Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

add patches to fix PyTorch 1.13.1 w/ foss/2022a on POWER + fix flaky test_jit_legacy test #18500

Merged
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
@@ -0,0 +1,25 @@
Fix use-after free leading to random failures in nn/test_embedding
on e.g. POWER platforms where FBGEMM isn't used

From https://github.com/pytorch/pytorch/pull/84750

Author: Alexander Grund (TU Dresden)

diff --git a/aten/src/ATen/native/quantized/cpu/qembeddingbag_prepack.cpp b/aten/src/ATen/native/quantized/cpu/qembeddingbag_prepack.cpp
index 224a66f8abf..f4d018007bf 100644
--- a/aten/src/ATen/native/quantized/cpu/qembeddingbag_prepack.cpp
+++ b/aten/src/ATen/native/quantized/cpu/qembeddingbag_prepack.cpp
@@ -252,9 +252,10 @@ Tensor& qembeddingbag_byte_prepack_out(Tensor& output, const Tensor& weight) {
}

#else
- const auto weight_data = weight_contig->scalar_type() == at::ScalarType::Half
- ? weight_contig->to(at::ScalarType::Float).data_ptr<float>()
- : weight_contig->data_ptr<float>();
+ const Tensor& float_weight = weight_contig->scalar_type() == at::ScalarType::Half
+ ? weight_contig->to(at::ScalarType::Float)
+ : *weight_contig;
+ const auto weight_data = float_weight.data_ptr<float>();
constexpr float kEpsilon = 1e-8f;
for (auto row : c10::irange(embedding_rows)) {
const float* input_row = weight_data + row * embedding_cols;
Original file line number Diff line number Diff line change
@@ -0,0 +1,48 @@
There is a bug in the fallback path for the case where FBGEMM isn't available (e.g. on POWER)
which leads to a race condition:
Data is "copied" for the full buffer while it is processed in chunks by different threads.
This a) duplicates the work and b) might write incomplete/wrong data to the output.

Found in failing test_embedding_bag_half_cpu_* of nn/test_embedding:
ERROR: test_embedding_bag_half_cpu_int32_int32 (__main__.TestEmbeddingNNDeviceTypeCPU)
----------------------------------------------------------------------
Traceback (most recent call last):
File "/dev/shm/s3248973-EasyBuild/PyTorch/1.13.1/foss-2022a/pytorch-v1.13.1/test/nn/test_embedding.py", line 936, in _test_EmbeddingBag_vs_Embedding
self.assertEqual(output, ref_output, atol=dtype2prec_DONTUSE[wdtype], rtol=0)
File "/tmp/eb-tmp-2022a/lib/python3.10/site-packages/torch/testing/_internal/common_utils.py", line 2470, in assertEqual
assert_equal(
File "/tmp/eb-tmp-2022a/lib/python3.10/site-packages/torch/testing/_comparison.py", line 1093, in assert_equal
raise error_metas[0].to_error(msg)
AssertionError: Tensor-likes are not close!

Mismatched elements: 1 / 4 (25.0%)
Greatest absolute difference: 1.18359375 at index (1, 1) (up to 0.01 allowed)
Greatest relative difference: 1.0 at index (1, 1) (up to 0 allowed)


Introduced by https://github.com/pytorch/pytorch/pull/74844

Author: Alexander Grund (TU Dresden)

