This paper discusses a simple yet powerful distributed fine-tuning topology using Dell servers and a Z9664F-ON switch in a distributed fine-tuning scenario.
- Three heterogeneous Dell PowerEdge servers with Broadcom NetXtreme BCM57508 NICs were used. Performance measurements were taken to show the differences between a three-node cluster compared with a single system.
- Testing shows high performance, scalability, and reliability using the setup in this paper.
- Scalers AI used Llama 2 from Meta AI to demonstrate how to create a chatbot for medical services.
- A developer template was generated to quickly help customers get up and running with distributed fine-tuning of GenAI.