Multi-GPU support ?

Hi,

I have 11 GPU in my server but the script don't spread across gpu, which end with an out of memory error.

Any multi-gpu support in the future ?

The issue names no files, tests, or entry points. Start by locating the script's GPU selection and model-loading paths, then reproduce the out-of-memory failure on the reported multi-GPU setup. Done means the workload is distributed across available GPUs without the reported memory failure and the behavior is covered by an appropriate test or documented usage.

Written by the indexing model from the issue text.