Issue Station
Sign In
Tools hereToolbox

pytorch - 💡(How to fix) Fix [Inductor] _benchmark_collective_with_cuda_events_impl aborts on scalar SymInt kwargs to functional collectives (dynamic shapes + collective_estimator="benchmark")

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…

Error Message

[rank3]: Traceback (most recent call last): [rank3]: File "/workspace/repro_benchmark_symint.py", line 62, in <module> [rank3]: main() [rank3]: File "/workspace/repro_benchmark_symint.py", line 54, in main [rank3]: f(torch.randn(n, HIDDEN, device="cuda"), prompt, w) [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_dynamo/eval_frame.py", line 1034, in compile_wrapper [rank3]: raise e.remove_dynamo_frames() from None # see TORCHDYNAMO_VERBOSE=1 [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1051, in _compile_fx_inner [rank3]: raise InductorError(e, currentframe()).with_traceback( [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1035, in _compile_fx_inner [rank3]: mb_compiled_graph = fx_codegen_and_compile( [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1796, in fx_codegen_and_compile [rank3]: return scheme.codegen_and_compile(gm, example_inputs, inputs_to_check, graph_kwargs) [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1342, in codegen_and_compile [rank3]: _recursive_post_grad_passes(gm, is_inference=is_inference) [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 581, in _recursive_post_grad_passes [rank3]: post_grad_passes(gm, is_inference) [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/post_grad.py", line 353, in post_grad_passes [rank3]: GraphTransformObserver(gm, "overlap_scheduling").apply_graph_pass( [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/fx/passes/graph_transform_observer.py", line 103, in apply_graph_pass [rank3]: return pass_fn(self.gm.graph) [rank3]: ^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/post_grad.py", line 354, in <lambda> [rank3]: lambda graph: schedule_overlap_bucketing_from_inductor_configs( [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 1720, in schedule_overlap_bucketing_from_inductor_configs [rank3]: return schedule_overlap_bucketing(gm, **kwargs) # type: ignore[arg-type] [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 1669, in schedule_overlap_bucketing [rank3]: ).run() [rank3]: ^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 936, in run [rank3]: self._align_compute_nodes_runtime_estimations_across_all_distributed_ranks() [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 758, in _align_compute_nodes_runtime_estimations_across_all_distributed_ranks [rank3]: cuda_val, cuda_key = benchmark_collective_with_cuda_events(n, nruns=5) [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/node_runtime_estimation.py", line 196, in benchmark_collective_with_cuda_events [rank3]: return benchmark_collective_with_cuda_events_impl(n, nruns) [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/node_runtime_estimation.py", line 262, in benchmark_collective_with_cuda_events_impl [rank3]: runtime = _benchmark_collective_with_cuda_events_impl(n, args, kwargs, nruns) [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/node_runtime_estimation.py", line 166, in _benchmark_collective_with_cuda_events_impl [rank3]: result = n.target(*args, **kwargs) # type: ignore[operator] [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 871, in call [rank3]: return self._op(*args, **kwargs) [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_library/autograd.py", line 112, in autograd_impl [rank3]: result = forward_no_grad(*args, Metadata(keyset, keyword_only_args)) [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_library/autograd.py", line 41, in forward_no_grad [rank3]: result = op.redispatch(keyset & _C._after_autograd_keyset, *args, **kwargs) [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 878, in redispatch [rank3]: return self._handle.redispatch_boxed(keyset, *args, **kwargs) # type: ignore[return-value] [rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank3]: torch._inductor.exc.InductorError: RuntimeError: when unpacking SymInt, expected int but got ((s77 + s9 + (PythonMod(-s77, 4)))//4) [rank3]: Exception raised from expect_int at /opt/pytorch/pytorch/c10/core/SymInt.h:132 (most recent call first): [rank3]: frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x7cccebffe578 in /usr/local/lib/python3.12/dist-packages/torch/lib/libc10.so) [rank3]: frame #1: <unknown function> + 0x153a3c5 (0x7ccd0af653c5 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cpu.so) [rank3]: frame #2: <unknown function> + 0x675e943 (0x7ccd10189943 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cpu.so) [rank3]: frame #3: <unknown function> + 0x828f6c (0x7ccd186e6f6c in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #4: <unknown function> + 0x821498 (0x7ccd186df498 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #5: <unknown function> + 0xe92864 (0x7ccd18d50864 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #6: <unknown function> + 0xe92c0a (0x7ccd18d50c0a in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #7: <unknown function> + 0x467292 (0x7ccd18325292 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #8: /usr/bin/python3() [0x581e9f] [rank3]: frame #9: _PyObject_MakeTpCall + 0x75 (0x548f75 in /usr/bin/python3) [rank3]: frame #10: /usr/bin/python3() [0x54cc29] [rank3]: frame #11: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3) [rank3]: frame #12: /usr/bin/python3() [0x54caad] [rank3]: frame #13: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3) [rank3]: frame #14: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3) [rank3]: frame #15: <unknown function> + 0xe99b91 (0x7ccd18d57b91 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #16: <unknown function> + 0xe9a212 (0x7ccd18d58212 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #17: <unknown function> + 0x828f6c (0x7ccd186e6f6c in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #18: <unknown function> + 0x821498 (0x7ccd186df498 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #19: <unknown function> + 0x6442c9a (0x7ccd0fe6dc9a in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cpu.so) [rank3]: frame #20: <unknown function> + 0xbe378b (0x7ccd18aa178b in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #21: <unknown function> + 0xbe3c34 (0x7ccd18aa1c34 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #22: <unknown function> + 0xac2b99 (0x7ccd18980b99 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #23: <unknown function> + 0x467292 (0x7ccd18325292 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so) [rank3]: frame #24: /usr/bin/python3() [0x581e9f] [rank3]: frame #25: PyObject_Call + 0x9c (0x54b11c in /usr/bin/python3) [rank3]: frame #26: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3) [rank3]: frame #27: _PyObject_Call_Prepend + 0xc2 (0x54a7c2 in /usr/bin/python3) [rank3]: frame #28: /usr/bin/python3() [0x5a3458] [rank3]: frame #29: PyObject_Call + 0x9c (0x54b11c in /usr/bin/python3) [rank3]: frame #30: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3) [rank3]: frame #31: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3) [rank3]: frame #32: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3) [rank3]: frame #33: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3) [rank3]: frame #34: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3) [rank3]: frame #35: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3) [rank3]: frame #36: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3) [rank3]: frame #37: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3) [rank3]: frame #38: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3) [rank3]: frame #39: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3) [rank3]: frame #40: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3) [rank3]: frame #41: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3) [rank3]: frame #42: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3) [rank3]: frame #43: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3) [rank3]: frame #44: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3) [rank3]: frame #45: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3) [rank3]: frame #46: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3) [rank3]: frame #47: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3) [rank3]: frame #48: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3) [rank3]: frame #49: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3) [rank3]: frame #50: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3) [rank3]: frame #51: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3) [rank3]: frame #52: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3) [rank3]: frame #53: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3) [rank3]: frame #54: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3) [rank3]: frame #55: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3) [rank3]: frame #56: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3) [rank3]: frame #57: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3) [rank3]: frame #58: /usr/bin/python3() [0x64f4d4] [rank3]: frame #59: _PyObject_MakeTpCall + 0x75 (0x548f75 in /usr/bin/python3) [rank3]: frame #60: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3) [rank3]: frame #61: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3) [rank3]: frame #62: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)

