2024-04-14

Announcing Stronk AI

 Announcing Stronk AI

As promised when receiving the `treasury rewards for developing orchestrators`, we've staked a significant amount of LPT with two small Orchestrators: 500 LPT at Bemanode and 2000 LPT at Epitonium

Another goal we had in mind was to get ready for AI inference jobs. So as of last week, we have the `Stronk-Gateway` loaded up with some ETH reserve, which has already sent out a couple of winning tickets to AI workers!

Of course we're also participating in the AI pipeline with our own prosumer-grade hardware:
MOTHERBOARD 1 x Asrock TRX50 WS
RAM 4 x Kingston FURY Renegade Pro / 32GB / DDR5 / 6000 MHz
HDD 2 x Seagate Exos X18 16TB
PSU 1 x Corsair AX1600i 1600W
CPU 1 x AMD Ryzen Threadripper 7970X CPU
CPU cooler 1 x Noctua NH-D9 TR5-SP6 CPU COOLER
SSD 1 x Samsung 990 PRO 4TB M.2 SSD
Case 1 x Corsair 7000D Airflow
GPU 2 x MSI GeForce RTX 4090 SUPRIM LIQUID X 24G

Community Contributions

But wait, there's more! The new AI pipeline is exciting stuff. While the core AI team has been working diligently on enabling Orchestrator discovery and selection, we went ahead and did some experimentation with the fun part: trying out new models and finetuning parameters. All of the changes we made in our fork are being upstreamed to enhance the AI pipeline for everyone.

* Added new `img2img` model `pix2pix` in this pull request
* Added new `img2vid` model `i2vgen` in this pull request.
* Added new `txt2vid` pipeline with models `ali-vilab` and `AnimateDiff-Lightning`. This is split up in one, two, three pull requests.
* Added a bunch of `txt2img` base models, which are still being evaluated before upstreaming.

Announcing your AI playground

In order to test all these changes to the core AI pipeline, we've created a simple dashboard which allows you to test our forked version of the AI pipleine even before our changes get merged upstream. Now with a mobile-friendly layout, visit https://inference.stronk.rocks/ to start crafting your own images and video's!