8000 Hang in iree-run-module on ONNX dequantizelinear test case · Issue #16666 · iree-org/iree · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Hang in iree-run-module on ONNX dequantizelinear test case #16666

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.

Sign up for GitHub

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

Jump to bottom
Closed
ScottTodd opened this issue Mar 5, 2024 · 6 comments
Closed

Hang in iree-run-module on ONNX dequantizelinear test case #16666

ScottTodd opened this issue Mar 5, 2024 · 6 comments
Labels
bug 🐞 Something isn't working integrations/onnx ONNX integration work integrations/pytorch PyTorch integration work

Comments

@ScottTodd
Copy link
Member

I'm seeing iree-run-module hang indefinitely with this call stack:

[Inline Frame] iree-run-module.exe!iree_futex_wait(void * expected_value, unsigned int) Line 148 (d:\dev\projects\iree\runtime\src\iree\base\internal\synchronization.c:148)
iree-run-module.exe!iree_notification_commit_wait(iree_notification_t * notification, unsigned int wait_token, __int64 spin_ns, __int64 deadline_ns) Line 819 (d:\dev\projects\iree\runtime\src\iree\base\internal\synchronization.c:819)
iree-run-module.exe!iree_notification_await(iree_notification_t * notification, bool(*)(void *) condition_fn, void * condition_arg, iree_timeout_t timeout) Line 875 (d:\dev\projects\iree\runtime\src\iree\base\internal\synchronization.c:875)
iree-run-module.exe!iree_hal_sync_semaphore_wait(iree_hal_semaphore_t * base_semaphore, unsigned __int64 value, iree_timeout_t timeout) Line 303 (d:\dev\projects\iree\runtime\src\iree\hal\drivers\local_sync\sync_semaphore.c:303)
iree-run-module.exe!iree_hal_semaphore_wait(iree_hal_semaphore_t * semaphore, unsigned __int64 value, iree_timeout_t timeout) Line 78 (d:\dev\projects\iree\runtime\src\iree\hal\semaphore.c:78)
iree-run-module.exe!iree_hal_semaphore_list_wait(iree_hal_semaphore_list_t semaphore_list, iree_timeout_t timeout) Line 201 (d:\dev\projects\iree\runtime\src\iree\hal\semaphore.c:201)
iree-run-module.exe!iree_hal_fence_wait(iree_hal_fence_t * fence, iree_timeout_t timeout) Line 253 (d:\dev\projects\iree\runtime\src\iree\hal\fence.c:253)
iree-run-module.exe!iree_hal_module_fence_await(iree_vm_stack_t * stack, void * module, iree_hal_module_state_t * state, iree_vm_abi_iCrD_t * args, iree_vm_abi_i_t * rets) Line 1521 (d:\dev\projects\iree\runtime\src\iree\modules\hal\module.c:1521)
iree-run-module.exe!iree_vm_shim_iCrD_i(iree_vm_stack_t * stack, unsigned int flags, iree_byte_span_t args_storage, iree_byte_span_t rets_storage, iree_status_handle_t *(*)(iree_vm_stack_t *, void *, void *, const void *, void *) target_fn, void * module, void * module_state) Line 76 (d:\dev\projects\iree\runtime\src\iree\vm\shims.c:76)
iree-run-module.exe!iree_vm_native_module_issue_call(iree_vm_native_module_t * module, iree_vm_stack_t * stack, iree_vm_stack_frame_t * callee_frame, unsigned int flags, iree_byte_span_t args_storage, iree_byte_span_t rets_storage) Line 342 (d:\dev\projects\iree\runtime\src\iree\vm\native_module.c:342)
iree-run-module.exe!iree_vm_native_module_begin_call(void * self, iree_vm_stack_t * stack, iree_vm_function_call_t call) Line 394 (d:\dev\projects\iree\runtime\src\iree\vm\native_module.c:394)
iree-run-module.exe!iree_vm_bytecode_issue_import_call(iree_vm_stack_t * stack, const iree_vm_function_call_t call, iree_string_view_t cconv_results, const iree_vm_register_list_t * dst_reg_list, iree_vm_stack_frame_t * * out_caller_frame, iree_vm_registers_t * out_caller_registers) Line 453 (d:\dev\projects\iree\runtime\src\iree\vm\bytecode\dispatch.c:453)
iree-run-module.exe!iree_vm_bytecode_call_import_variadic(iree_vm_stack_t * stack, const iree_vm_bytecode_module_state_t * module_state, unsigned int import_ordinal, const iree_vm_registers_t caller_registers, const iree_vm_register_list_t * segment_size_list, const iree_vm_register_list_t * src_reg_list, const iree_vm_register_list_t * dst_reg_list, iree_vm_stack_frame_t * * out_caller_frame, iree_vm_registers_t * out_caller_registers) Line 612 (d:\dev\projects\iree\runtime\src\iree\vm\bytecode\dispatch.c:612)
iree-run-module.exe!iree_vm_bytecode_dispatch(iree_vm_stack_t * stack, iree_vm_bytecode_module_t * module, iree_vm_stack_frame_t * current_frame, iree_vm_registers_t regs, iree_byte_span_t call_results) Line 1716 (d:\dev\projects\iree\runtime\src\iree\vm\bytecode\dispatch.c:1716)
iree-run-module.exe!iree_vm_bytecode_dispatch_begin(iree_vm_stack_t * stack, iree_vm_bytecode_module_t * module, const iree_vm_function_call_t call, iree_string_view_t cconv_arguments, iree_string_view_t cconv_results) Line 638 (d:\dev\projects\iree\runtime\src\iree\vm\bytecode\dispatch.c:638)

