[NNC] Added matmul for NNC lowering/unified dtypes

[Chillee]

Stack from ghstack:

• [WIP] Added skeleton of python dispatch key #56718 Added skeleton of python dispatch key
• [NNC] ported over sum and softmax to new scheme #56775 [NNC] ported over sum and softmax to new scheme
• [nnc] ported some more ops + added vectors to argvalue #56766 [nnc] ported some more ops + added vectors to argvalue
• [NNC] moved lowerings out of the TensorExprKernel and into independent functions #56679 moved lowerings out of the TensorExprKernel and into independent functions
• [NNC] Added matmul for NNC lowering/unified dtypes #56456 [NNC] Added matmul for NNC lowering/unified dtypes

Differential Revision: D27977532

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 88c77cc (more details on the Dr. CI page):

---

• 3/4 failures possibly* introduced in this PR
  • 2/3 non-scanned failure(s)
• 1/4 tentatively recognized as flaky ❄️
  • Click here to rerun these jobs

---

🕵️ 1 new failure recognized by patterns

The following CI failures do not appear to be due to upstream breakages:

pytorch_windows_vs2019_py36_cuda10.1_test1 (1/1)

Step: "Test" (full log | diagnosis details | 🔁 rerun)

RuntimeError: CUDA error: device-side assert triggered

Traceback (most recent call last):
  File "<string>", line 4, in <module>
  File "C:\Users\circleci\project\build\win_tmp\build\torch\cuda\__init__.py", line 444, in synchronize
    return torch._C._cuda_synchronize()
RuntimeError: CUDA error: device-side assert triggered
C:/Users/circleci/project/aten/src/ATen/native/cuda/TensorCompare.cu:68: block: [0,0,0], thread: [0,0,0] Assertion `input[0] != c10::complex<float>(0, 0)` failed.
Traceback (most recent call last):
  File "<string>", line 4, in <module>
  File "C:\Users\circleci\project\build\win_tmp\build\torch\cuda\__init__.py", line 444, in synchronize
    return torch._C._cuda_synchronize()
RuntimeError: CUDA error: device-side assert triggered
ok (55.422s)
  test_gather_bool (__main__.TestCuda) ... ok (0.005s)
  test_get_device_index (__main__.TestCuda) ... ok (0.005s)
  test_get_set_rng_state_all (__main__.TestCuda) ... skip (0.003s)
  test_grad_scaling_accumulation (__main__.TestCuda) ... ok (0.015s)
  test_grad_scaling_autocast (__main__.TestCuda) ... ok (0.058s)
  test_grad_scaling_clipping (__main__.TestCuda) ... ok (0.041s)
  test_grad_scaling_clipping_separate_unscale (__main__.TestCuda) ... test_cuda.py:2216: FutureWarning: Non-finite norm encountered in torch.nn.utils.clip_grad_norm_; continuing anyway. Note that the default behavior will change in a future release to error out if a non-finite total norm is encountered. At that point, setting error_if_nonfinite=false will be required to retain the old behavior.
  torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm, error_if_nonfinite=False)
ok (0.034s)

---

❄️ 1 failure tentatively classified as flaky

but reruns have not yet been triggered to confirm:

pytorch_macos_10_13_py3_lite_interpreter_build_test (1/1)

Step: "Test" (full log | diagnosis details | 🔁 rerun) ❄️

Apr 24 02:09:51 unknown file: Failure

Apr 24 02:09:47 /Applications/Xcode-12.GM.seed.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/ranlib: file: /Users/distiller/project/torch/lib/libCaffe2_perfkernels_avx512.a(common_avx512.cc.o) has no symbols
Apr 24 02:09:47 warning: /Applications/Xcode-12.GM.seed.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/ranlib: archive library: /Users/distiller/project/torch/lib/libCaffe2_perfkernels_avx512.a the table of contents is empty (no object file members in the library define global symbols)
Apr 24 02:09:49 + popd
Apr 24 02:09:49 ~/project
Apr 24 02:09:49 + /Users/distiller/project/../cpp-build/caffe2/build/bin/test_lite_interpreter_runtime
Apr 24 02:09:50 Note: Google Test filter = *-*_CUDA:*_MultiCUDA
Apr 24 02:09:50 [==========] Running 2 tests from 1 test case.
Apr 24 02:09:50 [----------] Global test environment set-up.
Apr 24 02:09:50 [----------] 2 tests from RunTimeTest
Apr 24 02:09:50 [ RUN      ] RunTimeTest.LoadAndForward
Apr 24 02:09:51 unknown file: Failure
Apr 24 02:09:51 C++ exception with description "open file failed, file path: /Users/distiller/project/test/cpp/lite_interpreter_runtime/sequence.ptl
Apr 24 02:09:51 Exception raised from FileAdapter at /Users/distiller/project/caffe2/serialize/file_adapter.cc:11 (most recent call first):
Apr 24 02:09:51 frame #0: decltype(std::__1::forward<c10::(anonymous namespace)::GetFetchStackTrace()::$_0&>(fp)()) std::__1::__invoke<c10::(anonymous namespace)::GetFetchStackTrace()::$_0&>(c10::(anonymous namespace)::GetFetchStackTrace()::$_0&) + 54 (0x10f65a746 in libc10.dylib)
Apr 24 02:09:51 frame #1: std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > std::__1::__invoke_void_return_wrapper<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > >::__call<c10::(anonymous namespace)::GetFetchStackTrace()::$_0&>(c10::(anonymous namespace)::GetFetchStackTrace()::$_0&) + 54 (0x10f65a6e6 in libc10.dylib)
Apr 24 02:09:51 frame #2: std::__1::__function::__alloc_func<c10::(anonymous namespace)::GetFetchStackTrace()::$_0, std::__1::allocator<c10::(anonymous namespace)::GetFetchStackTrace()::$_0>, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > ()>::operator()() + 54 (0x10f65a6a6 in libc10.dylib)
Apr 24 02:09:51 frame #3: std::__1::__function::__func<c10::(anonymous namespace)::GetFetchStackTrace()::$_0, std::__1::allocator<c10::(anonymous namespace)::GetFetchStackTrace()::$_0>, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > ()>::operator()() + 45 (0x10f6593dd in libc10.dylib)
Apr 24 02:09:51 frame #4: std::__1::__function::__value_func<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > ()>::operator()() const + 75 (0x10f66052b in libc10.dylib)
Apr 24 02:09:51 frame #5: std::__1::function<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > ()>::operator()() const + 35 (0x10f6504d3 in libc10.dylib)
Apr 24 02:09:51 frame #6: c10::Error::Error(c10::SourceLocation, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >) + 78 (0x10f650a4e in libc10.dylib)
Apr 24 02:09:51 frame #7: c10::detail::torchCheckFail(char const*, char const*, unsigned int, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&) + 130 (0x10f64c9a2 in libc10.dylib)

---

ci.pytorch.org: 1 failed

• Failed: pr/pytorch-linux-bionic-rocm4.1-py3.6

---

This comment was automatically generated by Dr. CI (expand for details).

Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions to the (internal) Dr. CI Users group.

[ZolotukhinM]

Overall looks good! Please include a more detailed description in the commit message and add a tag ([NNC] or [TensorExpr]) - it helps to find relevant commits for release notes.

[facebook-github-bot]

@Chillee merged this pull request in bcef7eb.