Bugs uncovered by the NNSmith project

PyTorch

• ✅🧮🚨 torch.compile produce wrong result in backward ad with conv2d + interpolate when interpolatemode=nearest/bilinear/bicubic · Issue #100794 · pytorch/pytorch
• ✅💥🚨 SIGIOT when running model with conv2d and avgpool2d after optimize_for_inference · Issue #86535 · pytorch/pytorch · GitHub
• ✅🧮 optimize_for_inference leads to wrong results for model with conv2d, max and clip · Issue #86556 · pytorch/pytorch · GitHub
• ✅💥🚨 RuntimeError: could not construct a memory descriptor using a format tag · Issue #86664 · pytorch/pytorch · GitHub
• ✅💥 [NNPack] Runtime error with padded Conv1d and >=16 batch size · Issue #90142 · pytorch/pytorch · GitHub
• ✅💥 stable torch.sort crash with expanded tensor · Issue #91420 · pytorch/pytorch · GitHub
• ✅💥 [Crash] torch.searchsorted with out-of-bound sorter · Issue #91606 · pytorch/pytorch · GitHub
• ✅🧮 index_select with scalar input and 0-dimed vector leads to undeterministic output · Issue #94340 · pytorch/pytorch · GitHub
• ✅🧮 index_select with scalar input and 0-dimed vector leads to undeterministic output · Issue #94340 · pytorch/pytorch · GitHub
• ✅💥 torch.compile failed on torch.add with a constant python number · Issue #92324 · pytorch/pytorch · GitHub
• ✅💥 torch.compile generates wrong profiling program for randn_like · Issue #92368 · pytorch/pytorch · GitHub
• ✅💥 torch.compile generates wrong profiling program for function having transpose and lerp · Issue #93229 · pytorch/pytorch · GitHub
• ✅💥 torch.compile triggers assertion error when explicitly provide out=None · Issue #92814 · pytorch/pytorch · GitHub
• ✅💥 INTERNAL ASSERT FAILED in torch.compile when the input tensor of torch.clamp has requires_grad=True · Issue #93225 · pytorch/pytorch · GitHub
• ✅💥 torch.compile failed to run in-place operation unsqueeze_(0) · Issue #93259 · pytorch/pytorch · GitHub
• ✅🧮 stack + inplace operator produce wrong results in torch.compile · Issue #93283 · pytorch/pytorch · GitHub
• ✅🧮 [pt2] compiled model with cat and expand gives wrong results · Issue #93357 · pytorch/pytorch · GitHub
• ✅🧮🚨 [pt2] compiled function with cat and mul gives wrong results · Issue #93365 · pytorch/pytorch · GitHub
• ✅🧮 [pt2] cannot compile model with linear layer when the input has rank 1 · Issue #93372 · pytorch/pytorch · GitHub
• ✅💥 softmax + transpose + div_ triggers assertion fail in compile mode · Issue #93371 · pytorch/pytorch · GitHub
• ✅🧮🚨 [pt2] torch.where gives wrong results with torch.compile · Issue #93374 · pytorch/pytorch · GitHub
• ✅💥 torch.rsub with alpha=xxx triggers assertion fail in compile mode · Issue #93376 · pytorch/pytorch · GitHub
• ✅🧮 [pt2] compile gives wrong result for function having expand and div_ · Issue #93377 · pytorch/pytorch · GitHub
• ✅💥 [pt2] torch._inductor.exc.CppCompileError: C++ compile error when compiling neg and max · Issue #93380 · pytorch/pytorch · GitHub
• ✅💥 [pt2] exception when compiling max_pool2d_with_indices · Issue #93384 · pytorch/pytorch · GitHub
• ✅💥 [pt2] cannot compile function having gt, expand and add_ · Issue #93386 · pytorch/pytorch · GitHub
• ✅💥🚨 torch.compile trigger assertion error when executing histogramdd · Issue #93274 · pytorch/pytorch · GitHub