Original test case: https://github.com/nod-ai/SHARK-TestSuite/tree/main/iree_tests/onnx/node/generated/test_dequantizelinear

Input MLIR:

module {
  func.func @test_dequantizelinear(%arg0: !torch.vtensor<[4],ui8>, %arg1: !torch.vtensor<[],f32>, %arg2: !torch.vtensor<[],ui8>) -> !torch.vtensor<[4],f32> attributes {torch.onnx_meta.ir_version = 10 : si64, torch.onnx_meta.opset_version = 21 : si64, torch.onnx_meta.producer_name = "backend-test", torch.onnx_meta.producer_version = ""} {
    %none = torch.constant.none
    %0 = torch.operator "onnx.DequantizeLinear"(%arg0, %arg1, %arg2) : (!torch.vtensor<[4],ui8>, !torch.vtensor<[],f32>, !torch.vtensor<[],ui8>) -> !torch.vtensor<[4],f32> 
    return %0 : !torch.vtensor<[4],f32>
  }
}

After iree-compile --compile-to=input dequantize_linear.mlir:

module attributes {hal.device.targets = [#hal.device.target<"llvm-cpu", [#hal.executable.target<"llvm-cpu", "embedded-elf-x86_64", {cpu = "generic", cpu_features = "", data_layout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128", native_vector_size = 16 : i64, target_triple = "x86_64-unknown-unknown-eabi-elf"}>]>]} {
  util.func public @test_dequantizelinear$async(%arg0: !hal.buffer_view, %arg1: !hal.buffer_view, %arg2: !hal.buffer_view, %arg3: !hal.fence, %arg4: !hal.fence) -> !hal.buffer_view attributes {iree.abi.model = "coarse-fences", iree.abi.stub} {
    %0 = hal.tensor.import wait(%arg3) => %arg0 : !hal.buffer_view -> tensor<4xi8>
    %1 = hal.tensor.import wait(%arg3) => %arg1 : !hal.buffer_view -> tensor<f32>
    %2 = hal.tensor.import wait(%arg3) => %arg2 : !hal.buffer_view -> tensor<i8>
    %extracted = tensor.extract %1[] : tensor<f32>
    %3 = arith.extf %extracted : f32 to f64
    %extracted_0 = tensor.extract %2[] : tensor<i8>
    %4 = arith.extsi %extracted_0 : i8 to i64
    %5 = tensor.empty() : tensor<4xf32>
    %6 = linalg.generic {indexing_maps = [affine_map<(d0) -> (d0)>, affine_map<(d0) -> (d0)>], iterator_types = ["parallel"]} ins(%0 : tensor<4xi8>) outs(%5 : tensor<4xf32>) {
    ^bb0(%in: i8, %out: f32):
      %9 = arith.extui %in : i8 to i32
      %10 = arith.trunci %4 : i64 to i32
      %11 = arith.subi %9, %10 : i32
      %12 = arith.uitofp %11 : i32 to f32
      %13 = arith.truncf %3 : f64 to f32
      %14 = arith.mulf %12, %13 : f32
      linalg.yield %14 : f32
    } -> tensor<4xf32>
    %7 = hal.tensor.barrier join(%6 : tensor<4xf32>) => %arg4 : !hal.fence
    %8 = hal.tensor.export %7 : tensor<4xf32> -> !hal.buffer_view
    util.return %8 : !hal.buffer_view
  }
  util.func public @test_dequantizelinear(%arg0: !hal.buffer_view, %arg1: !hal.buffer_view, %arg2: !hal.buffer_view) -> !hal.buffer_view attributes {iree.abi.stub} {
    %c-1_i32 = arith.constant -1 : i32
    %c0 = arith.constant 0 : index
    %device_0 = hal.devices.get %c0 : !hal.device
    %0 = util.null : !hal.fence
    %fence = hal.fence.create device(%device_0 : !hal.device) flags("None") : !hal.fence
    %1 = util.call @test_dequantizelinear$async(%arg0, %arg1, %arg2, %0, %fence) : (!hal.buffer_view, !hal.buffer_view, !hal.buffer_view, !hal.fence, !hal.fence) -> !hal.buffer_view
    %status = hal.fence.await until([%fence]) timeout_millis(%c-1_i32) : i32
    util.return %1 : !hal.buffer_view
  }
}

