Windows CI intermittent error: C2993: 'Derived': illegal type for non-type template parameter '__formal #25393

ezyang · 2019-08-29T14:25:06Z

Kind of similar to #25389

Sometimes our Eigen build fails this way:

C:\Jenkins\workspace\caffe2-builds\py2-cuda9.0-cudnn7-windows-build\third_party\eigen\Eigen\src/LU/InverseImpl.h(366): error C2993: 'Derived': illegal type for non-type template parameter '__formal'
C:\Jenkins\workspace\caffe2-builds\py2-cuda9.0-cudnn7-windows-build\third_party\eigen\Eigen\src/LU/InverseImpl.h(366): error C2065: 'RhsType': undeclared identifier
C:\Jenkins\workspace\caffe2-builds\py2-cuda9.0-cudnn7-windows-build\third_party\eigen\Eigen\src/LU/InverseImpl.h(366): error C2923: 'std::_Select<__formal>::_Apply': 'RhsType' is not a valid template type argument for parameter '<unnamed-symbol>'
C:\Jenkins\workspace\caffe2-builds\py2-cuda9.0-cudnn7-windows-build\third_party\eigen\Eigen\src/LU/InverseImpl.h(366): error C4430: missing type specifier - int assumed. Note: C++ does not support default-int
C:\Jenkins\workspace\caffe2-builds\py2-cuda9.0-cudnn7-windows-build\third_party\eigen\Eigen\src/LU/InverseImpl.h(366): error C2144: syntax error: 'unknown-type' should be preceded by ')'
C:\Jenkins\workspace\caffe2-builds\py2-cuda9.0-cudnn7-windows-build\third_party\eigen\Eigen\src/LU/InverseImpl.h(366): error C2144: syntax error: 'unknown-type' should be preceded by ';'
C:\Jenkins\workspace\caffe2-builds\py2-cuda9.0-cudnn7-windows-build\third_party\eigen\Eigen\src/LU/InverseImpl.h(367): error C2062: type 'bool' unexpected
C:\Jenkins\workspace\caffe2-builds\py2-cuda9.0-cudnn7-windows-build\third_party\eigen\Eigen\src/LU/InverseImpl.h(369): error C2059: syntax error: ')'
C:\Jenkins\workspace\caffe2-builds\py2-cuda9.0-cudnn7-windows-build\third_party\eigen\Eigen\src/LU/InverseImpl.h(371): error C2143: syntax error: missing ';' before '{'
C:\Jenkins\workspace\caffe2-builds\py2-cuda9.0-cudnn7-windows-build\third_party\eigen\Eigen\src/LU/InverseImpl.h(371): error C2447: '{': missing function header (old-style formal list?)

It doesn't seem to reliably repro.

cc @ezyang @gchanan @zou3519 @seemethere @peterjc123

The text was updated successfully, but these errors were encountered:

ezyang · 2019-09-03T17:57:14Z

I got some sort of similar-ish error persistently on a test branch:

14:06:51 L1Cost.cu
14:06:51 C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/include\thrust/detail/allocator/allocator_traits.inl(101): error C2993: 'T': illegal type for non-type template parameter '__formal'
14:06:51 C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/include\thrust/detail/allocator/allocator_traits.inl(101): note: see reference to class template instantiation 'thrust::detail::allocator_traits_detail::has_member_construct2_impl_has_member<T,Result(Arg1,Arg2)>' being compiled
14:06:51 C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/include\thrust/detail/allocator/allocator_traits.inl(101): error C2065: 't': undeclared identifier
14:06:51 C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/include\thrust/detail/allocator/allocator_traits.inl(101): error C2923: 'std::_Select<__formal>::_Apply': 't' is not a valid template type argument for parameter '<unnamed-symbol>'
14:06:51 C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/include\thrust/detail/allocator/allocator_traits.inl(101): error C2062: type 'unknown-type' unexpected
14:06:51 -- Removing C:/Jenkins/workspace/caffe2-builds/py2-cuda9.0-cudnn7-windows-build/build/caffe2/CMakeFiles/torch.dir/__/aten/src/THCUNN/./torch_generated_L1Cost.cu.obj

ezyang · 2019-09-03T18:16:31Z

Not that persistently; a PR stacked on the failing one succeeded. So there is something nondeterministic going on here.

