GPTQ - Move quantized_model to CUDA device

[samuel100]

Describe your changes

When using GPTQ the quantized_model must be moved to the CUDA device to avoid the "Expected all tensors to be on the same device" error in auto-gptq. See AutoGPTQ/AutoGPTQ#729

Checklist before requesting a review

• Add unit tests for this change.
• Make sure all tests can pass.
• Update documents if necessary.
• Lint and apply fixes to your code by running lintrunner -a
• Is this a user-facing change? If yes, give a description of this change to be included in the release notes.
• Is this PR including examples changes? If yes, please remember to update example documentation in a follow-up PR.

[xiaoyu-work]

According to the discussion thread, it seems this was already fixed by AutoGPTQ/AutoGPTQ#607?

[jambayk]

Like @xiaoyu-work mentioned, this issue should be fixed if you install autogptq from source https://github.com/AutoGPTQ/AutoGPTQ?tab=readme-ov-file#install-from-source. Could you try it to see if it works?

[samuel100]

@xiaoyu-work @jambayk -- I tested building from source and confirm that it fixed the issue. I'll close this PR and create a new one that will update the documentation with instructions.