--print-ir-after-all output: https://gist.github.com/ScottTodd/15ea43eae8733806b80d7d9ed1fb4f1a

Run command (using the input .npy files from the test case folder):

iree-run-module \
  --device=local-sync
  --module=dequantize_linear_cpu.vmfb \
  --input=@/path/to/input_0.npy \
  --input=@/path/to/input_1.npy \
  --input=@/path/to/input_2.npy

Running those .npy files through iree\tools\test\echo_npy.py produces

[    0.     2.     3.  1000.  -254. -1000.]
2.0
128

(first is type ui8 though, so why negative?)
@ScottTodd ScottTodd added bug 🐞 Something isn't working integrations/pytorch PyTorch integration work labels Mar 5, 2024
@benvanik
Copy link
Collaborator
benvanik commented Mar 5, 2024

try --trace_execution?

@ScottTodd
Copy link
Member Author

try --trace_execution?

[module.__init+00000000]    <block>
[module.__init+00000001]    %r0 = vm.const.ref.zero
[module.__init+00000004]    %i0 = vm.const.i32 2  // 0x00000002
[module.__init+0000000B]    %i1 = vm.const.i32 7  // 0x00000007
[module.__init+00000012]    %i2 = vm.const.i32 1  // 0x00000001
[module.__init+00000019]    %i3 = vm.const.i32 14  // 0x0000000E
[module.__init+00000020]    %i4:5 = vm.const.i64 -1  // 0xFFFFFFFFFFFFFFFF
[module.__init+0000002B]    %i6:7 = vm.const.i64.zero
[module.__init+0000002E]    %i8 = vm.const.i32.zero
[module.__init+00000031]    %r1 = vm.call @hal.devices.get(%i8(0))
[module.__init+0000003E]    %r2 = vm.const.ref.rodata 1  // 0x0000019750ECD048 21b
[module.__init+00000045]    %r3 = vm.const.ref.rodata 2  // 0x0000019750ECD06C 19b
[module.__init+0000004C]    %i9, %i10 = vm.call @hal.device.query.i64(%r1(!hal.device/0x0000019750EAC9C0), %r2(!vm.buffer/0x0000019750ECB690), %r3(!vm.buffer/0x0000019750ECB6B8))
[module.__init+00000060]    %i10 = vm.trunc.i64.i32 %i10:11(1)
[module.__init+00000065]    %i10 = vm.and.i32 %i10(1), %i2(1)
[module.__init+0000006C]    %i9 = vm.select.i32 %i9(1) ? %i10(1) : %i8(0)
[module.__init+00000075]    %r2 = vm.call.varadic @hal.descriptor_set_layout.create(%r1(!hal.device/0x0000019750EAC9C0), %i8(0), %i8(0), %i1(7), %i2(1), %i2(1), %i1(7), %i8(0))
[module.__init+00000098]    %r2 = vm.call.varadic @hal.pipeline_layout.create(%r1(!hal.device/0x0000019750EAC9C0), %i0(2), %r2(!hal.descriptor_set_layout/0x0000019750EBA180))
[module.__init+000000B2]    %i0:1 = vm.select.i64 %i9(1) ? %i6:7(0) : %i4:5(-1)
[module.__init+000000BB]    %i0 = vm.cmp.eq.i64 %i0:1(0), %i6:7(0)
[module.__init+000000C2]    vm.global.store.i32 %i9(1), .rwdata[0]
[module.__init+000000C9]    vm.global.store.ref %r2(!hal.pipeline_layout/0x0000019750EBABD0), .refs[0] : !hal.pipeline_layout
[module.__init+000000D4]    vm.cond_br %i0(1), ^000000E4(), ^00000126()
[module.__init+000000E5]    %r2 = vm.global.load.ref .refs[0](!hal.pipeline_layout/0x0000019750EBABD0) : !hal.pipeline_layout
[module.__init+000000F0]    %r4 = vm.const.ref.rodata 0  // 0x0000019750ECDBC0 3840b
[module.__init+000000F7]    %r0 = vm.call.varadic @hal.executable.create(%r1(!hal.device/0x0000019750EAC9C0), %r3(!vm.buffer/0x0000019750ECB6B8), %r4(!vm.buffer/0x0000019750ECB668), %r0(null), %r2(!hal.pipeline_layout/0x0000019750EBABD0))
[module.__init+00000118]    vm.global.store.ref %r0(!hal.executable/0x0000019750ED4630), .refs[1] : !hal.executable
