NVIDIA Tesla T4 torch.OutOfMemoryError: CUDA out of memory




system:windows server2022

**File "C:\Ai\framepack_cu126_torch26\webui\diffusers_helper\models\hunyuan_video_packed.py", line 122, in attn_varlen_func
x = torch.nn.functional.scaled_dot_product_attention(q.transpose(1, 2), k.transpose(1, 2), v.transpose(1, 2)).transpose(1, 2)
torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 28.87 GiB. GPU 0 has a total capacity of 15.83 GiB of which 4.00 GiB is free. Of the allocated memory 11.31 GiB is allocated by PyTorch, and 392.97 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 at diffusers_helper/models/hunyuan_video_packed.py line 122 and review the reported PyTorch CUDA out-of-memory traceback alongside the linked CUDA memory-management documentation. The issue is done when a reproducible resolution is identified and verified on the reported Windows Server 2022 and Tesla T4 setup.

Written by the indexing model from the issue text.