diff --git a/aten/src/ATen/native/EmbeddingBag.cpp b/aten/src/ATen/native/EmbeddingBag.cpp
index 6d8cea26f52..604ea16bace 100644
--- a/aten/src/ATen/native/EmbeddingBag.cpp
+++ b/aten/src/ATen/native/EmbeddingBag.cpp
@@ -246,7 +246,7 @@ index_select_add(const Tensor &select_indices,
/*scale_bias=*/nullptr,
/*normalize_by_lengths=*/false,
/*out=*/output_data_fp32 + start_idx * ddim);
- for (const auto i : c10::irange(output_size)) {
+ for (const auto i : c10::irange(start_idx, end_idx)) {
// Convert FP32 intermediate buffer result back to FP16 for output dtype
for (const auto d : c10::irange(ddim)) {
(output_data + i * ddim)[d] = static_cast<at::Half>((output_data_fp32 + ddim * i)[d]);
@@ -590,7 +590,7 @@ index_select_scale_add(const Tensor &select_indices,
/*scale_bias=*/nullptr,
/*normalize_by_lengths=*/false,
/*out=*/output_data_fp32 + start_idx * ddim);
- for (const auto i : c10::irange(output_size)) {
+ for (const auto i : c10::irange(start_idx, end_idx)) {
// Convert FP32 intermediate buffer result back to FP16 for output dtype
for (const auto d : c10::irange(ddim)) {
(output_data + i * ddim)[d] = static_cast<at::Half>((output_data_fp32 + ddim * i)[d]);
8 changes: 8 additions & 0 deletions easybuild/easyconfigs/p/PyTorch/PyTorch-1.13.1-foss-2022a.eb
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,9 @@ sources = ['%(namelower)s-v%(version)s.tar.gz']
patches = [
'PyTorch-1.7.0_disable-dev-shm-test.patch',
'PyTorch-1.10.0_fix-kineto-crash.patch',
'PyTorch-1.11.0_fix-fp16-quantization-without-fbgemm.patch',
'PyTorch-1.11.1_skip-test_init_from_local_shards.patch',
'PyTorch-1.12.0_fix-EmbeddingBag-without-fbgemm.patch',
'PyTorch-1.12.1_add-hypothesis-suppression.patch',
'PyTorch-1.12.1_fix-skip-decorators.patch',
'PyTorch-1.12.1_fix-test_cpp_extensions_jit.patch',
Expand All @@ -22,6 +24,7 @@ patches = [
'PyTorch-1.12.1_fix-vsx-loadu.patch',
'PyTorch-1.12.1_fix-vsx-vector-funcs.patch',
'PyTorch-1.12.1_skip-test_round_robin.patch',
'PyTorch-1.13.1_fix-flaky-jit-test.patch',
'PyTorch-1.13.1_fix-fsdp-fp16-test.patch',
'PyTorch-1.13.1_fix-pytest-args.patch',
'PyTorch-1.13.1_fix-test-ops-conf.patch',
Expand All @@ -36,8 +39,12 @@ checksums = [
{'pytorch-v1.13.1.tar.gz': 'dbc229ee9750b02b514937d017744443a269ea0241ed3f32b9af0703589d25d4'},
{'PyTorch-1.7.0_disable-dev-shm-test.patch': '622cb1eaeadc06e13128a862d9946bcc1f1edd3d02b259c56a9aecc4d5406b8a'},
{'PyTorch-1.10.0_fix-kineto-crash.patch': 'dc467333b28162149af8f675929d8c6bf219f23230bfc0d39af02ba4f6f882eb'},
{'PyTorch-1.11.0_fix-fp16-quantization-without-fbgemm.patch':
'cc526130b6446bbbf5f0f7372d3aeee3e7d4c4d6e471524dff028b430b152934'},
{'PyTorch-1.11.1_skip-test_init_from_local_shards.patch':
'4aeb1b0bc863d4801b0095cbce69f8794066748f0df27c6aaaf729c5ecba04b7'},
{'PyTorch-1.12.0_fix-EmbeddingBag-without-fbgemm.patch':
'090598592283e3fc46ee08a68b6a6afe07be41b26514afba51834408bf1c98ed'},
{'PyTorch-1.12.1_add-hypothesis-suppression.patch':
'e71ffb94ebe69f580fa70e0de84017058325fdff944866d6bd03463626edc32c'},
{'PyTorch-1.12.1_fix-skip-decorators.patch': 'e3ca6e42b2fa592ea095939fb59ab875668a058479407db3f3684cc5c6f4146c'},
Expand All @@ -51,6 +58,7 @@ checksums = [
{'PyTorch-1.12.1_fix-vsx-loadu.patch': '8bfe3c94ada1dd1f7974a1261a8b576fb7ae944050fa1c7830fca033831123b2'},
{'PyTorch-1.12.1_fix-vsx-vector-funcs.patch': 'caccbf60f62eac313896c1eaec78b08f5d0fdfcb907079087490bb13d1561aa2'},
{'PyTorch-1.12.1_skip-test_round_robin.patch': '63d4849b78605aa088fdff695637d9473ea60dee603a3ff7f788690d70c55349'},
{'PyTorch-1.13.1_fix-flaky-jit-test.patch': '71efdeb29b5e5b4982c9f5cb2182733654a34d52f85bb5487bc4d7d99b86101b'},
{'PyTorch-1.13.1_fix-fsdp-fp16-test.patch': '8ae68e60d6e1f92f50322b7f0381c7e65251fba32d7606e3a238a36a2f55b5cf'},
{'PyTorch-1.13.1_fix-pytest-args.patch': 'd3e3c841cf8d73683750f29326f2be56ee0bb5df7ff522baf7d7c3f301a91ec2'},
{'PyTorch-1.13.1_fix-test-ops-conf.patch': 'df652eec7753864ebebbfeca546929a53e3fb8f24259d5c9b964266a8551198c'},
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,21 @@
Especially `test_jit_legacy` seems to be flaky.
https://github.com/pytorch/pytorch/commit/316ba9e6fc9e2c309c1b3785e35393b4a727b918
makes the JIT tests run serially avoiding potential races.
So backport that commit.

Author: Alexander Grund (TU Dresden)

diff --git a/test/run_test.py b/test/run_test.py
index f7c80f3f0a6..9d9b0c553e9 100755
--- a/test/run_test.py
+++ b/test/run_test.py
@@ -994,7 +994,8 @@ def must_serial(file: str) -> bool:
"distributed" in file or
file in CUSTOM_HANDLERS or
file in RUN_PARALLEL_BLOCKLIST or
- file in CI_SERIAL_LIST
+ file in CI_SERIAL_LIST or
+ file in JIT_EXECUTOR_TESTS
)