Hugging Face, the company behind the popular open-source library for natural language processing, has partnered with AMD to accelerate state-of-the-art models. The partnership will allow developers to train and deploy large language models (LLMs) on AMD hardware, which will significantly improve performance and reduce costs.
“AMD and Hugging Face work together to deliver state-of-the-art transformer performance on AMD CPUs and GPUs. This partnership is excellent news for the Hugging Face community at large, which will soon benefit from the latest AMD platforms for training and inference. This new partnership will do more than match the competition and help alleviate market dynamics: it should also set new cost-performance standards.”
Hugging Face's library is used by millions of developers around the world to train and deploy LLMs. LLMs are a type of artificial intelligence model that can be used for a variety of tasks, including text generation, translation, and question answering. However, LLMs can be computationally expensive to train and deploy.
AMD's hardware is designed to accelerate the training and inference of LLMs. The partnership between Hugging Face and AMD will allow developers to take advantage of AMD's hardware to train and deploy LLMs more quickly and efficiently.
“On the GPU side, AMD and Hugging Face will first collaborate on the enterprise-grade Instinct MI2xx and MI3xx families, then on the customer-grade Radeon Navi3x family. In initial testing, AMD recently reported that the MI250 trains BERT-Large 1.2x faster and GPT2-Large 1.4x faster than its direct competitor.”
How the Partnership Aims to Accelerate AI Development
One of the key outcomes of the collaboration is the integration of Hugging Face Transformers, the most popular library for natural language processing, with AMD ROCm, the open-source software platform for accelerated computing. This integration will allow users to seamlessly train and run Hugging Face Transformers models on AMD GPUs and CPUs, with optimized performance and scalability.
Another outcome of the collaboration is the launch of Hugging Face Accelerated Inference API, a new service that allows users to run inference on large-scale natural language models using AMD hardware. The service will support models such as GPT-4, as well as custom models trained with Hugging Face Transformers.
Microsoft's Azure Machine Learning Foundation Models from Hugging Face
Microsoft and open source natural language processing (NLP) platform Hugging Face has been in partnership since last year. That initial collaboration focused on building Hugging Face Endpoints – a machine learning inference service that is underpinned by Azure ML Managed Endpoint. Last month, Microsoft announced that Azure Machine Learning now has Hugging Face foundation models.
At its Azure Open Source Day, Microsoft announced the new collaboration, which introduces a public preview of foundation models to Azure Machine Learning. The platform's customers can now create their own open-source foundation models and then scale them.
Users can use foundation models from various open source repositories (Hugging Face is one example) to help enhance their foundation models and then deploy them. Features include being able to use a pre-trained model for inference and deployment, as well as machine learning-supported tasks from customer data.