Fix Action

Fix / Workaround

torch._inductor.fx_passes.overlap_scheduling.OverlapScheduler._align_compute_nodes_runtime_estimations_across_all_distributed_ranks calls _benchmark_collective_with_cuda_events_impl to time each functional collective via n.target(*args, **kwargs). With aten_distributed_optimizations.collective_estimator = "benchmark" and dynamic input shapes (e.g. a sequence dim padded to a multiple of world_size, as ulysses-style sequence parallel does), the call crashes inside the C++ dispatcher with expect_int when a scalar SymInt argument reaches an op that takes an int.

[rank3]: Traceback (most recent call last):
[rank3]:   File "/workspace/repro_benchmark_symint.py", line 62, in <module>
[rank3]:     main()
[rank3]:   File "/workspace/repro_benchmark_symint.py", line 54, in main
[rank3]:     f(torch.randn(n, HIDDEN, device="cuda"), prompt, w)
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_dynamo/eval_frame.py", line 1034, in compile_wrapper
[rank3]:     raise e.remove_dynamo_frames() from None  # see TORCHDYNAMO_VERBOSE=1
[rank3]:     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1051, in _compile_fx_inner
[rank3]:     raise InductorError(e, currentframe()).with_traceback(
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1035, in _compile_fx_inner
[rank3]:     mb_compiled_graph = fx_codegen_and_compile(
[rank3]:                         ^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1796, in fx_codegen_and_compile
[rank3]:     return scheme.codegen_and_compile(gm, example_inputs, inputs_to_check, graph_kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1342, in codegen_and_compile
[rank3]:     _recursive_post_grad_passes(gm, is_inference=is_inference)
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 581, in _recursive_post_grad_passes
[rank3]:     post_grad_passes(gm, is_inference)
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/post_grad.py", line 353, in post_grad_passes
[rank3]:     GraphTransformObserver(gm, "overlap_scheduling").apply_graph_pass(
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/fx/passes/graph_transform_observer.py", line 103, in apply_graph_pass
[rank3]:     return pass_fn(self.gm.graph)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/post_grad.py", line 354, in <lambda>
[rank3]:     lambda graph: schedule_overlap_bucketing_from_inductor_configs(
[rank3]:                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 1720, in schedule_overlap_bucketing_from_inductor_configs
[rank3]:     return schedule_overlap_bucketing(gm, **kwargs)  # type: ignore[arg-type]
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 1669, in schedule_overlap_bucketing
[rank3]:     ).run()
[rank3]:       ^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 936, in run
[rank3]:     self._align_compute_nodes_runtime_estimations_across_all_distributed_ranks()
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 758, in _align_compute_nodes_runtime_estimations_across_all_distributed_ranks
[rank3]:     cuda_val, cuda_key = benchmark_collective_with_cuda_events(n, nruns=5)
[rank3]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/node_runtime_estimation.py", line 196, in benchmark_collective_with_cuda_events
[rank3]:     return benchmark_collective_with_cuda_events_impl(n, nruns)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/node_runtime_estimation.py", line 262, in benchmark_collective_with_cuda_events_impl
[rank3]:     runtime = _benchmark_collective_with_cuda_events_impl(n, args, kwargs, nruns)
[rank3]:               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/node_runtime_estimation.py", line 166, in _benchmark_collective_with_cuda_events_impl
[rank3]:     result = n.target(*args, **kwargs)  # type: ignore[operator]
[rank3]:              ^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 871, in __call__
[rank3]:     return self._op(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_library/autograd.py", line 112, in autograd_impl
[rank3]:     result = forward_no_grad(*args, Metadata(keyset, keyword_only_args))
[rank3]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_library/autograd.py", line 41, in forward_no_grad
[rank3]:     result = op.redispatch(keyset & _C._after_autograd_keyset, *args, **kwargs)
[rank3]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 878, in redispatch
[rank3]:     return self._handle.redispatch_boxed(keyset, *args, **kwargs)  # type: ignore[return-value]
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: torch._inductor.exc.InductorError: RuntimeError: when unpacking SymInt, expected int but got ((s77 + s9 + (PythonMod(-s77, 4)))//4)
[rank3]: Exception raised from expect_int at /opt/pytorch/pytorch/c10/core/SymInt.h:132 (most recent call first):
[rank3]: frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x7cccebffe578 in /usr/local/lib/python3.12/dist-packages/torch/lib/libc10.so)
[rank3]: frame #1: <unknown function> + 0x153a3c5 (0x7ccd0af653c5 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cpu.so)
[rank3]: frame #2: <unknown function> + 0x675e943 (0x7ccd10189943 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cpu.so)
[rank3]: frame #3: <unknown function> + 0x828f6c (0x7ccd186e6f6c in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #4: <unknown function> + 0x821498 (0x7ccd186df498 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #5: <unknown function> + 0xe92864 (0x7ccd18d50864 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #6: <unknown function> + 0xe92c0a (0x7ccd18d50c0a in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #7: <unknown function> + 0x467292 (0x7ccd18325292 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #8: /usr/bin/python3() [0x581e9f]
[rank3]: frame #9: _PyObject_MakeTpCall + 0x75 (0x548f75 in /usr/bin/python3)
[rank3]: frame #10: /usr/bin/python3() [0x54cc29]
[rank3]: frame #11: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #12: /usr/bin/python3() [0x54caad]
[rank3]: frame #13: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #14: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #15: <unknown function> + 0xe99b91 (0x7ccd18d57b91 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #16: <unknown function> + 0xe9a212 (0x7ccd18d58212 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #17: <unknown function> + 0x828f6c (0x7ccd186e6f6c in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #18: <unknown function> + 0x821498 (0x7ccd186df498 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #19: <unknown function> + 0x6442c9a (0x7ccd0fe6dc9a in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cpu.so)
[rank3]: frame #20: <unknown function> + 0xbe378b (0x7ccd18aa178b in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #21: <unknown function> + 0xbe3c34 (0x7ccd18aa1c34 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #22: <unknown function> + 0xac2b99 (0x7ccd18980b99 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #23: <unknown function> + 0x467292 (0x7ccd18325292 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #24: /usr/bin/python3() [0x581e9f]
[rank3]: frame #25: PyObject_Call + 0x9c (0x54b11c in /usr/bin/python3)
[rank3]: frame #26: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #27: _PyObject_Call_Prepend + 0xc2 (0x54a7c2 in /usr/bin/python3)
[rank3]: frame #28: /usr/bin/python3() [0x5a3458]
[rank3]: frame #29: PyObject_Call + 0x9c (0x54b11c in /usr/bin/python3)
[rank3]: frame #30: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #31: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #32: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #33: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #34: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #35: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #36: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #37: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #38: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #39: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #40: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #41: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)
[rank3]: frame #42: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #43: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #44: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #45: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #46: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #47: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #48: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #49: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #50: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #51: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #52: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #53: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #54: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #55: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)
[rank3]: frame #56: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #57: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)
[rank3]: frame #58: /usr/bin/python3() [0x64f4d4]
[rank3]: frame #59: _PyObject_MakeTpCall + 0x75 (0x548f75 in /usr/bin/python3)
[rank3]: frame #60: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #61: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #62: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)

