bug_list.md

Bug List

Note Annotations

• Status: ✅ fixed; 🚨 high-priority bug; 🔵 explicitly confirmed by developers; ❌ won't fix / not a bug; 🔁 duplicated;
• Symptom: 💥 Crash or exception; 🧮 Wrong Error Messages 🔀 inconsistency;

PyTorch

1. 🚨💥✅Segmentation fault in torch.nn.functional.cosine_embedding_loss with empty tensors and mixed dtypes #112732· pytorch/pytorch · GitHub

2. 🚨💥✅Segmentation Fault on torch.div with Empty Tensor Dimensions #113037· pytorch/pytorch · GitHub

3. 🚨🔀✅Inconsistent Behavior Between Eager Execution and Compiled Execution for adaptive_avg_pool1d and argmax #113013· pytorch/pytorch · GitHub

4. 🚨💥✅Compile error with nn.RReLU and inplace=True When Using torch.compile() #107993 · pytorch/pytorch · GitHub

5. 🚨💥✅torch.compile Raises Compilation Error for torch.logical_not with Empty Output Tensor #108000 · pytorch/pytorch · GitHub

6. 🚨💥✅AssertionError Triggered with Different Tensor Shapes of Input and Output in torch.cosh when Using torch.compile #108021· pytorch/pytorch · GitHub

7. 🚨🔀✅Behavior Discrepancy Between Eager Mode and Optimized Mode for torch.abs with transposed out #113010· pytorch/pytorch · GitHub

8. 🚨💥✅core dumped with torch.logical_or on torch.compile mode and eager mode #117624· pytorch/pytorch · GitHub

9. 🚨💥✅Trouble Pickling complex32 Tensor #120290 · GitHub

10. 🚨🔀✅different outputs of torch.dig_embed between torch.compile mode and eagermode #117019· pytorch/pytorch · GitHub

11. 🚨🔀✅Inconsistency bug in torch.bitwise_right_shift in torch.compile #118265· pytorch/pytorch · GitHub

12. 💥✅torch.compile Raises "NameError: name 'buf0' is not defined" for LeakyReLU with Empty Input Tensor #107995 · pytorch/pytorch · GitHub

13. 🧮✅Misleading error messages for torch.nn.ConstantPad2d and torch.nn.ConstantPad3d #104508 · pytorch/pytorch · GitHub

14. 🔀✅Inconsistency in nn.ChannelShuffle Behavior with Empty Input Tensor on CPU and CUDA #104603 · pytorch/pytorch · GitHub

15. 💥✅torch.compile fails with "BackendCompilerFailed" with AdaptiveAvgPool of output_size 0 #104618 · pytorch/pytorch · GitHub

16. 💥✅Assertion error for torch.nn.CrossMapLRN2d with wrong input dimension #104839 · pytorch/pytorch · GitHub

17. 💥✅[torch.compile] compile failed when nn.Linear with zero in_feature size #104937 · pytorch/pytorch · GitHub

18. 💥✅[torch.compile]LoweringException when using nn.InstanceNorm3d with certain input sizes #104948 · pytorch/pytorch · GitHub

19. 🔀✅Different behaviors ontorch.var_mean with torch.compile(default mode) #113167· pytorch/pytorch · GitHub

20. 🔀✅Inconsistency error with torch.sinh and torch.relu in PyTorch Model #117544· pytorch/pytorch · GitHub

21. 🧮✅Wrong error messages with torch.nn.AdaptiveMaxPool1d #104822 · pytorch/pytorch · GitHub

22. 💥✅[torch.compile] RuntimeError during Gradient Computation in torch.compile() #104943 · pytorch/pytorch · GitHub

23. 💥✅LazyBatchNorm1d requires previous instantiation before using dictionary input #105292 · pytorch/pytorch · GitHub

24. 🧮✅Assertion error related to weight dimension in MultiMarginLoss #106020 · pytorch/pytorch · GitHub

25. 💥✅Different Behaviors on nn.Dropout when p=1 When Using torch.compile() ##107942 · pytorch/pytorch · GitHub

26. 🧮✅AssertionError Triggered with 5D Input Tensor in torch.atan when Using torch.compile #108022· pytorch/pytorch · GitHub

