[PyTorch] Port ExecuTorch bfdot improvement back to ATen BlasKernel #136331

swolchok · 2024-09-19T19:54:15Z

Stack from ghstack (oldest at bottom):

ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. Supersedes #127488 . Includes pytorch/executorch#5444 .

Differential Revision: D63045939

ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. Supersedes #127488 . Includes not-yet-reviewed pytorch/executorch#5444 . Differential Revision: [D63045939](https://our.internmc.facebook.com/intern/diff/D63045939/) [ghstack-poisoned]

pytorch-bot · 2024-09-19T19:54:18Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/136331

📄 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.

❌ 5 New Failures, 1 Unrelated Failure

As of commit 63d29fa with merge base 99eb47f ():

NEW FAILURES - The following jobs have failed:

linux-aarch64 / linux-jammy-aarch64-py3.10 / build (gh)
/var/lib/jenkins/workspace/aten/src/ATen/native/BlasKernel.cpp:582:25: error: inlining failed in call to ‘always_inline’ ‘float at::native::blas_impl::{anonymous}::dot_with_fp32_arith_tail_after_main_loop(const T*, const T*, int64_t, float) [with T = c10::BFloat16]’: target specific option mismatch
linux-binary-libtorch-pre-cxx11 / libtorch-cpu-shared-with-deps-pre-cxx11-test / test (gh)
RuntimeError: recursive_directory_iterator in used pre-CXX11 binaries, see; ['std::filesystem::recursive_directory_iterator::recursion_pending() const', 'std::filesystem::recursive_directory_iterator::depth() const', 'std::filesystem::recursive_directory_iterator::options() const', 'std::filesystem::recursive_directory_iterator::operator*() const', 'std::filesystem::recursive_directory_iterator::disable_recursion_pending()', 'std::filesystem::recursive_directory_iterator::pop(std::error_code&)', 'std::filesystem::recursive_directory_iterator::pop()', 'std::filesystem::recursive_directory_iterator::pop() [clone .cold]', 'std::filesystem::recursive_directory_iterator::increment(std::error_code&)', 'std::filesystem::recursive_directory_iterator::increment(std::error_code&) [clone .cold]', 'std::filesystem::recursive_directory_iterator::operator=(std::filesystem::recursive_directory_iterator&&)', 'std::filesystem::recursive_directory_iterator::operator=(std::filesystem::recursive_directory_iterator const&)', 'std::filesystem::recursive_directory_iterator::recursive_directory_iterator(std::filesystem::path const&, std::filesystem::directory_options, std::error_code*)', 'std::filesystem::recursive_directory_iterator::recursive_directory_iterator(std::filesystem::path const&, std::filesystem::directory_options, std::error_code*)', 'std::filesystem::recursive_directory_iterator::recursive_directory_iterator(std::filesystem::path const&, std::filesystem::directory_options, std::error_code*) [clone .cold]', 'std::filesystem::recursive_directory_iterator::~recursive_directory_iterator()', 'std::filesystem::recursive_directory_iterator::~recursive_directory_iterator()', 'std::filesystem::recursive_directory_iterator::operator++()', 'std::filesystem::recursive_directory_iterator::operator++() [clone .cold]']
linux-binary-manywheel / manywheel-py3_9-cuda11_8-test / test (gh)
RuntimeError: recursive_directory_iterator in used pre-CXX11 binaries, see; ['std::filesystem::recursive_directory_iterator::recursion_pending() const', 'std::filesystem::recursive_directory_iterator::depth() const', 'std::filesystem::recursive_directory_iterator::options() const', 'std::filesystem::recursive_directory_iterator::operator*() const', 'std::filesystem::recursive_directory_iterator::disable_recursion_pending()', 'std::filesystem::recursive_directory_iterator::pop(std::error_code&)', 'std::filesystem::recursive_directory_iterator::pop()', 'std::filesystem::recursive_directory_iterator::pop() [clone .cold]', 'std::filesystem::recursive_directory_iterator::increment(std::error_code&)', 'std::filesystem::recursive_directory_iterator::increment(std::error_code&) [clone .cold]', 'std::filesystem::recursive_directory_iterator::operator=(std::filesystem::recursive_directory_iterator&&)', 'std::filesystem::recursive_directory_iterator::operator=(std::filesystem::recursive_directory_iterator const&)', 'std::filesystem::recursive_directory_iterator::recursive_directory_iterator(std::filesystem::path const&, std::filesystem::directory_options, std::error_code*)', 'std::filesystem::recursive_directory_iterator::recursive_directory_iterator(std::filesystem::path const&, std::filesystem::directory_options, std::error_code*)', 'std::filesystem::recursive_directory_iterator::recursive_directory_iterator(std::filesystem::path const&, std::filesystem::directory_options, std::error_code*) [clone .cold]', 'std::filesystem::recursive_directory_iterator::~recursive_directory_iterator()', 'std::filesystem::recursive_directory_iterator::~recursive_directory_iterator()', 'std::filesystem::recursive_directory_iterator::operator++()', 'std::filesystem::recursive_directory_iterator::operator++() [clone .cold]']