• ✅🧮 [pt2] torch.compile produces wrong results for masked_fill · Issue #93823 · pytorch/pytorch · GitHub
• ✅🧮 [pt2] torch.compile produces wrong results for function with reciprocal_ · Issue #93824 · pytorch/pytorch · GitHub
• ✅🧮 [pt2] torch.compile produces wrong results for function with neg on uint8 tensor · Issue #93829 · pytorch/pytorch · GitHub
• ✅💥 log_softmax + pad triggers assertion fail in compile mode · Issue #93819 · pytorch/pytorch · GitHub
• ✅💥 [pt2] Cannot compile model with neg and linear · Issue #93836 · pytorch/pytorch · GitHub
• ✅🧮 pad + gt produce wrong results in compile mode · Issue #93351 · pytorch/pytorch · GitHub
• ✅💥 [pt2] (interpolate with mode=nearest) + kthvalue triggers assertion fail in compile mode · Issue #93849 · pytorch/pytorch · GitHub
• ✅💥 [pt2] torch._inductor.exc.CppCompileError: C++ compile error when compiling argmax and min · Issue #94055 · pytorch/pytorch · GitHub
• ✅💥 Tensor.select + add_ triggers C++ Compile Error · Issue #94960 · pytorch/pytorch · GitHub
• ✅💥 torch.compile fails when using torch.sub with python constant · Issue #95181 · pytorch/pytorch · GitHub
• ✅💥 Tensor.copy_ + moveaxis Trigger Exception in Compile Mode · Issue #95262 · pytorch/pytorch · GitHub
• ✅🧮 torch.ge produces wrong results in compile mode when given int tensors · Issue #95695 · pytorch/pytorch · GitHub
• ✅💥 [pt2] bitwise_and + clamp_max Triggers Compilation Error · Issue #97968 · pytorch/pytorch · GitHub
• ✅🧮 [pt2] add + unfold + abs_ produces wrong results · Issue #98143 · pytorch/pytorch · GitHub
• ✅🧮 [pt2] pow + cos produces wrong result · Issue #98149 · pytorch/pytorch · GitHub
• ✅💥 torch._C._nn.fractional_max_pool3d Trigger Segmentation Fault · Issue #89648 · pytorch/pytorch · GitHub
• ✅💥🚨 torch.nn.functional.embedding_bag Trigger "IOT instruction" Failure · Issue #89677 · pytorch/pytorch · GitHub
• ✅🧴 torch.Tensor.index_select Trigger heap-buffer-overflow with AddressSanitizer · Issue #88940 · pytorch/pytorch · GitHub
• ✅🧴 nn.utils.rnn.pack_sequence Trigger heap-buffer-overflow with AddressSanitizer · Issue #88334 · pytorch/pytorch · GitHub
• ✅🚨🧴 MultiMarginLoss Trigger out-of-bound Read under Compute Sanitizer · Issue #88724 · pytorch/pytorch · GitHub
• ✅🧴 nn.functional.max_unpool3d Trigger heap-buffer-overflow with AddressSanitizer · Issue #88032 · pytorch/pytorch · GitHub
• ✅🧴 torch.nn.functional.interpolate Trigger heap-buffer-overflow with AddressSanitizer · Issue #88939 · pytorch/pytorch · GitHub
• ✅🧴 torch.fft.hfft Trigger RuntimeError under UndefinedBehaviorSanitizer · Issue #88985 · pytorch/pytorch · GitHub
• ✅🧴 torch.nn.functional.interpolate Trigger RuntimeError under UndefinedBehaviorSanitizer · Issue #88951 · pytorch/pytorch · GitHub
• ✅💥🚨 torch.compile failed on torch.bitwise_xor with a constant python number · Issue #93224 · pytorch/pytorch · GitHub
• ✅💥 [CPU Inductor] Compile error when passing float16 tensors to vector_norm + remainder · Issue #97758 · pytorch/pytorch · GitHub
• ✅💥 [pt2] movedim + add_ + cat triggers exception · Issue #98122 · pytorch/pytorch · GitHub
• ✅🧮 dstack + reciprocal produce wrong result in compile mode · Issue #93078 · pytorch/pytorch · GitHub
• ✅💥 min reduction on float16 tensor failed on certain shapes · Issue #93249 · pytorch/pytorch · GitHub
