Here’s the translation into American English:
In a move that reinforces its commitment to artificial intelligence (AI), AMD has recently released a detailed guide for running DeepSeek R1 Distilled Reasoning models on AMD Ryzen™ AI processors and Radeon™ graphics cards. These models, regarded as cutting-edge in the field of large language models (LLMs), are designed to tackle complex tasks through chain-of-thought reasoning (CoT), a technique that enables AI to “think” before providing a definitive answer.
DeepSeek R1: An Advanced Reasoning Model
DeepSeek R1 is one of the most advanced reasoning models currently available. What sets it apart from traditional LLMs is its ability to analyze problems from multiple angles before generating a final response. Unlike traditional models that provide immediate answers, CoT models, like DeepSeek R1, conduct an internal reflection process that may involve thousands of tokens before arriving at a conclusion. This approach makes them ideal tools for complex tasks such as solving mathematical or scientific problems.
The breakthrough is that AMD has simplified the implementation of these models on its Ryzen AI processors and Radeon graphics cards, making it accessible even for non-expert users. This advancement is available through LM Studio, a platform that allows for easy downloading and running of these models.
How to Run DeepSeek R1 on AMD Hardware
To get the DeepSeek R1 models running on AMD hardware, users need to follow a few simple steps. First, ensure that Adrenalin driver version 25.1.1 or later is installed. Next, download and install LM Studio, a tool that facilitates AI model management. Once the installation is complete, users can select the DeepSeek R1 model that best fits their hardware and download it in Q4 K M format, a configuration recommended by AMD for optimal performance.
For Ryzen AI processors, the largest compatible model is the DeepSeek-R1-Distill-Llama-70B, although it is only available on devices with 64 GB or 128 GB of memory. For Radeon graphics cards, the RX 7900 XTX can run the DeepSeek-R1-Distill-Qwen-32B, while more modest models like the RX 7600 support the DeepSeek-R1-Distill-Llama-8B.
Innovation and Versatility
The implementation of these models on AMD hardware not only democratizes access to advanced AI but also opens new possibilities in fields such as education, research, and software development. The ability to observe the AI’s “thought process” before obtaining a final answer adds a layer of transparency and accuracy that is invaluable in critical tasks.
With this initiative, AMD reaffirms its commitment to innovation in AI by offering its users cutting-edge tools that enhance the performance of their devices. For those interested in exploring these capabilities, AMD invites them to try the DeepSeek R1 models on their systems, ensuring a smooth and efficient user experience.
via: AMD Blogs