CPU: Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Address sizes: 52 bits physical, 57 bits virtual Byte Order: Little Endian CPU(s): 224 On-line CPU(s) list: 0-223 Vendor ID: GenuineIntel Model name: INTEL(R) XEON(R) PLATINUM 8570 CPU family: 6 Model: 207 Thread(s) per core: 2 Core(s) per socket: 56 Socket(s): 2 Stepping: 2 CPU(s) scaling MHz: 20% CPU max MHz: 4000.0000 CPU min MHz: 800.0000 BogoMIPS: 4200.00 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cat_l2 cdp_l3 intel_ppin cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect user_shstk avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req vnmi avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr ibt amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities ibpb_exit_to_user Virtualization: VT-x L1d cache: 5.3 MiB (112 instances) L1i cache: 3.5 MiB (112 instances) L2 cache: 224 MiB (112 instances) L3 cache: 600 MiB (2 instances) NUMA node(s): 2 NUMA node0 CPU(s): 0-55,112-167 NUMA node1 CPU(s): 56-111,168-223 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling: Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization Vulnerability Spectre v2: Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Mitigation; IBPB before exit to userspace

Code Example

from __future__ import annotations

import torch
import torch.distributed as dist
import torch.distributed._functional_collectives as funcol
from torch._inductor import config as ind_config


def main() -> None:
    dist.init_process_group(backend="nccl")
    rank = dist.get_rank()
    world_size = dist.get_world_size()
    torch.cuda.set_device(rank)
    group = dist.group.WORLD

    aten = ind_config.aten_distributed_optimizations
    aten.enable_overlap_scheduling = True
    aten.collective_bucketing = True
    aten.insert_overlap_deps = True
    aten.collective_estimator = "benchmark"

    HIDDEN = 384
    PROMPT = 32

    @torch.compile(backend="inductor", dynamic=True)
    def f(x: torch.Tensor, prompt: torch.Tensor, w: torch.Tensor) -> torch.Tensor:
        seq = x.shape[0]
        pad = (-seq) % world_size
        x_padded = torch.nn.functional.pad(x, (0, 0, 0, pad))
        combined = torch.cat([prompt, x_padded], dim=0)
        chunk = combined.shape[0] // world_size
        split_sizes = [chunk] * world_size
        gathered = funcol.all_to_all_single(
            combined, split_sizes, split_sizes, group
        )
        return (gathered @ w).sum()

    prompt = torch.randn(PROMPT, HIDDEN, device="cuda")
    w = torch.randn(HIDDEN, HIDDEN, device="cuda")
    for n in (7, 11, 13):
        f(torch.randn(n, HIDDEN, device="cuda"), prompt, w)

    if rank == 0:
        print("OK: no crash (bug not reproduced on this PyTorch)")
    dist.destroy_process_group()


if __name__ == "__main__":
    main()

