How to reduce GPU usage ?

GPU 1050Ti with 4G memory, got this error.

```
RuntimeError: CUDA out of memory. Tried to allocate 768.00 MiB (GPU 0; 4.00 GiB total capacity; 2.71 GiB already allocated; 0 bytes free; 2.81 GiB reserved in total by PyTorch)
```

The report names no file, test, or entry point, only a PyTorch CUDA out-of-memory error on a 4 GB GPU. Start by locating the inference example or notebook that produces the allocation, then establish a reproducible lower-memory configuration and verify that inference completes without the reported error.

Written by the indexing model from the issue text.