Motivaiton
Akash is the first decentralized network with GPUs capable of general-purpose computing.
The absence of foundation models trained on decentralized networks such as Akash presents a significant challenge in establishing user confidence and adoption.
Training a foundation model on Akash establishes trust and provides valuable information to the Akash core team about various issues, if any exist.
Summary
We’re proposing (and seeking community funds for) training a foundation AI model on Akash Network, resulting in an “Akash” named open-source AI model, archived/shared on Huggingface. The full details can be found in the original Github discussion thread here (https://github.com/orgs/akash-network/discussions/300).
Benefits to the Akash community include:
- Demonstrating that AI model training can be conducted on Akash’s decentralized cloud platform.
- Attracting AI and ML developers (web3 and web2) to the platform.
- Positioning GPU providers for success, following the mainnet launch (by generating demand).
- Rounding out AI use cases by adding model training (Inference and fine-tuning have already been demonstrated in the testnet).
- Strengthening Akash Networks’ open-source commitment by building and contributing an open-source AI model to the broader OSS community.
This is a funding proposal asking for $48,000, which is 54,055 AKT (at the 30-day moving average price as of October 5, 2023, of 0.88798). Since the duration of the incentives in this initial experiment is long (3-6 months), using an average price mitigates some of the volatility.
Any unspent funds at the end of the experiment will be returned to the community pool OR used towards future tenant/provider incentives ONLY. In the event of a shortfall, OCL will cover the difference to avoid tapping into the community pool again. The wallet that holds the funds will be akash1ajwyre772xyu8m94j7prtvz57w020aqnpf62lx
Posts
Copyright
All content herein is licensed under Apache 2.0.