• ✅💥 argmin + view Trigger Exception in compile mode · Issue #95370 · pytorch/pytorch · GitHub
• 🔵💥 [JIT] Zero-channel conv2d cannot be applied with optimize_for_inference · Issue #91396 · pytorch/pytorch · GitHub
• 🔵💥 [JIT] Applying conv2d over Constants Leads to Exception · Issue #92740 · pytorch/pytorch · GitHub
• ✅🧮🚨 torch.compile produce wrong result in interpolate when mode=bilinear · Issue #93262 · pytorch/pytorch · GitHub
• 🔵🧮 torch.fmod produces inconsistent results in eager and compile mode · Issue #97333 · pytorch/pytorch · GitHub
• 🔵💥 torch.Tensor.flatten Trigger Segmentation Fault when trying to provide and output named dim · Issue #89718 · pytorch/pytorch · GitHub
• 🔵🧴 nn.functional.embedding_bag Trigger out-of-bound Read under Compute Sanitizer · Issue #88563 · pytorch/pytorch · GitHub
• ✅🔵🧴 torch.nn.CTCLoss Trigger heap-buffer-overflow under AddressSanitizer · Issue #88047 · pytorch/pytorch · GitHub
• 🔵🧴 torch.nn.ReplicationPad2D Report "invalid configuration argument" Error under Compute Sanitizer · Issue #89254 · pytorch/pytorch · GitHub
• 🔵🧴 torch.nn.LayerNorm Abort with "invalid device ordinal" Error · Issue #89218 · pytorch/pytorch · GitHub
• 🔵🧴 torch.svd_lowrank Trigger RuntimeError under UndefinedBehaviorSanitizer · Issue #88942 · pytorch/pytorch · GitHub
• 🔵🧴 torch.linalg.lstsq Trigger RuntimeError under UndefinedBehaviorSanitizer · Issue #88941 · pytorch/pytorch · GitHub
• 💥 Adding a linear layer leads to failure of optimize_for_mobile · Issue #86667 · pytorch/pytorch · GitHub
• 💥 [JIT] INTERNAL ASSERT FAILED when dispatching for torch.Tensor.view · Issue #90365 · pytorch/pytorch · GitHub
• 💥 [JIT] INTERNAL ASSERT FAILED torch.add with boolean primitive constant · Issue #90367 · pytorch/pytorch · GitHub
• 💥 [JIT] INTERNAL ASSERT FAILED torch.mul with boolean primitive constant · Issue #90366 · pytorch/pytorch · GitHub
• 💥 [JIT] Wrong type inference leads to misleading error message · Issue #90369 · pytorch/pytorch · GitHub
• 💥 [JIT] INTERNAL ASSERT FAILED when Conv2d and clamp used together · Issue #92563 · pytorch/pytorch · GitHub
• 💥 [JIT] Inconsistency in tensor shape between eager mode and JIT · Issue #92548 · pytorch/pytorch · GitHub
• 💥 [JIT][TracingCheckError] inplace ops incompatible with contiguous(.., channels_last) · Issue #92558 · pytorch/pytorch · GitHub
• 💥 [JIT] Consecutive use of addmm Leads to Exception · Issue #92742 · pytorch/pytorch · GitHub
• 🧴 torch.topk Trigger RuntimError under UndefinedBehaviorSanitizer · Issue #88944 · pytorch/pytorch · GitHub
• 🧴 torch.vander Trigger RuntimeError with UndefinedBehaviorSanitizer · Issue #88943 · pytorch/pytorch · GitHub
• ✅🧴 torch.nn.CTCLoss Trigger out-of-bound Read under Compute Sanitizer · Issue #89208 · pytorch/pytorch · GitHub
• 🧴 torch.nn.functional.embedding_bag Trigger RuntimeError under UndefinedBehaviorSanitizer · Issue #88950 · pytorch/pytorch · GitHub
• 🧴 torch.set_rng_state Trigger RuntimeError under UndefinedBehaviorSanitizer · Issue #88949 · pytorch/pytorch · GitHub
• 🧴 torch.Tensor.msort Trigger RuntimeError under UndefinedBehaviorSanitizer · Issue #88947 · pytorch/pytorch · GitHub
• 🧴 torch.linalg.eigvals Trigger RuntimeError under UndefinedBehaviorSanitizer · Issue #88945 · pytorch/pytorch · GitHub