27. 💥✅RuntimeError with nn.ConstantPad when using torch.compile in max-autotune mode #108043· pytorch/pytorch · GitHub

28. 💥✅Compilation failure: torch.mm out variant w/ dynamic=True #112489· pytorch/pytorch · GitHub

29. 💥✅torch.addcdiv INTERNAL ASSERT FAILURE #112490· pytorch/pytorch · GitHub

30. 💥✅Compilation Failure of torch.Var in torch.compile Optimized Mode #112494· pytorch/pytorch · GitHub

31. 💥✅Compilation Failure of torch.nn.functional.adaptive_max_pool3d_with_indices in torch.compile Optimized Mode #112496· pytorch/pytorch · GitHub

32. 🔀✅Discrepancy Between Eager Mode and Optimized Mode Results for hinge_embedding_loss #113012· pytorch/pytorch · GitHub

33. 💥✅ZeroDivisionError in torch.compile with torch.unfold_copy and Zero-Dimension Tensor #113026· pytorch/pytorch · GitHub

34. 💥✅RuntimeError on linalg_matrix_power Function with torch.compile #113003· pytorch/pytorch · GitHub

35. 💥✅With torch.compile in default setting special_hermite_polynomial_he failed running call_function #113119· pytorch/pytorch · GitHub

36. 🔀✅Inconsistency with torch.acosh with torch.compile #118267· pytorch/pytorch · GitHub

37. 🔀✅Inconsistency with torch.log1p combined with torch.argmax in torch.compile #118266· pytorch/pytorch · GitHub

38. 💥✅Floating Point Exception on Empty Input Tensor with torch.batch_norm_update_stats #120803 · GitHub

39. 💥✅Segmentation Fault due to Invalid out_dim Type in torch.tensor.flatten #120762 · GitHub

40. 💥✅torch.native_channel_shuffle with empty tensor trigger floating point exception. #121092· pytorch/pytorch · GitHub

41. 💥✅Segmentation Fault in torch._C._nn.replication_pad2d with Incorrect Inputs #121093· pytorch/pytorch · GitHub

42. 💥✅Core dumped with torch._C.nn.rrelu_with_noise #121094· pytorch/pytorch · GitHub

43. 💥✅Segmentation Fault in torch._C._nn.thnn_conv2d Due to Invalid Padding Argument #121188· pytorch/pytorch · GitHub

44. 🔀✅Discrepancy in Behavior of torch.diag_embed Between Eager Execution and 'torch.compiled' Optimized Mode #112713· pytorch/pytorch · GitHub

45. 💥✅RuntimeError on torch.unqiue_consecutive with torch.compile( fullgraph = true) #113118· pytorch/pytorch · GitHub

46. 🔀✅different outputs on torch.asinh between eager mode and torch.compile #113151· pytorch/pytorch · GitHub

47. 🔀✅Different behaviors in torch.special.sinc between eagermode and torch.compile #118176· pytorch/pytorch · GitHub

48. 💥🔵Segmentation Fault Due to Empty Stride Argument in torch._C._nn.slow_conv3d #121095· pytorch/pytorch · GitHub

49. 🧮🔵Unrelated error messages with torch.nn.AdaptiveAvgPool3d #104823 · pytorch/pytorch · GitHub

50. 🧮🔵Incorrect Error Message Ordering for nn.AdaptiveAvgPool2d with Incorrect output_size #104698 · pytorch/pytorch · GitHub

51. 🧮🔵Misleading error message in multilabel_margin_loss when passing incompatible tensor dimensions #106011 · pytorch/pytorch · GitHub

52. 🧮🔵Improve Error Message in MultiMarginLoss for Inconsistent Target Size #106251 · pytorch/pytorch · GitHub

53. 🧮🔵Inconsistent Keyword Arguments behaviors in torch.triangular_solve() #111908· pytorch/pytorch · GitHub

54. 🧮🔵Contradictory Error Message for stride Argument in torch.conv_transpose3d() #112376· pytorch/pytorch · GitHub