smessmer · 2019-09-03T18:48:31Z

another instance: CI on #23888

peterjc123 · 2019-09-06T11:52:07Z

Looks like an NVCC bug. The workground is protecting eigen headers with #ifndef __CUDACC__. Can we extract some basic code example and compile it against NVCC?

peterjc123 · 2019-12-12T07:22:50Z

Another one in https://app.circleci.com/jobs/github/pytorch/pytorch/3919969.

unique_ops.cu
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/include\thrust/system/cuda/detail/unique_by_key.h(211): error C2993: 'KeyInputIt': illegal type for non-type template parameter '__formal'
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/include\thrust/system/cuda/detail/unique_by_key.h(246): note: see reference to class template instantiation 'thrust::cuda_cub::__unique_by_key::UniqueByKeyAgent<KeyInputIt,ValInputIt,KeyOutputIt,ValOutputIt,BinaryPred,Size,NumSelectedOutIt>::PtxPlan<Arch>' being compiled
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/include\thrust/system/cuda/detail/unique_by_key.h(589): note: see reference to class template instantiation 'thrust::cuda_cub::__unique_by_key::UniqueByKeyAgent<KeyInputIt,ValInputIt,KeyOutputIt,ValOutputIt,BinaryPred,Size,NumSelectedOutIt>' being compiled
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/include\thrust/system/cuda/detail/unique_by_key.h(211): error C2065: 'IS_FIRST_TILE': undeclared identifier
C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/include\thrust/system/cuda/detail/unique_by_key.h(211): error C2993: 'KeyInputIt': illegal type for non-type template parameter '__formal'

gottbrath · 2019-12-16T21:54:35Z

Can you try with 10.2?

gottbrath · 2019-12-16T22:48:14Z

Mingbo, can you put general information on the issue you are seeing here in this issue?

gottbrath · 2019-12-16T23:00:34Z

latest word on this is

from mingbo:
"from what I can tell, it's a flaky one. might depend on which VS version. I hit the problem before and problem fixed (at least seem so) after upgrading VS 2019 latest version" and "the same error showed up once last week, and re-run the job finished without problem."

jmlundberg · 2020-02-04T09:07:35Z

Hi. Did anyone have this issue persistently, ie that a rebuild did not "fix" it? We've had this issue in another (closed source) project. Exposing cuda to less includes (such as boost) seems to help but is not predictable.

ezyang · 2020-02-04T21:07:57Z

It's never been persistent for us, but this is mostly from our Windows CI builds where we blow everything away and rebuild from scratch, which may help.

peterjc123 · 2020-02-10T03:27:53Z

Yet another occurrence in the forum: https://discuss.pytorch.org/t/how-do-i-get-my-older-gpu-that-supports-cuda-to-work-with-pytorch-1-4/69005/13. It seems to be fixed by set MAX_JOBS=1, which implies there may be a race condition in reading the include files or generating the intermediate source files.

Summary: ## Several flags `/MP[M]`: It is a flag for the compiler `cl`. It leads to object-level multiprocessing. By default, it spawns M processes where M is the number of cores on the PC. `/maxcpucount:[M]`: It is a flag for the generator `msbuild`. It leads to project-level multiprocessing. By default, it spawns M processes where M is the number of cores on the PC. `/p:CL_MPCount=[M]`: It is a flag for the generator `msbuild`. It leads the generator to pass `/MP[M]` to the compiler. `/j[M]`: It is a flag for the generator `ninja`. It leads to object-level multiprocessing. By default, it spawns M processes where M is the number of cores on the PC. ## Reason for the change 1. Object-level multiprocessing is preferred over project-level multiprocessing. 2. ~For ninja, we don't need to set `/MP` otherwise M * M processes will be spawned.~ Actually, it is not correct because in ninja configs, there are only one source file in the command. Therefore, the `/MP` switch should be useless. 3. For msbuild, if it is called through Python configuration scripts, then `/p:CL_MPCount=[M]` will be added, otherwise, we add `/MP` to `CMAKE_CXX_FLAGS`. 4. ~It may be a possible fix for #28271, #27463 and #25393. Because `/MP` is also passed to `nvcc`.~ It is probably not true. Because `/MP` should not be effective given there is only one source file per command. ## Reference 1. https://docs.microsoft.com/en-us/cpp/build/reference/mp-build-with-multiple-processes?view=vs-2019 2. https://github.com/Microsoft/checkedc-clang/wiki/Parallel-builds-of-clang-on-Windows 3. https://blog.kitware.com/cmake-building-with-all-your-cores/ Pull Request resolved: #33120 Differential Revision: D19817227 Pulled By: ezyang fbshipit-source-id: f8d01f835016971729c7a8d8a0d1cb8a8c2c6a5f