linux-binary-manywheel / manywheel-py3_9-cuda12_1-test / test (gh)
RuntimeError: recursive_directory_iterator in used pre-CXX11 binaries, see; ['std::filesystem::recursive_directory_iterator::recursion_pending() const', 'std::filesystem::recursive_directory_iterator::depth() const', 'std::filesystem::recursive_directory_iterator::options() const', 'std::filesystem::recursive_directory_iterator::operator*() const', 'std::filesystem::recursive_directory_iterator::disable_recursion_pending()', 'std::filesystem::recursive_directory_iterator::pop(std::error_code&)', 'std::filesystem::recursive_directory_iterator::pop()', 'std::filesystem::recursive_directory_iterator::pop() [clone .cold]', 'std::filesystem::recursive_directory_iterator::increment(std::error_code&)', 'std::filesystem::recursive_directory_iterator::increment(std::error_code&) [clone .cold]', 'std::filesystem::recursive_directory_iterator::operator=(std::filesystem::recursive_directory_iterator&&)', 'std::filesystem::recursive_directory_iterator::operator=(std::filesystem::recursive_directory_iterator const&)', 'std::filesystem::recursive_directory_iterator::recursive_directory_iterator(std::filesystem::path const&, std::filesystem::directory_options, std::error_code*)', 'std::filesystem::recursive_directory_iterator::recursive_directory_iterator(std::filesystem::path const&, std::filesystem::directory_options, std::error_code*)', 'std::filesystem::recursive_directory_iterator::recursive_directory_iterator(std::filesystem::path const&, std::filesystem::directory_options, std::error_code*) [clone .cold]', 'std::filesystem::recursive_directory_iterator::~recursive_directory_iterator()', 'std::filesystem::recursive_directory_iterator::~recursive_directory_iterator()', 'std::filesystem::recursive_directory_iterator::operator++()', 'std::filesystem::recursive_directory_iterator::operator++() [clone .cold]']
linux-binary-manywheel / manywheel-py3_9-cuda12_4-test / test (gh)
RuntimeError: recursive_directory_iterator in used pre-CXX11 binaries, see; ['std::filesystem::recursive_directory_iterator::recursion_pending() const', 'std::filesystem::recursive_directory_iterator::depth() const', 'std::filesystem::recursive_directory_iterator::options() const', 'std::filesystem::recursive_directory_iterator::operator*() const', 'std::filesystem::recursive_directory_iterator::disable_recursion_pending()', 'std::filesystem::recursive_directory_iterator::pop(std::error_code&)', 'std::filesystem::recursive_directory_iterator::pop()', 'std::filesystem::recursive_directory_iterator::pop() [clone .cold]', 'std::filesystem::recursive_directory_iterator::increment(std::error_code&)', 'std::filesystem::recursive_directory_iterator::increment(std::error_code&) [clone .cold]', 'std::filesystem::recursive_directory_iterator::operator=(std::filesystem::recursive_directory_iterator&&)', 'std::filesystem::recursive_directory_iterator::operator=(std::filesystem::recursive_directory_iterator const&)', 'std::filesystem::recursive_directory_iterator::recursive_directory_iterator(std::filesystem::path const&, std::filesystem::directory_options, std::error_code*)', 'std::filesystem::recursive_directory_iterator::recursive_directory_iterator(std::filesystem::path const&, std::filesystem::directory_options, std::error_code*)', 'std::filesystem::recursive_directory_iterator::recursive_directory_iterator(std::filesystem::path const&, std::filesystem::directory_options, std::error_code*) [clone .cold]', 'std::filesystem::recursive_directory_iterator::~recursive_directory_iterator()', 'std::filesystem::recursive_directory_iterator::~recursive_directory_iterator()', 'std::filesystem::recursive_directory_iterator::operator++()', 'std::filesystem::recursive_directory_iterator::operator++() [clone .cold]']