---

[rank3]: Traceback (most recent call last):
[rank3]:   File "/workspace/repro_benchmark_symint.py", line 62, in <module>
[rank3]:     main()
[rank3]:   File "/workspace/repro_benchmark_symint.py", line 54, in main
[rank3]:     f(torch.randn(n, HIDDEN, device="cuda"), prompt, w)
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_dynamo/eval_frame.py", line 1034, in compile_wrapper
[rank3]:     raise e.remove_dynamo_frames() from None  # see TORCHDYNAMO_VERBOSE=1
[rank3]:     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1051, in _compile_fx_inner
[rank3]:     raise InductorError(e, currentframe()).with_traceback(
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1035, in _compile_fx_inner
[rank3]:     mb_compiled_graph = fx_codegen_and_compile(
[rank3]:                         ^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1796, in fx_codegen_and_compile
[rank3]:     return scheme.codegen_and_compile(gm, example_inputs, inputs_to_check, graph_kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1342, in codegen_and_compile
[rank3]:     _recursive_post_grad_passes(gm, is_inference=is_inference)
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 581, in _recursive_post_grad_passes
[rank3]:     post_grad_passes(gm, is_inference)
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/post_grad.py", line 353, in post_grad_passes
[rank3]:     GraphTransformObserver(gm, "overlap_scheduling").apply_graph_pass(
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/fx/passes/graph_transform_observer.py", line 103, in apply_graph_pass
[rank3]:     return pass_fn(self.gm.graph)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/post_grad.py", line 354, in <lambda>
[rank3]:     lambda graph: schedule_overlap_bucketing_from_inductor_configs(
[rank3]:                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 1720, in schedule_overlap_bucketing_from_inductor_configs
[rank3]:     return schedule_overlap_bucketing(gm, **kwargs)  # type: ignore[arg-type]
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 1669, in schedule_overlap_bucketing
[rank3]:     ).run()
[rank3]:       ^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 936, in run
[rank3]:     self._align_compute_nodes_runtime_estimations_across_all_distributed_ranks()
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 758, in _align_compute_nodes_runtime_estimations_across_all_distributed_ranks
[rank3]:     cuda_val, cuda_key = benchmark_collective_with_cuda_events(n, nruns=5)
[rank3]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/node_runtime_estimation.py", line 196, in benchmark_collective_with_cuda_events
[rank3]:     return benchmark_collective_with_cuda_events_impl(n, nruns)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/node_runtime_estimation.py", line 262, in benchmark_collective_with_cuda_events_impl
[rank3]:     runtime = _benchmark_collective_with_cuda_events_impl(n, args, kwargs, nruns)
[rank3]:               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/node_runtime_estimation.py", line 166, in _benchmark_collective_with_cuda_events_impl
[rank3]:     result = n.target(*args, **kwargs)  # type: ignore[operator]
[rank3]:              ^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 871, in __call__
[rank3]:     return self._op(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_library/autograd.py", line 112, in autograd_impl
[rank3]:     result = forward_no_grad(*args, Metadata(keyset, keyword_only_args))
[rank3]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_library/autograd.py", line 41, in forward_no_grad
[rank3]:     result = op.redispatch(keyset & _C._after_autograd_keyset, *args, **kwargs)
[rank3]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 878, in redispatch
[rank3]:     return self._handle.redispatch_boxed(keyset, *args, **kwargs)  # type: ignore[return-value]
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: torch._inductor.exc.InductorError: RuntimeError: when unpacking SymInt, expected int but got ((s77 + s9 + (PythonMod(-s77, 4)))//4)
[rank3]: Exception raised from expect_int at /opt/pytorch/pytorch/c10/core/SymInt.h:132 (most recent call first):
[rank3]: frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x7cccebffe578 in /usr/local/lib/python3.12/dist-packages/torch/lib/libc10.so)
[rank3]: frame #1: <unknown function> + 0x153a3c5 (0x7ccd0af653c5 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cpu.so)
[rank3]: frame #2: <unknown function> + 0x675e943 (0x7ccd10189943 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cpu.so)
[rank3]: frame #3: <unknown function> + 0x828f6c (0x7ccd186e6f6c in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #4: <unknown function> + 0x821498 (0x7ccd186df498 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #5: <unknown function> + 0xe92864 (0x7ccd18d50864 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #6: <unknown function> + 0xe92c0a (0x7ccd18d50c0a in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #7: <unknown function> + 0x467292 (0x7ccd18325292 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #8: /usr/bin/python3() [0x581e9f]
[rank3]: frame #9: _PyObject_MakeTpCall + 0x75 (0x548f75 in /usr/bin/python3)
[rank3]: frame #10: /usr/bin/python3() [0x54cc29]
[rank3]: frame #11: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #12: /usr/bin/python3() [0x54caad]
[rank3]: frame #13: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #14: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #15: <unknown function> + 0xe99b91 (0x7ccd18d57b91 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #16: <unknown function> + 0xe9a212 (0x7ccd18d58212 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #17: <unknown function> + 0x828f6c (0x7ccd186e6f6c in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #18: <unknown function> + 0x821498 (0x7ccd186df498 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #19: <unknown function> + 0x6442c9a (0x7ccd0fe6dc9a in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cpu.so)
[rank3]: frame #20: <unknown function> + 0xbe378b (0x7ccd18aa178b in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #21: <unknown function> + 0xbe3c34 (0x7ccd18aa1c34 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #22: <unknown function> + 0xac2b99 (0x7ccd18980b99 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #23: <unknown function> + 0x467292 (0x7ccd18325292 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #24: /usr/bin/python3() [0x581e9f]
[rank3]: frame #25: PyObject_Call + 0x9c (0x54b11c in /usr/bin/python3)
[rank3]: frame #26: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #27: _PyObject_Call_Prepend + 0xc2 (0x54a7c2 in /usr/bin/python3)
[rank3]: frame #28: /usr/bin/python3() [0x5a3458]
[rank3]: frame #29: PyObject_Call + 0x9c (0x54b11c in /usr/bin/python3)
[rank3]: frame #30: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #31: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #32: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #33: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #34: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #35: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #36: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #37: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #38: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #39: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #40: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #41: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)
[rank3]: frame #42: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #43: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #44: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #45: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #46: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #47: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #48: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #49: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #50: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #51: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #52: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #53: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #54: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #55: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)
[rank3]: frame #56: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #57: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)
[rank3]: frame #58: /usr/bin/python3() [0x64f4d4]
[rank3]: frame #59: _PyObject_MakeTpCall + 0x75 (0x548f75 in /usr/bin/python3)
[rank3]: frame #60: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #61: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #62: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)
RAW_BUFFERClick to expand / collapse

🐛 Describe the bug

torch._inductor.fx_passes.overlap_scheduling.OverlapScheduler._align_compute_nodes_runtime_estimations_across_all_distributed_ranks calls _benchmark_collective_with_cuda_events_impl to time each functional collective via n.target(*args, **kwargs). With aten_distributed_optimizations.collective_estimator = "benchmark" and dynamic input shapes (e.g. a sequence dim padded to a multiple of world_size, as ulysses-style sequence parallel does), the call crashes inside the C++ dispatcher with expect_int when a scalar SymInt argument reaches an op that takes an int.

