pytorch

mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 12:54:11 +08:00

Files

rzou 889e3eeed3 Avoid cuda init to FakeTensorMode (#124413 )

Also partially fixes #122109

This PR:
- We add a C++ flag (only_lift_cpu_tensors) to toggle the
  torch.tensor(1, device='cuda') ctor strategy.
  When false (default), it does the current PyTorch behavior
  of unconditionally constructing a concrete CUDA tensor then calling
  lift_fresh on it. When true, we instead construct a concrete CPU
  tensor, call lift_fresh, and then call Tensor.to(device) (under any ambient
  modes).
- FakeTensorMode flips this flag depending on if CUDA is available or
  not. We don't unconditionally set the flag to True because that is
  likely BC-breaking.

Test Plan:
- existing tests
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124413
Approved by: https://github.com/eellison

2024-04-19 02:39:35 +00:00

__init__.py

Add Fake Cross Ref Mode, migrate sparse to it (#85382 )

2022-09-21 17:15:47 +00:00

fake_impls.py

Add fake impl for aten.unique2 (#124306 )

2024-04-17 22:55:27 +00:00

fake_tensor.py

Avoid cuda init to FakeTensorMode (#124413 )

2024-04-19 02:39:35 +00:00

fake_utils.py

Revert "[fake_impls] Fix seed/offset device for attention kernels (#120839 )" (#121447 )