CPU or GPU

[sahulsumra]

is this code is going to use GPU or CPU because currently its using CPU how to make it run on GPU

1.I am using OLLMA to test the accuracy but its taking lots of time for the output...

"""
import vision_agent as va
agent = va.agent.OllamaVisionAgentCoder()
agent("Count the people in the image", media="image (4).png")

"""
I am using above code to check the accuracy ....

2. What if need to test this model on some drawing images it going to work and how its will work .....do we need to train the model give some ideas about it .....Library looks very intresting

[dillonalaird]

1. Ollama should support GPU and CPU but you will have to check the Ollama documentation to get that working https://github.com/ollama/ollama/tree/main CPU will run a lot slower. I think I just installed it on a GPU machine and it automatically used the GPU. There are also some issues with performance using Ollama I discuss on the Discord https://discord.gg/RVcW3j9RgR

2. You can see all the tools and models the agent has access to here https://github.com/landing-ai/vision-agent/blob/main/vision_agent/tools/tools.py#L1948 so it largely depends on how well those models are able to accomplish your task. We also offer the ability to fine-tune the florence2 model, which can be easily done through the UI at https://va.landing.ai it can also be done through the repo but you have to convert your bounding boxes to a string format so it's more difficult