Software Choices for a Multi-GPU AI Workstation

When I set out to build my local AI Workstation I did so with the expectation that I wanted something that could dabble at every level of the current AI landscape.  While I am not using it for production coding efforts I do enjoy reading through the code and making changes to suit my needed or even fix bugs that happen to crop up.

For instance, when working with Augment Toolkit I found a couple issues that were causing the pipeline to not generate any data.  I was able to track the issue back to a simple “copy paste” problem as the author was building out the various pipelines.  A very common thing to overlook especially considering that the author admitted to using cloud based services for running Augment Toolkit and the subsequent training run.  Unless you scour over the logs you will completely miss it.

In addition to code, I rather enjoy general content creation tasks and gave the example about the three strawberries and then tweak the prompt to change the story.  After a couple rounds of text have been generated I can enable the ComfyUI image tool and generate a title image for the story.  Overall, I have still only scratched the surface on what this AI Workstation can do including the ability to Fine Tune Train an LLM, which I will cover in the next article.