peterjc123 · 2020-02-24T17:37:40Z

I was able to reproduce the issue with a more verbose output with #33693: https://app.circleci.com/jobs/github/pytorch/pytorch/3919969. And then I inspected the variables carefully and found out that the VC env is activated twice. We should really avoid that.

ezyang · 2020-02-25T03:24:12Z

You are SO COOL!!!

Summary: Possibly get rid of #28271, #27463 and #25393. Pull Request resolved: #33700 Differential Revision: D20089251 Pulled By: ezyang fbshipit-source-id: 0cfe62b869fb874e25f06894aa76fadc44cf6817

peterjc123 · 2020-02-25T06:34:49Z

I tried to build from scratch several times with this PR and could not reproduce the issue anymore. Let's assume it's fixed.

peterjc123 · 2020-02-26T04:10:38Z

Looks like it was not resolved. https://circleci.com/gh/pytorch/pytorch/4599162?utm_campaign=vcs-integration-link&utm_medium=referral&utm_source=github-build-link

mstfbl · 2021-03-09T10:21:36Z

I have also obtained this error with CUDA 10.1/cuDNN 7.6.4 on Win Server 2019.

[4153/5198] Building NVCC (Device) object caffe2/CMakeFiles/torch_cuda.dir/__/aten/src/THC/generated/torch_cuda_generated_THCTensorTopKHalf.cu.obj
THCTensorTopKHalf.cu
THCTensorTopKHalf.cu
[4154/5198] Building NVCC (Device) object caffe2/CMakeFiles/torch_cuda.dir/__/aten/src/THC/generated/torch_cuda_generated_THCTensorMathReduceLong.cu.obj
FAILED: caffe2/CMakeFiles/torch_cuda.dir/__/aten/src/THC/generated/torch_cuda_generated_THCTensorMathReduceLong.cu.obj 
cmd.exe /C "cd /D F:\agent\_work\1\s\build\caffe2\CMakeFiles\torch_cuda.dir\__\aten\src\THC\generated && C:\Miniconda\envs\windows_2019_py_37\Library\bin\cmake.exe -E make_directory F:/agent/_work/1/s/build/caffe2/CMakeFiles/torch_cuda.dir/__/aten/src/THC/generated/. && C:\Miniconda\envs\windows_2019_py_37\Library\bin\cmake.exe -D verbose:BOOL=OFF -D build_configuration:STRING=Release -D generated_file:STRING=F:/agent/_work/1/s/build/caffe2/CMakeFiles/torch_cuda.dir/__/aten/src/THC/generated/./torch_cuda_generated_THCTensorMathReduceLong.cu.obj -D generated_cubin_file:STRING=F:/agent/_work/1/s/build/caffe2/CMakeFiles/torch_cuda.dir/__/aten/src/THC/generated/./torch_cuda_generated_THCTensorMathReduceLong.cu.obj.cubin.txt -P F:/agent/_work/1/s/build/caffe2/CMakeFiles/torch_cuda.dir/__/aten/src/THC/generated/torch_cuda_generated_THCTensorMathReduceLong.cu.obj.Release.cmake"
THCTensorMathReduceLong.cu
THCTensorMathReduceLong.cu
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\include\thrust\system\cuda\detail\cub\util_type.cuh(889): error C2993: 'T': is not a valid type for non-type template parameter '__formal'
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\include\thrust\system\cuda\detail\cub\util_type.cuh(909): note: see reference to class template instantiation 'thrust::cuda_cub::cub::BinaryOpHasIdxParam<T,BinaryOp>' being compiled
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\include\thrust\system\cuda\detail\cub\util_type.cuh(889): error C2065: '__T3': undeclared identifier
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\include\thrust\system\cuda\detail\cub\util_type.cuh(889): error C2923: 'std::_Select<__formal>::_Apply': '__T3' is not a valid template type argument for parameter '<unnamed-symbol>'
CMake Error at torch_cuda_generated_THCTensorMathReduceLong.cu.obj.Release.cmake:281 (message):
  Error generating file
  F:/agent/_work/1/s/build/caffe2/CMakeFiles/torch_cuda.dir/__/aten/src/THC/generated/./torch_cuda_generated_THCTensorMathReduceLong.cu.obj

leezu · 2021-03-09T15:26:40Z

@mstfbl this is a cuda bug. Please see above discussion and upgrade to cuda 11.