[module.__init+00000123]    vm.return
EXEC @test_dequantizelinear
[module.test_dequantizelinear+00000000]    <block>
[module.test_dequantizelinear+00000001]    %i0 = vm.const.i32 3  // 0x00000003
[module.test_dequantizelinear+00000008]    %i1 = vm.const.i32 48  // 0x00000030
[module.test_dequantizelinear+0000000F]    %i2 = vm.const.i32 4  // 0x00000004
[module.test_dequantizelinear+00000016]    %i3 = vm.const.i32 13  // 0x0000000D
[module.test_dequantizelinear+0000001D]    %i4 = vm.const.i32 28  // 0x0000001C
[module.test_dequantizelinear+00000024]    %i5 = vm.const.i32 150998019  // 0x09000C03
[module.test_dequantizelinear+0000002B]    %i6 = vm.const.i32 86  // 0x00000056
[module.test_dequantizelinear+00000032]    %i7 = vm.const.i32 553648160  // 0x21000020
[module.test_dequantizelinear+00000039]    %i8 = vm.const.i32 3075  // 0x00000C03
[module.test_dequantizelinear+00000040]    %i9 = vm.const.i32 16  // 0x00000010
[module.test_dequantizelinear+00000047]    %i10 = vm.const.i32 1  // 0x00000001
[module.test_dequantizelinear+0000004E]    %i11 = vm.const.i32 268435464  // 0x10000008
[module.test_dequantizelinear+00000055]    %r3 = vm.const.ref.zero
[module.test_dequantizelinear+00000058]    %i12 = vm.const.i32.zero
[module.test_dequantizelinear+0000005B]    %i13 = vm.const.i32 -1  // 0xFFFFFFFF
[module.test_dequantizelinear+00000062]    %i14:15 = vm.const.i64 4  // 0x0000000000000004
[module.test_dequantizelinear+0000006D]    %i16:17 = vm.const.i64 1  // 0x0000000000000001
[module.test_dequantizelinear+00000078]    %i18:19 = vm.const.i64 16  // 0x0000000000000010
[module.test_dequantizelinear+00000083]    %i20:21 = vm.const.i64.zero
[module.test_dequantizelinear+00000086]    %i22:23 = vm.const.i64 64  // 0x0000000000000040
[module.test_dequantizelinear+00000091]    %i24:25 = vm.const.i64 128  // 0x0000000000000080
[module.test_dequantizelinear+0000009C]    %i26:27 = vm.const.i64 -1  // 0xFFFFFFFFFFFFFFFF
[module.test_dequantizelinear+000000A7]    %i28 = vm.global.load.i32 .rwdata[0](1)
[module.test_dequantizelinear+000000AE]    %r4 = vm.global.load.ref .refs[0](!hal.pipeline_layout/0x0000019750EBABD0) : !hal.pipeline_layout
[module.test_dequantizelinear+000000B9]    %r5 = vm.global.load.ref .refs[1](!hal.executable/0x0000019750ED4630) : !hal.executable
[module.test_dequantizelinear+000000C4]    %r6 = vm.call @hal.devices.get(%i12(0))
[module.test_dequantizelinear+000000D2]    %r7 = vm.call @hal.fence.create(%r6(!hal.device/0x0000019750EAC9C0), %i12(0))
[module.test_dequantizelinear+000000E2]    %r8 = vm.const.ref.rodata 3  // 0x0000019750ECD08C 6b
[module.test_dequantizelinear+000000E9]    vm.call.varadic @hal.buffer_view.assert(%r0(!hal.buffer_view/0x0000019750EBA590), %r8(!vm.buffer/0x0000019750ECB6E0), %i11(268435464), %i10(1), %i14(4))
[module.test_dequantizelinear+00000108]    %r0 = vm.call @hal.buffer_view.buffer(%r0(!hal.buffer_view/0x0000019750EBA590))
[module.test_dequantizelinear+00000116]    %r9 = vm.call @hal.device.allocator(%r6(!hal.device/0x0000019750EAC9C0))
[module.test_dequantizelinear+00000124]    vm.call @hal.buffer.assert(%r0(!hal.buffer/0x0000019750EC6380), %r8(!vm.buffer/0x0000019750ECB6E0), %r9(!hal.allocator/0x0000019750ED55D0), %i14(4), %i9(16), %i8(3075))
