* Dynex Machine Learning Update *.
12 Aug 2023, 06:34
* Dynex Machine Learning Update *
We would like to share an update on the progress since we upgraded all miners with Ising/QUBO sampling capabilities to also support the a to job type "Machine Learning":
- So far 3,382 ML-jobs have been computed the Dynex main-net, ranging up to 25M concurrent chips
- Around 180k GPUs with Ising/QUBO sampling capabilities are being active in the moment
- A Dynex Job Board ("Dynex PoUW Dashboard") has been published where users can follow computing jobs in real-time; accessible for example at and also published in our Git
- We published the first version of our Dynex SDK Wiki for our testers and the community in general to outline the concepts, models and sampling on the Dynex Platform:
- The Dynex SDK has been provided to our external security expert @y3ti | deepminerz.com who has been assisting us with phase I of the rollout, which includes functionality and scalability testing; network security and stability; power draw/utilisation of GPUs; pool/mallob reliability and functionality during large scale tests and more. This is an important phase to ensure the quality and reliability of our Dynex SDK and the platform in general. We can report that the testing exceeded our expectations and only minor required fixes have been identified so far.
- Initially planned for later, we decided to incorporate "Parallel Jobs" already in the initial release of the Dynex SDK. This enables Python programmers to push multiple jobs to be run on parallel on the Dynex Platform, which allows realisation of concepts like Federated Machine Learning for example; as our computing platform is decentralised in terms of participating workers, this feature is a natural evolvement
- Phase II, which includes the rollout of the Dynex SDK to the first batch of beta testers will be starting in the next few days and we will provide further updates as they progress.