DeepSeek v4 (Flash) running on DGX Spark

Описание: This is a real-time video showing the performance you can get when you run the DeepSeek V4 (Flash) model on a Nvidia DGX Spark.

Here I have connected it to my Xcode project running on a Mac, called "Solve it now!" which is published in the App Store. It is a medium level complexity project.

The server software is a fork off the amazing work of https://github.com/antirez/ds4 called

https://github.com/faisalmemon/ds4/

Use branch faisalm/dockerised

The purpose of my fork is to demonstrate how to compile the source code within a Docker container. This is the normal workflow for a AI server, despite the DGX Spark having a desktop form factor.