[module.test_dequantizelinear+0000013A]    vm.call.varadic @hal.buffer_view.assert(%r1(!hal.buffer_view/0x0000019750F3AA10), %r8(!vm.buffer/0x0000019750ECB6E0), %i7(553648160), %i10(1))
[module.test_dequantizelinear+00000158]    %r1 = vm.call @hal.buffer_view.buffer(%r1(!hal.buffer_view/0x0000019750F3AA10))
[module.test_dequantizelinear+00000166]    vm.call @hal.buffer.assert(%r1(!hal.buffer/0x0000019750EC6280), %r8(!vm.buffer/0x0000019750ECB6E0), %r9(!hal.allocator/0x0000019750ED55D0), %i14(4), %i9(16), %i8(3075))
[module.test_dequantizelinear+0000017C]    vm.call.varadic @hal.buffer_view.assert(%r2(!hal.buffer_view/0x0000019750F3A550), %r8(!vm.buffer/0x0000019750ECB6E0), %i11(268435464), %i10(1))
[module.test_dequantizelinear+0000019A]    %r2 = vm.call @hal.buffer_view.buffer(%r2(!hal.buffer_view/0x0000019750F3A550))
[module.test_dequantizelinear+000001A8]    vm.call @hal.buffer.assert(%r2(!hal.buffer/0x0000019750EC5B80), %r8(!vm.buffer/0x0000019750ECB6E0), %r9(!hal.allocator/0x0000019750ED55D0), %i16(1), %i9(16), %i8(3075))
[module.test_dequantizelinear+000001BE]    %r8 = vm.call @hal.fence.create(%r6(!hal.device/0x0000019750EAC9C0), %i12(0))
[module.test_dequantizelinear+000001CE]    %r9 = vm.call @hal.device.queue.alloca(%r6(!hal.device/0x0000019750EAC9C0), %i26(4294967295), %r3(null), %r8(!hal.fence/0x0000019750F3A690), %i12(0), %i6(86), %i5(150998019), %i24(128))
[module.test_dequantizelinear+000001EA]    %r10 = vm.call @hal.command_buffer.create(%r6(!hal.device/0x0000019750EAC9C0), %i10(1), %i10(1), %i12(0))
[module.test_dequantizelinear+000001FE]    vm.call @hal.command_buffer.copy_buffer(%r10(!hal.command_buffer/0x0000019750F02B50), %r1(!hal.buffer/0x0000019750EC6280), %i20(0), %r9(!hal.buffer/0x0000019750F08440), %i20(0), %i14(4))
[module.test_dequantizelinear+00000214]    vm.call @hal.command_buffer.copy_buffer(%r10(!hal.command_buffer/0x0000019750F02B50), %r2(!hal.buffer/0x0000019750EC5B80), %i20(0), %r9(!hal.buffer/0x0000019750F08440), %i22(64), %i16(1))
[module.test_dequantizelinear+0000022A]    vm.call @hal.command_buffer.execution_barrier(%r10(!hal.command_buffer/0x0000019750F02B50), %i4(28), %i3(13), %i12(0))
[module.test_dequantizelinear+0000023C]    vm.call @hal.command_buffer.finalize(%r10(!hal.command_buffer/0x0000019750F02B50))
[module.test_dequantizelinear+00000248]    %r1 = vm.call @hal.fence.create(%r6(!hal.device/0x0000019750EAC9C0), %i12(0))
[module.test_dequantizelinear+00000258]    vm.call.varadic @hal.device.queue.execute(%r6(!hal.device/0x0000019750EAC9C0), %i26(4294967295), %r8(!hal.fence/0x0000019750F3A690), %r1(!hal.fence/0x0000019750F3A1D0), %r10(!hal.command_buffer/0x0000019750F02B50))
[module.test_dequantizelinear+00000278]    %i5 = vm.call.varadic @hal.fence.await(%i13(4294967295), %r1(!hal.fence/0x0000019750F3A1D0))
[module.test_dequantizelinear+0000028E]    vm.cond_br %i5(0), ^00000348(), ^0000029E()
[module.test_dequantizelinear+0000029F]    %i5 = vm.call @hal.buffer.load(%r9(!hal.buffer/0x0000019750F08440), %i22(64), %i10(1))
[module.test_dequantizelinear+000002B0]    %i6 = vm.call.varadic @hal.fence.await(%i13(4294967295), %r7(!hal.fence/0x0000019750F39D90))

