pytorch

dcac4dd5 - Add int32_t range check in packed_accessor32 in PyTorch TensorBase (#86085)

Add int32_t range check in packed_accessor32 in PyTorch TensorBase (#86085) Summary: As ajtulloch suggested, we can make tensor.packed_accessor32<...>() raise an exception if tensor.numel() > std::numeric_limits<uint32_t>::max(). Trade-off: run-time check overhead (one-time) when doing `packed_accessor32` accessor. Differential Revision: D39996275 Pull Request resolved: https://github.com/pytorch/pytorch/pull/86085 Approved by: https://github.com/ngimel