mstfbl · 2021-03-11T01:18:30Z

@leezu I fixed the error by installing sscache for use with cuda 10.1 and cuda 10.2, as done here. https://github.com/pytorch/builder/blob/993e8b275e313641796db8a0b2869d5f3dd13828/windows/build_pytorch.bat#L98-L124

leezu · 2021-03-11T18:00:52Z

@mstfbl it's an intermittent error and the probability of occurrence varies depending on your system. So it's not really fixed, but you may have found a way to reduce the occurrence in your system, which may be sufficient in your case :)

peterbell10 · 2021-05-10T23:51:35Z

Another instance, this time inside protobuf:

roi_align_op.cu
C:/Users/circleci/project/third_party/protobuf/src\google/protobuf/arena.h(427): error C2993: 'T': is not a valid type for non-type template parameter '__formal'
C:/Users/circleci/project/third_party/protobuf/src\google/protobuf/arena.h(459): note: see reference to class template instantiation 'google::protobuf::Arena::InternalHelper<T>' being compiled
C:/Users/circleci/project/third_party/protobuf/src\google/protobuf/arena.h(427): error C2065: 'Second': undeclared identifier
C:/Users/circleci/project/third_party/protobuf/src\google/protobuf/arena.h(427): error C2923: 'std::_Select<__formal>::_Apply': 'Second' is not a valid template type argument for parameter '<unnamed-symbol>'
C:/Users/circleci/project/third_party/protobuf/src\google/protobuf/arena.h(427): error C4430: missing type specifier - int assumed. Note: C++ does not support default-int
C:/Users/circleci/project/third_party/protobuf/src\google/protobuf/arena.h(427): error C2144: syntax error: 'unknown-type' should be preceded by ')'
C:/Users/circleci/project/third_party/protobuf/src\google/protobuf/arena.h(427): error C2144: syntax error: 'unknown-type' should be preceded by ';'
C:/Users/circleci/project/third_party/protobuf/src\google/protobuf/arena.h(427): error C2238: unexpected token(s) preceding ';'
C:/Users/circleci/project/third_party/protobuf/src\google/protobuf/arena.h(427): error C2059: syntax error: ')'
C:/Users/circleci/project/third_party/protobuf/src\google/protobuf/arena.h(427): error C2988: unrecognizable template declaration/definition
C:/Users/circleci/project/third_party/protobuf/src\google/protobuf/arena.h(427): error C2059: syntax error: '<end Parse>'

https://app.circleci.com/pipelines/github/pytorch/pytorch/317648/workflows/504abdce-7887-47a1-bf6a-de417da17503/jobs/13229319

Running nvcc with --verbose --keep I found the syntax error is introduced by cudafe++ which is the cuda compiler front end. cudafe++ operates on the output from the compiler's pre-processor and that output barely even changed to trigger the build error, only line number metadata and some extra whitespace was changed in the pre-processed output.

Here is the diff of the two inputs, one of which failed and the other didn't:

$ diff nvcc-output/roi_align_op.cpp4.ii nvcc-output-2/roi_align_op.cpp4.ii
353964d353963
< #line 229 "C:/Users/circleci/project\\c10/macros/Macros.h"
353966c353965,353966
< #line 231 "C:/Users/circleci/project\\c10/macros/Macros.h"
---
> 
>

And the corresponding change in the output:

$ diff nvcc-output/roi_align_op.compute_75.cudafe1.cpp nvcc-output-2/roi_align_op.compute_75.cudafe1.cpp
124779c124779
< template< class U> static char ArenaConstructable(const typename std::_Select< T> ::template _Apply< U, Second> ::InternalArenaConstructable_ *); 
---
> template< class U> static char ArenaConstructable(const typename U::InternalArenaConstructable_ *); 
162181c162181
< namespace _GLOBAL__N__31_roi_align_op_compute_75_cpp1_ii_c6fef629 { }; using namespace ::at::_GLOBAL__N__31_roi_align_op_compute_75_cpp1_ii_c6fef629; namespace _GLOBAL__N__31_roi_align_op_compute_75_cpp1_ii_c6fef629 { 
---
> namespace _GLOBAL__N__31_roi_align_op_compute_75_cpp1_ii_6d8ec415 { }; using namespace ::at::_GLOBAL__N__31_roi_align_op_compute_75_cpp1_ii_6d8ec415; namespace _GLOBAL__N__31_roi_align_op_compute_75_cpp1_ii_6d8ec415 { 
201406c201406
< #define _NV_ANON_NAMESPACE _GLOBAL__N__31_roi_align_op_compute_75_cpp1_ii_c6fef629
---
> #define _NV_ANON_NAMESPACE _GLOBAL__N__31_roi_align_op_compute_75_cpp1_ii_6d8ec415

This failure reproduced reliably on the CI machine. Running nvcc manually multiple times, it always fails or passes given the same inputs. I think it's a deterministic failure, but is annoyingly sensitive to any changes in the input file whatsoever.

There is an issue with the CUDA implementation which prevents a proper execution. See pytorch/pytorch#25393. Tweaking the compiler settings would allow to get less errors, but it seems impossible to prevent the errors altogether.

Fix Readme and disable MSVC-CUDA 10.2 + Update to the new package status. Simplify the HIP-related INSTALL.md section. + Disable the MSVC-CUDA 10.2 job. There is an issue with the CUDA implementation which prevents a proper execution. See pytorch/pytorch#25393 and NVIDIA/thrust#1090. Tweaking the compiler settings would allow getting fewer errors, but it seems impossible to prevent the errors altogether. Related PR: #852

There is an issue with the CUDA implementation which prevents a proper execution. See pytorch/pytorch#25393. Tweaking the compiler settings would allow to get less errors, but it seems impossible to prevent the errors altogether.

See pytorch#65612 and pytorch#25393 Fixes pytorch#65648

Summary: See #65612 and #25393 Fixes #65648 Pull Request resolved: #65649 Reviewed By: janeyx99 Differential Revision: D31189692 Pulled By: malfet fbshipit-source-id: 6ec0548d5833f3428d882071d26c357d89b0a9ba

ngimel · 2022-02-07T19:43:44Z

closing due to age

ezyang · 2022-02-10T15:56:18Z

Wait, do we think this is fixed? It's not clear to me it is...

seemethere · 2022-02-10T16:41:35Z

I think this was related to CUDA 10.2 builds on windows which we don't actually support anymore so it is probably safe to keep closed.

pytorchbot added the module: windows Windows support for PyTorch label Aug 29, 2019

VitalyFedyunin added module: build Build system issues triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module labels Aug 29, 2019

ezyang changed the title ~~Windows CI error: C2993: 'Derived': illegal type for non-type template parameter '__formal~~ Windows CI intermittent error: C2993: 'Derived': illegal type for non-type template parameter '__formal Sep 3, 2019

ezyang added module: ci Related to continuous integration module: flaky-tests Problem is a flaky test in CI labels Sep 3, 2019