FLAKY - The following job failed but was likely due to flakiness present on trunk:

trunk / linux-focal-cuda12.4-py3.10-gcc9-experimental-split-build-test / test (default, 3, 5, lf.linux.4xlarge.nvidia.gpu) (gh) (similar failure)
'test/inductor/test_cudacodecache.py::TestCUDACodeCache::test_async_compile'

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

facebook-github-bot · 2024-09-19T19:54:27Z

This pull request was exported from Phabricator. Differential Revision: D63045939

ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. Supersedes #127488 . Includes not-yet-reviewed pytorch/executorch#5444 . Differential Revision: [D63045939](https://our.internmc.facebook.com/intern/diff/D63045939/) ghstack-source-id: 243596406 Pull Request resolved: #136331

swolchok · 2024-09-19T19:55:34Z

Note that this code is gated off under C10_MOBILE, but we have ExecuTorch for that anyway, right?

aten/src/ATen/native/BlasKernel.cpp

malfet · 2024-09-20T20:51:39Z

c10/macros/Macros.h

+#if defined(_MSC_VER)
+#define C10_ALWAYS_INLINE_ATTRIBUTE
+#elif __has_attribute(always_inline) || defined(__GNUC__)
+#define C10_ALWAYS_INLINE_ATTRIBUTE __attribute__((__always_inline__))
+#else
+#define C10_ALWAYS_INLINE_ATTRIBUTE
+#endif


Can you please move it to a separate PR? Also, why is this needed?(any perf numbers) And why exclude clang?

why is this needed?(any perf numbers)

not having it prevents ForcedUnroll from working properly under -Oz. See pytorch/executorch#5247 .

why exclude clang

clang defines __GNUC__.

separate PR

done. #136445

…lasKernel" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. Supersedes #127488 . Includes not-yet-reviewed pytorch/executorch#5444 . Differential Revision: [D63045939](https://our.internmc.facebook.com/intern/diff/D63045939/) [ghstack-poisoned]

facebook-github-bot · 2024-09-23T17:11:01Z

This pull request was exported from Phabricator. Differential Revision: D63045939

…lasKernel" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. Supersedes #127488 . Includes not-yet-reviewed pytorch/executorch#5444 . Differential Revision: [D63045939](https://our.internmc.facebook.com/intern/diff/D63045939/) [ghstack-poisoned]

facebook-github-bot · 2024-09-23T17:18:02Z

This pull request was exported from Phabricator. Differential Revision: D63045939

Pull Request resolved: #136331 ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. Supersedes #127488 . Includes not-yet-reviewed pytorch/executorch#5444 . ghstack-source-id: 244187352 Differential Revision: [D63045939](https://our.internmc.facebook.com/intern/diff/D63045939/)

…lasKernel" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. Supersedes #127488 . Includes pytorch/executorch#5444 . Differential Revision: [D63045939](https://our.internmc.facebook.com/intern/diff/D63045939/) [ghstack-poisoned]

facebook-github-bot · 2024-09-30T16:23:56Z

This pull request was exported from Phabricator. Differential Revision: D63045939

Pull Request resolved: #136331 ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. Supersedes #127488 . Includes pytorch/executorch#5444 . ghstack-source-id: 245455630 Differential Revision: [D63045939](https://our.internmc.facebook.com/intern/diff/D63045939/)

albanD

CI failure look very much relevant!

aten/src/ATen/native/BlasKernel.cpp

albanD · 2024-10-01T14:43:25Z

aten/src/ATen/native/BlasKernel.cpp

@@ -301,7 +302,7 @@ static constexpr auto kF16RegistersPerIterationShift = kF16ElementsPerIterationS
 static constexpr auto kF16RegistersPerIteration = 1 << kF16RegistersPerIterationShift;
 static_assert(kF16RegistersPerIteration == kF16ElementsPerIteration / kF16ElementsPerRegister);

-static inline double reduce(float16x8_t x[kF16RegistersPerIteration]) {
+static inline float reduce(float16x8_t x[kF16RegistersPerIteration]) {


Wasn't this done on purpose to preserve accuracy?