Bug originally hit on a diffusion image editing workload with Ulysses degree 8 (Qwen-Image-Edit), while enabling aten_distributed_optimizations flags.

Reproducer

Run with torchrun --nproc-per-node=4 repro_benchmark_symint.py:

from __future__ import annotations

import torch
import torch.distributed as dist
import torch.distributed._functional_collectives as funcol
from torch._inductor import config as ind_config


def main() -> None:
    dist.init_process_group(backend="nccl")
    rank = dist.get_rank()
    world_size = dist.get_world_size()
    torch.cuda.set_device(rank)
    group = dist.group.WORLD

    aten = ind_config.aten_distributed_optimizations
    aten.enable_overlap_scheduling = True
    aten.collective_bucketing = True
    aten.insert_overlap_deps = True
    aten.collective_estimator = "benchmark"

    HIDDEN = 384
    PROMPT = 32

    @torch.compile(backend="inductor", dynamic=True)
    def f(x: torch.Tensor, prompt: torch.Tensor, w: torch.Tensor) -> torch.Tensor:
        seq = x.shape[0]
        pad = (-seq) % world_size
        x_padded = torch.nn.functional.pad(x, (0, 0, 0, pad))
        combined = torch.cat([prompt, x_padded], dim=0)
        chunk = combined.shape[0] // world_size
        split_sizes = [chunk] * world_size
        gathered = funcol.all_to_all_single(
            combined, split_sizes, split_sizes, group
        )
        return (gathered @ w).sum()

    prompt = torch.randn(PROMPT, HIDDEN, device="cuda")
    w = torch.randn(HIDDEN, HIDDEN, device="cuda")
    for n in (7, 11, 13):
        f(torch.randn(n, HIDDEN, device="cuda"), prompt, w)

    if rank == 0:
        print("OK: no crash (bug not reproduced on this PyTorch)")
    dist.destroy_process_group()


if __name__ == "__main__":
    main()

