AMD has introduced a new feature in its AI Workbench that enables users to fine-tune large language models (LLMs) for specific use cases without writing any code. The platform streamlines the process of adapting pre-trained models to meet particular requirements, making it accessible to a broader audience. This tool is part of AMD's broader initiative to simplify AI development and deployment.
The fine-tuning process involves uploading a dataset, selecting a model, configuring training parameters, and running the training job. Once completed, the model can be deployed using AMD Inference Microservices (AIMs), which provide standardized, portable inference microservices for serving AI models. AIMs abstract away the complexities of model serving by offering an intelligent orchestration layer that automatically configures runtime environments and selects an optimized performance profile.
The blog post outlines a real-world example where a model is fine-tuned to answer questions about the AMD enterprise AI reference stack and to refuse discussing unrelated topics. This demonstrates how fine-tuning can be used to specialize models for specific tasks, showcasing the flexibility and utility of the AMD AI Workbench.
Source: amd
