windows10+16G(GPU)+32g

torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 29.76 GiB. GPU 0 has a total capacity of 16.00 GiB of which 3.88 GiB is free. Of the allocated memory 10.72 GiB is allocated by PyTorch, and 407.65 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)

Start with the reported torch.OutOfMemoryError and the linked PyTorch CUDA Memory Management documentation, comparing the requested allocation with the GPU capacity and free memory shown. No repository file or test is named; completion criteria are not specified in the issue, so the first step is to establish the intended fix and a reproducible validation.

Written by the indexing model from the issue text.