transformers

cdea840f - [Qwen2.5-Omni] Update stale expected values for cuda sm_86 (#48164)

[Qwen2.5-Omni] Update stale expected values for cuda sm_86 (#48164) The Nvidia CI docker was updated to torch 2.10 in commit c0fe6164d0, which changed model numerics slightly. Update ("cuda", (8, 6)) and ("cuda", 8) expected values in test_small_model_integration_test and test_small_model_integration_test_batch, verified on A10G CUDA 8.6. Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>