smessmer mentioned this issue Sep 3, 2019

Allow TensorMethods.h to include Dispatcher.h (alternative) #23888

Closed

ezyang mentioned this issue Sep 6, 2019

Nightly build failure on Windows (convert_to_int_of_same_size related) #25763

Closed

peterjc123 mentioned this issue Sep 18, 2019

[C++] Projects fail to build with LibTorch Nightly + CUDA 10.1 #26010

Closed

peterjc123 mentioned this issue Feb 8, 2020

Correct /MP usage in MSVC #33120

Closed

peterjc123 mentioned this issue Feb 24, 2020

Don't activate vc env again for cuda with ninja on Windows #33700

Closed

peterjc123 closed this as completed Feb 25, 2020

peterjc123 reopened this Feb 26, 2020

mszhanyi mentioned this issue Nov 13, 2020

update windows build guide #47840

Closed

peterjc123 mentioned this issue Mar 3, 2021

Windows Nightly failures with syntax errors #53131

Closed

david-wahlberg mentioned this issue Apr 25, 2021

LNK2019 error after static compilation of Libtorch #44814

Open

seemethere removed the triage review label Jul 15, 2021

tcojean mentioned this issue Aug 4, 2021

Fix Readme and disable MSVC-CUDA 10.2 ginkgo-project/ginkgo#852

Merged

ngimel mentioned this issue Sep 24, 2021

win-vs2019-cuda10.2-py3 builds fail with non-deterministic pattern #65612

Closed

malfet mentioned this issue Sep 25, 2021

Nondeterministic Windows CUDA compilation failure #64605

Closed

ngimel mentioned this issue Sep 25, 2021

Stop windows cuda 10.2 CI builds and releases #65648

Closed

malfet mentioned this issue Sep 25, 2021

Stop Win+CUDA-10.2 builds #65649

Closed

malfet added a commit to malfet/pytorch that referenced this issue Sep 29, 2021

Stop Win+CUDA-10.2 builds

d66b3ea

See pytorch#65612 and pytorch#25393 Fixes pytorch#65648

malfet mentioned this issue Nov 22, 2021

1.10 build error on windows (CUDA 10.2) #68685

Closed

ngimel closed this as completed Feb 7, 2022

ezyang unassigned nbcsm Feb 10, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Windows CI intermittent error: C2993: 'Derived': illegal type for non-type template parameter '__formal #25393

Windows CI intermittent error: C2993: 'Derived': illegal type for non-type template parameter '__formal #25393

ezyang commented Aug 29, 2019 •

edited by pytorch-probot bot

Loading

ezyang commented Sep 3, 2019

ezyang commented Sep 3, 2019

smessmer commented Sep 3, 2019

peterjc123 commented Sep 6, 2019

peterjc123 commented Dec 12, 2019

gottbrath commented Dec 16, 2019

gottbrath commented Dec 16, 2019

gottbrath commented Dec 16, 2019

jmlundberg commented Feb 4, 2020

ezyang commented Feb 4, 2020

peterjc123 commented Feb 10, 2020 •

edited

Loading

peterjc123 commented Feb 24, 2020

ezyang commented Feb 25, 2020

peterjc123 commented Feb 25, 2020

peterjc123 commented Feb 26, 2020

mstfbl commented Mar 9, 2021 •

edited

Loading

leezu commented Mar 9, 2021

mstfbl commented Mar 11, 2021

leezu commented Mar 11, 2021

peterbell10 commented May 10, 2021

ngimel commented Feb 7, 2022

ezyang commented Feb 10, 2022

seemethere commented Feb 10, 2022

Windows CI intermittent error: C2993: 'Derived': illegal type for non-type template parameter '__formal #25393

Windows CI intermittent error: C2993: 'Derived': illegal type for non-type template parameter '__formal #25393

Comments

ezyang commented Aug 29, 2019 • edited by pytorch-probot bot Loading

ezyang commented Sep 3, 2019

ezyang commented Sep 3, 2019

smessmer commented Sep 3, 2019

peterjc123 commented Sep 6, 2019

peterjc123 commented Dec 12, 2019

gottbrath commented Dec 16, 2019

gottbrath commented Dec 16, 2019

gottbrath commented Dec 16, 2019

jmlundberg commented Feb 4, 2020

ezyang commented Feb 4, 2020

peterjc123 commented Feb 10, 2020 • edited Loading

peterjc123 commented Feb 24, 2020

ezyang commented Feb 25, 2020

peterjc123 commented Feb 25, 2020

peterjc123 commented Feb 26, 2020

mstfbl commented Mar 9, 2021 • edited Loading

leezu commented Mar 9, 2021

mstfbl commented Mar 11, 2021

leezu commented Mar 11, 2021

peterbell10 commented May 10, 2021

ngimel commented Feb 7, 2022

ezyang commented Feb 10, 2022

seemethere commented Feb 10, 2022

ezyang commented Aug 29, 2019 •

edited by pytorch-probot bot

Loading

peterjc123 commented Feb 10, 2020 •

edited

Loading

mstfbl commented Mar 9, 2021 •

edited

Loading