55. 🔀🔵different behaviors in cumsum between torch.compile mode and eager mode #117623· pytorch/pytorch · GitHub

56. 🔀🔵Different behaviors with torch.isfinite and torch.tril between eagermode and JIT #118086· pytorch/pytorch · GitHub

57. 🔀🔵Different behaviors in torch.nn.functional.hinge_embedding_loss between eagermode and torch.compile #118175· pytorch/pytorch · GitHub

58. 🔀🔵 Inconsistent Behavior of torch.dsplit with torch.compile #118741 · GitHub

59. 💥🔵Compilation Failure with torch.nn.functional.rrelu(training=True) in Torch Compile #120292 · GitHub

60. 💥🔵torch compile error with torch.Tensor.unsqueeze_ #129673 · GitHub

61. 💥🔵torch.lt with out parameter does not return the correct shape #129820 · GitHub

62. 🔀🔵different behaviors in kl_div between torch.compile mode and eager mode #117034· pytorch/pytorch · GitHub

63. 🔀🔵Inconsistent outputs of torch.adaptive_avg_pool1d and torch.threshold_ between eagermode and torch.compile #118302· pytorch/pytorch · GitHub

64. 🔀🔵Inconsistent outputs of mish and log10 between eagermode and torch.jit.trace #118303· pytorch/pytorch · GitHub

65. 🔀🔵 Inconsistent outputs in linalg_ldl_factor_ex and special_spheriacal_bessel_j0 between eagermode and torch.compile #118300· pytorch/pytorch · GitHub

66. 🔀Different behaviors in torch.nn.linear between torch.compile and eagermode #118087· pytorch/pytorch · GitHub

67. 🔀Inconsistent outputs of cosine_similarity and linalg_svdvals between eagermode and torch.compile #118297· pytorch/pytorch · GitHub

68. 🔀Inconsistent output of torch.lodget with JIT compilation #118400· pytorch/pytorch · GitHub

69. 🔀Inconsistency in Output When Using torch.nn.functional.selu with JIT Compilation #118401· pytorch/pytorch · GitHub

70. 🔀Inconsistency in Output for torch.nn.functional.poisson_nll_loss in JIT Compiled Model #118462 · GitHub

71. 🔀Inconsistency in Output for torch.argmin in JIT Compiled Model #118463 · GitHub

72. 🔀Different behaviors of torch._C._linalg.linalg_qr and torch.pow in torch.compile #118964 · GitHub

73. 🔀Inconsistency error of torch.Tensor.atan2_ with torch.compile #124020 · GitHub

74. 🔀torch.compile Fails to Compile for nn.SELU(inplace=True) with Empty Tensor Input #108019 · pytorch/pytorch · GitHub

TensorFlow

1. 💥✅core dumped with Conv2DBackpropFilter #62142 · tensorflow/tensorflow · GitHub
2. 🔀✅Different Behavior of tf.raw_ops.Cosh with jit_compile=True #62236 · tensorflow/tensorflow · GitHub
3. 🔀✅Different behaviors of raw_ops.Sigmoid can be observed when jitcompiled=true. #62212 · tensorflow/tensorflow · GitHub
4. 💥✅core dumped Error with tf.raw_ops.QuantizeAndDequantizeV4 · Issue #62055 · tensorflow/tensorflow · GitHub
5. 💥🔵core dumped with tf.raw_ops.BiasAdd #62146 · tensorflow/tensorflow · GitHub
6. 💥🔵core dumped Error with tf.raw_ops.TensorScatterUpdate #62184 · tensorflow/tensorflow · GitHub
7. 🔀🔵Different Behavior of tf.raw_ops.Asinh with jit_compile=True #62220 · tensorflow/tensorflow · GitHub
8. 🔀🔵Different Behavior of tf.raw_ops.Asin with jit_compile=True #62280 · tensorflow/tensorflow · GitHub
9. 🔀🔵Different Behavior of tf.raw_ops.SqrtGrad with jit_compile=True #62263 · tensorflow/tensorflow · GitHub
10. 🔀Different Behavior of tf.raw_ops.Cos+tf.raw_ops.Erfc with jit_compile=True #62287 · tensorflow/tensorflow · GitHub