@benvanik
Copy link
Collaborator
benvanik commented Mar 5, 2024

feels like a miscompile to me - I have no idea what's going on there (copies and then a load and such)

@ScottTodd
Copy link
Member Author

That was my feeling too, or a divergent / infinite loop based on the input values (wondering about the ui8 inputs).

@ScottTodd ScottTodd mentioned this issue Mar 7, 2024
Add timeouts to pytest cases in pkgci/iree_tests. #16703
Merged
ScottTodd added a commit that referenced this issue Mar 7, 2024
@ScottTodd
Add timeouts to pytest cases in pkgci/iree_tests. (#16703)
f513fe2 
When a test times out the error messages are basically useless (for the
entire run, not just the specific test cases), since
https://pypi.org/project/pytest-timeout/ terminates the test rather
forcefully... but timing out is still better than running forever. For
the test cases in #16666 that
timed out, I just excluded them from the test job configs.
@ScottTodd
Copy link
Member Author

I no longer see these hangs locally. Weird, since I want to debug an unrelated hang and thought these could be helpful with debugging :/

This was referenced May 10, 2024
Closed
Closed
@ScottTodd
Copy link
Member Author

Starting to narrow down the poor failure mode here. Seems like pytest-timeout and pytest-retry are (sometimes?) incompatible with one another.

Still not sure why the tests are actually hanging though.

@ScottTodd ScottTodd mentioned this issue May 13, 2024
Merged
ScottTodd added a commit that referenced this issue May 13, 2024
@ScottTodd
Drop --retries=2 from pytest to fix --timeout behavior. (#17384)
07d6508 
Progress on #16666.

The [pytest-retry](https://pypi.org/project/pytest-retry/) and
[pytest-timeout](https://pypi.org/project/pytest-timeout/) plugins
initially appeared to be compatible. However, we've been observing tests
hanging without ever tripping the timeout. Hangs under
[pytest-xdist](https://pypi.org/project/pytest-xdist/) don't log
information about individual tests until after the full suite finishes,
making such failures impossible to debug directly ([sample
logs](https://github.com/iree-org/iree/actions/runs/9038964717/job/24841208573#step:9:43)).
Conversely, test failure flakes that would benefit from a retry show
which test failed and can be attempted again from the GitHub UI.

