Using CUDA in Colab

Question

Is it necessary to convert tensors and models to CUDA with tensor.to in colab when I've chosen runtime type as GPU?

I want use CUDA for training my model

user15813867 · Accepted Answer · 2021-09-08 10:54:29Z

1

tensor.to(device) transfer data to the given device.
Yes, you need to transfer model, input, labels etc to whatever device you are intending to use

answered Sep 8, 2021 at 10:54

user15813867

Sign up to request clarification or add additional context in comments.

1 Comment

Do note: torch.Tensor.to makes a copy while torch.nn.Module.to is an in-place operation.