Submit a Contribution

The commitment to open-source principles in developing character models offers a broad avenue for innovation and contribution. Here's a detailed approach to how developers and contributors from diverse backgrounds can participate and how their contributions can be integrated and recognized.

Join our Discord community to ignite your creativity and collaborate with like-minded individuals. Engage in brainstorming sessions, foster discussions about your contributions, and seize the opportunity to form dynamic teams for even greater impact!

Who can contribute?

Anyone who is interested in improving the model and able to meet the submission requirements is welcome to participate.

Would you love to contribute but are unsure where to start? Refer to our beginner's guide to learn about Model Fine-tuning and Dataset Curation!

Get Started

Step 1 Set your challenge

To get started, you need to set your challenge. There are a few ways to participate:

1. Create a dataset for LLM models

2. Create a voice data

3. Finetune a LLM

4. Finetune a Voice Model

If can be a combination of both, below are the examples of contributions

• Andrew created a dataset for LLM (1) and finetune a LLM with the dataset (3)

• Serena created a voice data (2) and finetune a voice model with the data (4)

• Liam downloaded a dataset and finetune a LLM with the dataset (3)

• Jess dowlonaded a voice data and finetune a voice model with the data (4)

Step 2 Select a character for your challenge.

1. Head over to app.virtuals.io/contribution

2. Choose one Virtual that you are interested in contributing.

3. Below is an example of a Virtual. In each Card, there display the model used by each Core. For example, Cognitive Core is using siliconmaid-7b-seraphinal; Voice Core is using kq-seraphina-tts.Depending on your challenge in Step 1. Take a look a tthe respective Core.

4. If you intend to do (3) or (4) in Step 1, you may download the model file under the File section after you click into each Core. Note that it is an ipfs storage. You have to rename the file to the right extension format to open the file successfully. Alternatively, you may use any other model found.
5. If you do not intend to do (1) or (2), click into Contribution Tab. You may find the contribution list and download any dataset with status "Finetune Required" to finetune.

Step 3 Start working on your challenge

Prepare your model or data independently. Ensure they are tested and ready for submission.

We also have a few guides curated by the community to help you get started!

Step 4 Submit your work

To submit your work, you will need to login/ sign up with Protocol App. Submission of Contribution is an off-chain process that does not require transaction signing. However, a wallet is required to register the contribution on-chain and for Protocol to airdrop the NFT. It can be done by registering an account at Protocol App.

Before you submit please make sure that

• Your folder and files required as per explained in the Contribution section.

• Please make sure you adhere to the guidelines specifically to each Core.

Cognitive Core

pageContribute to Cognitive Core

Voice Core

pageContribute to Voice Core

Visual Core

pageVisual Core

• Ensure you identify how to contribute. Below are the examples.

• Andrew created a dataset for LLM (1) and finetune a LLM with the dataset (3). Andrew will submit two contributions for Cognitive Core: One for dataset, another one for Model.

• Serena created a voice data (2) and finetune a voice model with the data (4). Serena will submit two contribution for Voice Core: One for Voice Data, another one for Model.

• Liam downloaded a dataset and finetune a LLM with the dataset (3) Liam will submit one contribution for Cognitive Core and will label the dataset used in his submission.

• Jess dowlonaded a voice data and finetune a voice model with the data (4) Jess will submit one contribution for Voice Core and will label the dataset used in his submission.

• Ensure contributions align with the objectives and scope of the platform. Irrelevant submissions will be rejected.

• Provide detailed information about your submission, as it forms the basis of the voting proposal for Validators.

• Avoid plagiarism; it will lead to automatic rejection.

• Credit any external sources or references used.

Refer to the link for the step-by-step guide on submitting a contribution.

What should you be expecting post Submission?

Enhance your experience by adding our Protocol App to your home screen and enabling notifications. Stay up-to-date on your contributions and reward distributions, ensuring you never miss a beat in your journey with us! Alternatively, join our Discord community for the updates.

Contributors can track their submission status in the Protocol App, with statuses including:

Draft: When the submission is in draft.

Orchestrating: A playground is in creation for Validators to interact with your model.

Processing: Submission is submitted and off-chain agent will validate the submission to ensure it fits the requirements such as

• formatting of the file

• output of the file is correct

• there is no similar submission

Voting: NFT is created and airdropped to the wallet registered. Voting proposal is made available for Voting. At this stage, user will be able to see a Contribution NFT being minted to their wallet and a Voting Proposal.

Approved: The submission has been accepted

Available: The submission is made available on marketplace for usage and contributions

Rejected: The submission has been rejected, no change is done to existing ecosystem

Only applicable to dataset

Finetune Required: Dataset is being accepted and now model finetuning is required to use the dataset.

Merging: A model trained with the dataset has been submitted for voting.

Merged: Dataset has been used to finetune the model and the model has been accepted.

Get Rewarded

Successful contributions will be rewarded based on their daily utilization and the impact of the contribution.

pageContributors get revenue based on impact scores and contribution citations