Thanks to ChatGPT Plus, the paid version of the most popular chatbot in the world, we can create our own AI with the information we want.
A few weeks ago NVIDIA presented Chat with RTX, an application that allows us to create a personalized chatbot based on an LLM that we can train with the data we want and that allows us to create personalized AI, provided that we have an RTX 30 or higher.
If we have an RTX 20 graphics card we can forget about it, but what about AMD? If we have an AMD graphics card, this manufacturer does not offer us an application like NVIDIA to train a personalized AI, but it shows us the steps to follow to create it.
If we have an APU Ryzen 7000 and Ryzen 8000, the models include an NPU or a graphics card from the Radeon 7000 series, we have the possibility of creating an AI and interacting with it through a chatbot, in the same way as we let’s do it with ChatGPT, Microsoft Copilot, Google Gemini or any other.
How to create a custom chatbot with an AMD graphics card or APU
AMD has published a guide on its website with all the steps to follow to create a GPT-based LLM-powered chatbot, using the compatible APUs and graphics cards we mentioned above.
If the equipment has a Ryzen APUs 7000 series or 8000 series, you must download the LM Studio – Windows application here. Yes, you have one Radeon Graphics of the RX 7000 series, the application to download is LM Studio – ROCm from here, a version which is still in the beta phase.
Next we run the version that we downloaded and we must select the type of LLM that we want to use, just like with Chat with RTX. The LLMs available are Mistral 7b and LAMA V2 7b, both give us very similar results so unless you need to process a lot of data, in principle it doesn’t matter which model you choose.
In the right panel, we select, we search for the file corresponding to the LLM that we have selected and we download the Q4_K_M model and we click on the Download button.
Once we have downloaded it, we go to the chat window and select the LLM model that we have selected and downloaded so that it can finish loading into the application.
If the computer is run by an AMD APU, we can now start chatting with the chatbot, once we have added the content we want to train the AI with.
If instead of having an APU you have a Radeon RX 7000, you will need to do some additional settings to start chatting. To do this, you must display the settings box and check the GPU Offload boxes and move the scroll bar to the right.
After making this change, the detected GPU type should display AMD ROCm. Otherwise, either our chart does not belong to the RX 7000 series or we did not perform the steps correctly.