Tests that hang still need to be added to `skip_run_tests` instead of
`expected_run_failures`, since our conftest.py logic is only looking for
failures in `iree-compile` or `iree-run-module`, not timeout signals
from the test runner. See [these sample
logs](https://github.com/iree-org/iree/actions/runs/9069578568/job/24919962287#step:16:4699)
for what a timeout failure now looks like.

ci-exactly:
build_packages,regression_test_amdgpu_rocm,regression_test_cpu,regression_test_nvidiagpu_cuda,regression_test_nvidiagpu_vulkan,test_tensorflow_cpu
bangtianliu pushed a commit to bangtianliu/iree that referenced this issue Jun 5, 2024
@ScottTodd @bangtianliu
Drop --retries=2 from pytest to fix --timeout behavior. (iree-org…
554f488 
…#17384)

Progress on iree-org#16666.

The [pytest-retry](https://pypi.org/project/pytest-retry/) and
[pytest-timeout](https://pypi.org/project/pytest-timeout/) plugins
initially appeared to be compatible. However, we've been observing tests
hanging without ever tripping the timeout. Hangs under
[pytest-xdist](https://pypi.org/project/pytest-xdist/) don't log
information about individual tests until after the full suite finishes,
making such failures impossible to debug directly ([sample
logs](https://github.com/iree-org/iree/actions/runs/9038964717/job/24841208573#step:9:43)).
Conversely, test failure flakes that would benefit from a retry show
which test failed and can be attempted again from the GitHub UI.

Tests that hang still need to be added to `skip_run_tests` instead of
`expected_run_failures`, since our conftest.py logic is only looking for
failures in `iree-compile` or `iree-run-module`, not timeout signals
from the test runner. See [these sample
logs](https://github.com/iree-org/iree/actions/runs/9069578568/job/24919962287#step:16:4699)
for what a timeout failure now looks like.

ci-exactly:
build_packages,regression_test_amdgpu_rocm,regression_test_cpu,regression_test_nvidiagpu_cuda,regression_test_nvidiagpu_vulkan,test_tensorflow_cpu
@ScottTodd ScottTodd mentioned this issue Jun 6, 2024
Open
LLITCHEV pushed a commit to LLITCHEV/iree th B370 at referenced this issue Jul 30, 2024
@ScottTodd @LLITCHEV
Drop --retries=2 from pytest to fix --timeout behavior. (iree-org…
ae75534 
…#17384)

Progress on iree-org#16666.

The [pytest-retry](https://pypi.org/project/pytest-retry/) and
[pytest-timeout](https://pypi.org/project/pytest-timeout/) plugins
initially appeared to be compatible. However, we've been observing tests
hanging without ever tripping the timeout. Hangs under
[pytest-xdist](https://pypi.org/project/pytest-xdist/) don't log
information about individual tests until after the full suite finishes,
making such failures impossible to debug directly ([sample
logs](https://github.com/iree-org/iree/actions/runs/9038964717/job/24841208573#step:9:43)).
Conversely, test failure flakes that would benefit from a retry show
which test failed and can be attempted again from the GitHub UI.

Tests that hang still need to be added to `skip_run_tests` instead of
`expected_run_failures`, since our conftest.py logic is only looking for
failures in `iree-compile` or `iree-run-module`, not timeout signals
from the test runner. See [these sample
logs](https://github.com/iree-org/iree/actions/runs/9069578568/job/24919962287#step:16:4699)
for what a timeout failure now looks like.

ci-exactly:
build_packages,regression_test_amdgpu_rocm,regression_test_cpu,regression_test_nvidiagpu_cuda,regression_test_nvidiagpu_vulkan,test_tensorflow_cpu

Signed-off-by: Lubo Litchev <lubol@google.com>
@ScottTodd ScottTodd mentioned this issue Aug 13, 2024
Open
@ScottTodd ScottTodd added the integrations/onnx ONNX integration work label Aug 19, 2024
@ScottTodd ScottTodd mentioned this issue Aug 25, 2024
Open
@benvanik benvanik closed this as not planned Won't fix, can't repro, duplicate, stale Apr 29, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
bug 🐞 Something isn't working integrations/onnx ONNX integration work integrations/pytorch PyTorch integration work
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

Re-enable skipped tests that no longer hang. ScottTodd/iree
2 participants
@benvanik @ScottTodd
0