Want to use it for learning about autonomous driving and using of LLMs in robotics.
Unfortunately, it is not possible to purchase the Picar including a Raspberry, so I will need to buy a Raspberry seperately. I found an online shop with all sorts of them and in a video I saw, the RPI 4B is recommended. There are versions with 2 GB, 4 GB or 8 GB of RAM. Are the larger versions worth it for the Picar? What model do you recommend? Anything special I need to consider so that the Picar works without problems?
Just my own personal opinion. Im sure you’ll get conflicting advice though.
The picar will run perfectly fine on any rpi model. However, your requirement of LLM could, depending upon which LLM models you plan to use, will be what is really dictating your requirements, even as much as an 8gb rpi5 or, maybe even a 16gb. Conversely small models can be run on a 4gb rpi4.
Many Llm specify their performance and memory requirements on their requirements page. Just take a look for your intended models.
Hi, thanks for the reply. I guess it is more about learning how to use a LLM on RPI in combination with the picar. It would be ok if at least a small model would work. For really good results I agree that I cant expect either of the RAM options to be sufficient. For this I guess the only way is to connect to the internet and use the API of one of the big tech companies. But if the 4 gb version could run a small model for testing purposes that would be ok with me.
However, it would be important for me that the autonomous driving features and obstacle recognition works smoothly.
Exactly. I learned a lot with small models on a 4gb rpi4, which is also enough for basic object recognition and driving around. For me, I could not justify the cost of hardware for using a larger LLM just for personal learning. Im sure you’ve seen Sunfounders table for onboard Ollama as a typical example…