I believe @malfet asked for it. perhaps I got confused though.

Ok, in any case, the numerical error on macos arm seem relevant and needs investigation right?

Ok, in any case, the numerical error on macos arm seem relevant and needs investigation right?

yes, it's fixed. I accidentally replaced the float16 kernel with the bfloat16 kernel due to name changes

…lasKernel" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. Supersedes #127488 . Includes pytorch/executorch#5444 . Differential Revision: [D63045939](https://our.internmc.facebook.com/intern/diff/D63045939/) [ghstack-poisoned]

facebook-github-bot · 2024-10-02T15:37:55Z

This pull request was exported from Phabricator. Differential Revision: D63045939

Pull Request resolved: #136331 ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. Supersedes #127488 . Includes pytorch/executorch#5444 . ghstack-source-id: 245903000 Differential Revision: [D63045939](https://our.internmc.facebook.com/intern/diff/D63045939/)

swolchok · 2024-10-04T20:18:27Z

I will probably reapply this and just disable the BF16 option for GCC.

abhishek-iitmadras · 2024-10-04T20:30:09Z

I will probably reapply this and just disable the BF16 option for GCC.

Hi @swolchok (cc @malfet )

If I add below code:

TARGET_ARM_BF16_ATTRIBUTE
static C10_ALWAYS_INLINE float dot_with_fp32_arith_tail_after_main_loop_bf16(
    const BFloat16* vec1,
    const BFloat16* vec2,
    int64_t len,
    float reducedSum) {
  const auto len_aligned = len & ~(kF32ElementsPerIteration - 1);
  for (int j = len_aligned; j < len; ++j) {
    reducedSum += static_cast<float>(vec1[j]) * static_cast<float>(vec2[j]);
  }
  return reducedSum;
}

And modified dot_with_fp32_arith_bfdot to use dot_with_fp32_arith_tail_after_main_loop_bf16 specific to bf16 then build pass successfully.

#if COMPILER_SUPPORTS_BF16_TARGET
TARGET_ARM_BF16_ATTRIBUTE float
dot_with_fp32_arith_bfdot(const BFloat16* vec1, const BFloat16* vec2, int64_t len) {
  auto reducedSum = dot_with_fp32_arith_main_loop_bfdot(vec1, vec2, len);
  return dot_with_fp32_arith_tail_after_main_loop_bf16(vec1, vec2, len, reducedSum);
}
#endif // COMPILER_SUPPORTS_BF16_TARGET

So the issue here might me like in function dot_with_fp32_arith_bfdot, You try to return dot_with_fp32_arith_tail_after_main_loop and in dot_with_fp32_arith_tail_after_main_loop , you use dot_with_fp32_arith_vectorized_tail_inner_loop which is not specific to bf16.

Correct me if I am wrong.

swolchok · 2024-10-04T20:33:35Z

you use dot_with_fp32_arith_vectorized_tail_inner_loop which is not specific to bf16.

Correct me if I am wrong.

It should be fine to inline a not-specific-to-bf16 function in a specific-to-bf16 function. In fact, clang does it correctly. I would prefer not to duplicate code to work around compiler bugs.

swolchok · 2024-10-04T22:21:41Z

prefer not to duplicate code

it will be slightly ugly, but we can put the body of the one function we'd need to duplicate in a macro so as not to punish people for using ARM GCC. I will try that.