Error logs

<details> <summary>Click to reveal error log (one rank)</summary> 
[rank3]: Traceback (most recent call last):
[rank3]:   File "/workspace/repro_benchmark_symint.py", line 62, in <module>
[rank3]:     main()
[rank3]:   File "/workspace/repro_benchmark_symint.py", line 54, in main
[rank3]:     f(torch.randn(n, HIDDEN, device="cuda"), prompt, w)
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_dynamo/eval_frame.py", line 1034, in compile_wrapper
[rank3]:     raise e.remove_dynamo_frames() from None  # see TORCHDYNAMO_VERBOSE=1
[rank3]:     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1051, in _compile_fx_inner
[rank3]:     raise InductorError(e, currentframe()).with_traceback(
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1035, in _compile_fx_inner
[rank3]:     mb_compiled_graph = fx_codegen_and_compile(
[rank3]:                         ^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1796, in fx_codegen_and_compile
[rank3]:     return scheme.codegen_and_compile(gm, example_inputs, inputs_to_check, graph_kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 1342, in codegen_and_compile
[rank3]:     _recursive_post_grad_passes(gm, is_inference=is_inference)
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py", line 581, in _recursive_post_grad_passes
[rank3]:     post_grad_passes(gm, is_inference)
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/post_grad.py", line 353, in post_grad_passes
[rank3]:     GraphTransformObserver(gm, "overlap_scheduling").apply_graph_pass(
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/fx/passes/graph_transform_observer.py", line 103, in apply_graph_pass
[rank3]:     return pass_fn(self.gm.graph)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/post_grad.py", line 354, in <lambda>
[rank3]:     lambda graph: schedule_overlap_bucketing_from_inductor_configs(
[rank3]:                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 1720, in schedule_overlap_bucketing_from_inductor_configs
[rank3]:     return schedule_overlap_bucketing(gm, **kwargs)  # type: ignore[arg-type]
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 1669, in schedule_overlap_bucketing
[rank3]:     ).run()
[rank3]:       ^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 936, in run
[rank3]:     self._align_compute_nodes_runtime_estimations_across_all_distributed_ranks()
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/overlap_scheduling.py", line 758, in _align_compute_nodes_runtime_estimations_across_all_distributed_ranks
[rank3]:     cuda_val, cuda_key = benchmark_collective_with_cuda_events(n, nruns=5)
[rank3]:                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/node_runtime_estimation.py", line 196, in benchmark_collective_with_cuda_events
[rank3]:     return benchmark_collective_with_cuda_events_impl(n, nruns)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/node_runtime_estimation.py", line 262, in benchmark_collective_with_cuda_events_impl
[rank3]:     runtime = _benchmark_collective_with_cuda_events_impl(n, args, kwargs, nruns)
[rank3]:               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_inductor/fx_passes/node_runtime_estimation.py", line 166, in _benchmark_collective_with_cuda_events_impl
[rank3]:     result = n.target(*args, **kwargs)  # type: ignore[operator]
[rank3]:              ^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 871, in __call__
[rank3]:     return self._op(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_library/autograd.py", line 112, in autograd_impl
[rank3]:     result = forward_no_grad(*args, Metadata(keyset, keyword_only_args))
[rank3]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_library/autograd.py", line 41, in forward_no_grad
[rank3]:     result = op.redispatch(keyset & _C._after_autograd_keyset, *args, **kwargs)
[rank3]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 878, in redispatch
[rank3]:     return self._handle.redispatch_boxed(keyset, *args, **kwargs)  # type: ignore[return-value]
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: torch._inductor.exc.InductorError: RuntimeError: when unpacking SymInt, expected int but got ((s77 + s9 + (PythonMod(-s77, 4)))//4)
[rank3]: Exception raised from expect_int at /opt/pytorch/pytorch/c10/core/SymInt.h:132 (most recent call first):
[rank3]: frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x7cccebffe578 in /usr/local/lib/python3.12/dist-packages/torch/lib/libc10.so)
[rank3]: frame #1: <unknown function> + 0x153a3c5 (0x7ccd0af653c5 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cpu.so)
[rank3]: frame #2: <unknown function> + 0x675e943 (0x7ccd10189943 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cpu.so)
[rank3]: frame #3: <unknown function> + 0x828f6c (0x7ccd186e6f6c in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #4: <unknown function> + 0x821498 (0x7ccd186df498 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #5: <unknown function> + 0xe92864 (0x7ccd18d50864 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #6: <unknown function> + 0xe92c0a (0x7ccd18d50c0a in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #7: <unknown function> + 0x467292 (0x7ccd18325292 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #8: /usr/bin/python3() [0x581e9f]
[rank3]: frame #9: _PyObject_MakeTpCall + 0x75 (0x548f75 in /usr/bin/python3)
[rank3]: frame #10: /usr/bin/python3() [0x54cc29]
[rank3]: frame #11: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #12: /usr/bin/python3() [0x54caad]
[rank3]: frame #13: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #14: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #15: <unknown function> + 0xe99b91 (0x7ccd18d57b91 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #16: <unknown function> + 0xe9a212 (0x7ccd18d58212 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #17: <unknown function> + 0x828f6c (0x7ccd186e6f6c in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #18: <unknown function> + 0x821498 (0x7ccd186df498 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #19: <unknown function> + 0x6442c9a (0x7ccd0fe6dc9a in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cpu.so)
[rank3]: frame #20: <unknown function> + 0xbe378b (0x7ccd18aa178b in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #21: <unknown function> + 0xbe3c34 (0x7ccd18aa1c34 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #22: <unknown function> + 0xac2b99 (0x7ccd18980b99 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #23: <unknown function> + 0x467292 (0x7ccd18325292 in /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_python.so)
[rank3]: frame #24: /usr/bin/python3() [0x581e9f]
[rank3]: frame #25: PyObject_Call + 0x9c (0x54b11c in /usr/bin/python3)
[rank3]: frame #26: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #27: _PyObject_Call_Prepend + 0xc2 (0x54a7c2 in /usr/bin/python3)
[rank3]: frame #28: /usr/bin/python3() [0x5a3458]
[rank3]: frame #29: PyObject_Call + 0x9c (0x54b11c in /usr/bin/python3)
[rank3]: frame #30: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #31: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #32: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #33: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #34: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #35: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #36: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #37: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #38: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #39: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #40: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #41: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)
[rank3]: frame #42: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #43: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #44: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #45: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #46: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #47: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #48: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #49: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #50: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #51: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #52: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #53: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #54: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #55: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)
[rank3]: frame #56: _PyEval_EvalFrameDefault + 0x4c3a (0x5db0ba in /usr/bin/python3)
[rank3]: frame #57: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)
[rank3]: frame #58: /usr/bin/python3() [0x64f4d4]
[rank3]: frame #59: _PyObject_MakeTpCall + 0x75 (0x548f75 in /usr/bin/python3)
[rank3]: frame #60: PyObject_Vectorcall + 0x35 (0x549975 in /usr/bin/python3)
[rank3]: frame #61: _PyEval_EvalFrameDefault + 0xa89 (0x5d6f09 in /usr/bin/python3)
[rank3]: frame #62: PyObject_Call + 0x119 (0x54b199 in /usr/bin/python3)
</details>

Other ranks produce similar error tracebacks.

