🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/120076

• 📄 Preview Python docs built from this PR
• 📄 Preview C++ docs built from this PR
• ❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure

As of commit 231c64d with merge base 52e9049 ():

This comment was automatically generated by Dr. CI and updates every 15 minutes.

This is still not an optimal type for Generator, because if you have a Generator and you need to create a const optional<Generator>&, you still need to do a copy assignment into the optional (because the optional is designated as owning the Generator it has internally).

We have an analogous problem with optional tensors, and ultimately we had to do a trick with OptionalTensorRef to get the efficient version. If we're going through all the trouble of changing the API here, we probably should do this here too.

@pytorchbot revert -m "Reverting in order to check if this will fix XLA trunk jobs" -c ignoredsignal

@pytorchbot successfully started a revert job. Check the current status here.
Questions? Feedback? Please reach out to the PyTorch DevX Team

@cyyever your PR has been successfully reverted.

@pytorchmergebot merge -i

Merge started

Your change will be merged while ignoring the following 2 checks: linux-binary-conda / conda-py3_10-cuda11_8-build / build, trunk / win-vs2019-cpu-py3 / test (default, 1, 3, windows.4xlarge.nonephemeral)

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

HI This fails internally with error:

stderr:
terminate called after throwing an instance of 'c10::Error'
  what():  
Mismatch in kernel C++ signatures
  operator: aten::random_.from(Tensor(a!) self, int from, int? to, *, Generator? generator=None) -> Tensor(a!)
    registered at buck-out/v2/gen/fbcode/45fe2882ccb6f7da/caffe2/__gen_aten__/out/RegisterSchema.cpp:6
  kernel 1: at::Tensor& (at::Tensor&, long, std::optional<long>, std::optional<at::Generator> const&)
    dispatch key: VmapMode
    registered at fbcode/caffe2/aten/src/ATen/VmapModeRegistrations.cpp:37
  kernel 2: at::Tensor& (at::Tensor&, long, std::optional<long>, std::optional<at::Generator>)
    dispatch key: CustomRNGKeyId
    registered at fbcode/pytorch/csprng/torchcsprng/csrc/csprng.cpp:344

Exception raised from registerKernel at fbcode/caffe2/aten/src/ATen/core/dispatch/OperatorEntry.cpp:130 (most recent call first):
# 0  c10::get_backtrace[abi:cxx11](unsigned long, unsigned long, bool)
# 1  0x00000000246579dc
# 2  c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >)
# 3  c10::detail::torchCheckFail(char const*, char const*, unsigned int, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)
# 4  c10::impl::OperatorEntry::registerKernel(c10::Dispatcher const&, std::optional<c10::DispatchKey>, c10::KernelFunction, std::optional<c10::impl::CppSignature>, std::unique_ptr<c10::FunctionSchema, std::default_delete<c10::FunctionSchema> >, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >)
# 5  c10::Dispatcher::registerImpl(c10::OperatorName, std::optional<c10::DispatchKey>, c10::KernelFunction, std::optional<c10::impl::CppSignature>, std::unique_ptr<c10::FunctionSchema, std::default_delete<c10::FunctionSchema> >, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >)
# 6  torch::Library::_impl(char const*, torch::CppFunction&&, torch::_RegisterOrVerify) &
# 7  0x000000001fcae596
# 8  _ZN5torch6detail16TorchLibraryInitC2ENS_7Library4KindEPFvRS2_EPKcSt8optionalIN3c1011DispatchKeyEES8_j_torchcsprng$_C
# 9  0x000000001fcc4738
# 10 __libc_start_main_alias_2
# 11 _start


exit_code:"Process terminated by the signal: 6"

Pointing to this csprng : https://github.com/pytorch/csprng/blob/main/torchcsprng/csrc/csprng.cpp#L344

@pytorchbot revert -m "breaking internal builds" -c ghfirst

@pytorchbot successfully started a revert job. Check the current status here.
Questions? Feedback? Please reach out to the PyTorch DevX Team

@cyyever your PR has been successfully reverted.

@alanwaketan @ezyang https://github.com/pytorch/csprng should be updated for this PR, how can we do that because we didn't have a coordination protocol like PyTorch/XLA?

@cyyever I guess I will need to revert your change in PyTorch/XLA for now?

@cyyever At this point, you might have to just take the long way around and add BC logic that can handle the old overload types. Fortunately, it should be possible to detect them from C++ metaprogramming and then you can transparently shim them to the new type.

If you still want to try landing directly again, we will need to make this an fbcode first diff, as the csprng will need to be landed simultaneously to fbcode. That means someone at Meta has to volunteer to shepherd the rest of the patch.

Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as Stale.
Feel free to remove the Stale label if you feel this was a mistake.
If you are unable to remove the Stale label please contact a maintainer in order to do so.
If you want the bot to never mark this PR stale again, add the no-stale label.
Stale pull requests will automatically be closed after 30 days of inactivity.