yes, I did. using Intel extension for pytorch instead PyTorch 2.5 (I know it has supported but it's not stable and library like torchvision still build manually for windows). btw I'm using Intel Arc A530M (mobile version), I'm just using for light weight training task for instance classification (CNN) by using ShuffleNet V2 architecture. works great for me (I don't have another gpu for be compared)

Give this GPU a try if you're not in a hurry. There is still a lack of community support outside of PyTorch.