…Try #2 ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

…Try #2 ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) ghstack-source-id: 246411194 Pull Request resolved: #137377

swolchok · 2024-10-04T22:30:29Z

#137377 is the second attempt

…t back to ATen BlasKernel, Try #2" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

…lasKernel, Try #2" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

…t back to ATen BlasKernel, Try #2" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

…lasKernel, Try #2" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

…Try #2 Pull Request resolved: #137377 ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . ghstack-source-id: 246616406 Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/)

…t back to ATen BlasKernel, Try #2" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

…lasKernel, Try #2" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

…fdot improvement back to ATen BlasKernel, Try #2" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

… back to ATen BlasKernel, Try #2" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

…vement back to ATen BlasKernel, Try #2" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

…Ten BlasKernel, Try #2" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

…Try #2 Pull Request resolved: #137377 ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . ghstack-source-id: 246956192 Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/)

…t back to ATen BlasKernel, Try #2" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

…lasKernel, Try #2" ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

…Try #2 (#137377) ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) Pull Request resolved: #137377 Approved by: https://github.com/malfet

…Try #2 ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. First attempt was #136331 . Differential Revision: [D63923166](https://our.internmc.facebook.com/intern/diff/D63923166/) [ghstack-poisoned]

facebook-github-bot added the fb-exported label Sep 19, 2024

swolchok requested review from albanD and malfet September 19, 2024 19:54

swolchok added the topic: performance topic category label Sep 19, 2024

swolchok mentioned this pull request Sep 19, 2024

Use bfdot instruction in ARM bfloat16 dot product if compiling with it available #127488

Closed

swolchok added the release notes: performance_as_product label Sep 19, 2024

malfet reviewed Sep 20, 2024

View reviewed changes

swolchok mentioned this pull request Sep 23, 2024

[PyTorch] Add attribute version of C10_ALWAYS_INLINE #136445

Closed

swolchok requested a review from malfet September 23, 2024 22:08

malfet approved these changes Sep 30, 2024

View reviewed changes

pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Sep 30, 2024

albanD requested changes Oct 1, 2024

View reviewed changes

swolchok requested a review from albanD October 2, 2024 15:40

swolchok mentioned this pull request Oct 4, 2024

[PyTorch] Port ExecuTorch bfdot improvement back to ATen BlasKernel, Try #2 #137377

Closed

swolchok closed this Oct 4, 2024

swolchok mentioned this pull request Oct 10, 2024

[PyTorch] Port ExecuTorch bfdot improvement back to ATen BlasKernel, Try #2 #137727

Closed

github-actions bot deleted the gh/swolchok/647/head branch November 6, 2024 02:07

[PyTorch] Port ExecuTorch bfdot improvement back to ATen BlasKernel #136331

[PyTorch] Port ExecuTorch bfdot improvement back to ATen BlasKernel #136331

Uh oh!

Conversation

swolchok commented Sep 19, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Sep 19, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/136331

❌ 5 New Failures, 1 Unrelated Failure

Uh oh!

facebook-github-bot commented Sep 19, 2024

Uh oh!

swolchok commented Sep 19, 2024

Uh oh!

Uh oh!

Uh oh!

malfet Sep 20, 2024

Choose a reason for hiding this comment

Uh oh!

swolchok Sep 23, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

swolchok Sep 23, 2024

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented Sep 23, 2024

Uh oh!

facebook-github-bot commented Sep 23, 2024

Uh oh!

facebook-github-bot commented Sep 30, 2024

Uh oh!

albanD left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

albanD Oct 1, 2024

Choose a reason for hiding this comment

Uh oh!

swolchok Oct 1, 2024

Choose a reason for hiding this comment

Uh oh!

albanD Oct 1, 2024

Choose a reason for hiding this comment

Uh oh!

swolchok Oct 2, 2024

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented Oct 2, 2024

Uh oh!

swolchok commented Oct 4, 2024

Uh oh!

abhishek-iitmadras commented Oct 4, 2024

Uh oh!

swolchok commented Oct 4, 2024

Uh oh!

swolchok commented Oct 4, 2024

Uh oh!

swolchok commented Oct 4, 2024

Uh oh!

Uh oh!

swolchok commented Sep 19, 2024 •

edited

Loading

pytorch-bot bot commented Sep 19, 2024 •

edited

Loading

swolchok Sep 23, 2024 •

edited

Loading