Versions

Collecting environment information... PyTorch version: 2.12.0a0+0291f960b6.nv26.04.48445190 Is debug build: False CUDA used to build PyTorch: 13.2 ROCM used to build PyTorch: N/A

OS: Ubuntu 24.04.4 LTS (x86_64) GCC version: (Ubuntu 13.3.0-6ubuntu2~24.04.1) 13.3.0 Clang version: Could not collect CMake version: version 3.31.6 Libc version: glibc-2.39

Python version: 3.12.3 (main, Mar 3 2026, 12:15:18) [GCC 13.3.0] (64-bit runtime) Python platform: Linux-6.8.0-106-generic-x86_64-with-glibc2.39 Is CUDA available: True CUDA runtime version: 13.2.78 CUDA_MODULE_LOADING set to: LAZY GPU models and configuration: GPU 0: NVIDIA B200 GPU 1: NVIDIA B200 GPU 2: NVIDIA B200 GPU 3: NVIDIA B200 GPU 4: NVIDIA B200 GPU 5: NVIDIA B200 GPU 6: NVIDIA B200 GPU 7: NVIDIA B200

Nvidia driver version: 580.159.03 cuDNN version: Probably one of the following: /usr/lib/x86_64-linux-gnu/libcudnn.so.9.21.0 /usr/lib/x86_64-linux-gnu/libcudnn_adv.so.9.21.0 /usr/lib/x86_64-linux-gnu/libcudnn_cnn.so.9.21.0 /usr/lib/x86_64-linux-gnu/libcudnn_engines_precompiled.so.9.21.0 /usr/lib/x86_64-linux-gnu/libcudnn_engines_runtime_compiled.so.9.21.0 /usr/lib/x86_64-linux-gnu/libcudnn_engines_tensor_ir.so.9.21.0 /usr/lib/x86_64-linux-gnu/libcudnn_graph.so.9.21.0 /usr/lib/x86_64-linux-gnu/libcudnn_heuristic.so.9.21.0 /usr/lib/x86_64-linux-gnu/libcudnn_ops.so.9.21.0 Is XPU available: False HIP runtime version: N/A MIOpen runtime version: N/A Is XNNPACK available: True Caching allocator config: N/A

CPU: Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Address sizes: 52 bits physical, 57 bits virtual Byte Order: Little Endian CPU(s): 224 On-line CPU(s) list: 0-223 Vendor ID: GenuineIntel Model name: INTEL(R) XEON(R) PLATINUM 8570 CPU family: 6 Model: 207 Thread(s) per core: 2 Core(s) per socket: 56 Socket(s): 2 Stepping: 2 CPU(s) scaling MHz: 20% CPU max MHz: 4000.0000 CPU min MHz: 800.0000 BogoMIPS: 4200.00 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cat_l2 cdp_l3 intel_ppin cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect user_shstk avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req vnmi avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr ibt amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities ibpb_exit_to_user Virtualization: VT-x L1d cache: 5.3 MiB (112 instances) L1i cache: 3.5 MiB (112 instances) L2 cache: 224 MiB (112 instances) L3 cache: 600 MiB (2 instances) NUMA node(s): 2 NUMA node0 CPU(s): 0-55,112-167 NUMA node1 CPU(s): 56-111,168-223 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling: Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization Vulnerability Spectre v2: Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Mitigation; IBPB before exit to userspace

Versions of relevant libraries: [pip3] intel-openmp==2021.4.0 [pip3] mkl==2021.1.1 [pip3] mkl-devel==2021.1.1 [pip3] mkl-include==2021.1.1 [pip3] mypy_extensions==1.1.0 [pip3] numpy==1.26.4 [pip3] nvidia-cudnn-frontend==1.22.1 [pip3] nvtx==0.2.15 [pip3] onnx==1.21.0 [pip3] onnx-ir==0.2.0 [pip3] onnxscript==0.6.2 [pip3] optree==0.19.0 [pip3] tbb==2021.13.1 [pip3] torch==2.12.0a0+0291f960b6.nv26.4.48445190 [pip3] torch_c_dlpack_ext==0.1.5 [pip3] torch_tensorrt==2.12.0a0 [pip3] torchao==0.17.0+git42bcdc49 [pip3] torchdata==0.11.0 [pip3] torchtitan==0.2.2+git6d1ff9e6 [pip3] torchvision==0.26.0a0+48956e05.nv26.4.48445190 [pip3] triton==3.6.0+git5d72932fc5.nv26.4 [pip3] triton_kernels==1.0.0+git5d72932fc5.nv26.4 [conda] Could not collect

cc @chauhang @penguinwu @ezyang @bobrenjc93 @aditvenk @laithsakka @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @kadeng @muchulee8 @amjames @aakhundov @coconutruben @jataylo

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING

pytorch - 💡(How to fix) Fix [Inductor] _benchmark_collective_with_cuda_events_impl aborts on scalar SymInt kwargs to functional collectives (dynamic shapes + collective_estimator="benchmark")