⛔[SD3 branch] Huber+SNR RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!

Pytorch 2.7.0, CUDA 12.8

What: RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! when using Huber + SNR

Where: train_util.py, line 6025

Fix:

From
`alphas_cumprod = torch.index_select(noise_scheduler.alphas_cumprod, 0, timesteps.cpu())
`
To
`alphas_cumprod = torch.index_select(noise_scheduler.alphas_cumprod, 0, timesteps.to(noise_scheduler.alphas_cumprod.device))
`

Start at train_util.py line 6025 and reproduce the Huber + SNR path with PyTorch 2.7.0 and CUDA 12.8. Check the tensor devices used by the alphas_cumprod lookup, then verify that the RuntimeError no longer occurs when the lookup indices and source tensor use the same device.

Written by the indexing model from the issue text.