Speedup patch embedding by moving coord creation on device

[Tom-Julux]

While experimenting with image segmentation and using the prun function on predictor.set_image, I discovered a significant performance issue in the get_rel_pos function.

The problem arises from the lack of device assignment for the tensor containing relative positions, causing unnecessary data transfers when using accelerators like GPUs.

The solution is rather simple: create the coordinate tensor on the same device as the underlying data.
In my specific case (vit-b, P100 GPU, 128x128 image size), this change reduced the get_rel_pos execution time from 197ms to 3ms, cutting down the predictor.set_image runtime from 220ms to 20ms.

Given that there's no reason to keep the coordinate tensor creation outside the device, I suggest incorporating this change into the official release for the benefit of all users.

[facebook-github-bot]

Hi @Tom-Julux!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at cla@meta.com. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!