voice clone audios.

python neurons/miner.py --netuid 16 --wallet.name {wallet_name} --wallet.hotkey {hotkey_name} --logging.debug --clone_model elevenlabs/eleven --model elevenlabs/eleven --eleven_api {elevenlabs_api_key} --axon.port {machine_port}

*Note: Ensure the file paths, wallet details, and API keys are correctly set for seamless operation of the validators and miners.

☑️ You can acquire your ElevenLabs API key through the following links: For a referral link, visit: http://elevenlabs.io/?from=partnerpalmer2912 For a non-referral link, visit: http://elevenlabs.io/

Important notes:

• Each miner is required to obtain an ElevenLabs API key. Please note that one API key is needed for each hotkey.
• We strongly recommend opting for the "Growing Business" plan or the "Enterprise" plan.
• It takes around 2 minutes to create your account & get your API key.

@BittAudio Test version of AudioSubnet is now available in the new branch named "Audio_testnet". Details above 👆

@BittAudio is boosting this subnet, nice work @BittAudio

@BittAudio Hey, Miners and Validators, please join our testing phase on the Mainnet.

In order to test the Audio Subnet, the immunity period has been set to 15000 blocks (previously 5000).

Follow the GitHub instructions and start mining. https://github.com/UncleTensor/AudioSubnet

Installation git clone https://github.com/UncleTensor/AudioSubnet.git cd AudioSubnet pip install -r requirements.txt python -m pip install -e . Available models: microsoft/speecht5_tts facebook/mms-tts-eng suno/bark

*Please let us know if you encounter any issues with your miner or validator. *

Thank you!