Hardware optimization for computer vision
Hello, I’m currently developing a computer vision model. Due to the need for computational resources, I’m here. Honestly, I don’t have much knowledge about hardware, and I’d like to know if you have any documentation or material that could help me make the most out of the tool
1 Reply
A very rough estimate is to base it on your model size and make an approximation of the VRAM needed:
For inference: 1B parameters ≈ 2GB VRAM (FP16)
For training: 1B parameters ≈ 6-12GB VRAM