PyTorch-ONNX Converter

• ✅ [ONNX] f64 * LeakyReLU(f64) mistakingly returns f32 · Issue #85316 · pytorch/pytorch · GitHub
• ✅ [ONNX] Converter did not consider the implicit casting specifically for Max · Issue #87609 · pytorch/pytorch · GitHub
• ✅ fix: onnx PReLU unidirectional broadcasting by ganler · Pull Request #70571 · pytorch/pytorch · GitHub
• ✅ [Clip] [ONNX] Make Non-Float Op Exportation Compatible to Avoid Invalid ONNX Models by ganler · Pull Request #72401 · pytorch/pytorch · GitHub
• ✅ [Min] [ONNX] Make Non-Float Op Exportation Compatible to Avoid Invalid ONNX Models by ganler · Pull Request #72401 · pytorch/pytorch · GitHub
• ✅ [Max] [ONNX] Make Non-Float Op Exportation Compatible to Avoid Invalid ONNX Models by ganler · Pull Request #72401 · pytorch/pytorch · GitHub
• ✅ [ReLU] [ONNX] Make Non-Float Op Exportation Compatible to Avoid Invalid ONNX Models by ganler · Pull Request #72401 · pytorch/pytorch · GitHub
• ✅ [Pad] [ONNX] Make Non-Float Op Exportation Compatible to Avoid Invalid ONNX Models by ganler · Pull Request #72401 · pytorch/pytorch · GitHub
• ✅ [onnx export] Add broadcast to matmul shape inference by lazycal · Pull Request #70534 · pytorch/pytorch · GitHub
• ✅ [Bug][ONNX] Specification Inconsistency in Flatten · Issue #74142 · pytorch/pytorch · GitHub
• ✅ [ONNX] Fix shape inconsistency when exporting scalar log2 by lazycal · Pull Request #78701 · pytorch/pytorch · GitHub
• ✅ [ONNX Export] Interpolation likely should be exported with half_pixel instead of pytorch_half_pixel · Issue #79361 · pytorch/pytorch · GitHub

ONNX

• ✅ [Bug] Checker misses data type mismatch for Max · Issue #4619 · onnx/onnx · GitHub

ONNXRuntime

• ✅ Crashes when relu is followed by a clip · Issue #9753 · microsoft/onnxruntime · GitHub
• ✅ MatMul fusion failed at scalar input · Issue #10950 · microsoft/onnxruntime · GitHub
• ✅ GemmTransposeFusion error when C is transposed (Gemm(A,B,Transpose(C)), complained with confusing name _transformed_transformed_transformed... · Issue #12071 · microsoft/onnxruntime · GitHub
• 🔵 [Bug] Mixing negative and positive paddings causes segfault/uninitialized memory values produced in reflected pad · Issue #11828 · microsoft/onnxruntime · GitHub
• 🔵 Runtime Exception when relu is followed by a clip · Issue #10936 · microsoft/onnxruntime · GitHub
• 🔵 Inconsistent result to NumPy and PyTorch when consecutively casting a float tensor to int32 and then to bool · Issue #11994 · microsoft/onnxruntime · GitHub
• 🔵 Wrong output shape due to MergeShape failure · Issue #11870 · microsoft/onnxruntime · GitHub
• Wrong Floor output on very large input · Issue #12076 · microsoft/onnxruntime · GitHub
• Resize with mode linear always produces 0.5 on GPU regardless of the input · Issue #12091 · microsoft/onnxruntime · GitHub
• Resize with nearest mode have inconsistent results compared to PyTorch and TVM · Issue #12098 · microsoft/onnxruntime · GitHub
• Parameters are optimized out even if it is a needed return value · Issue #13425 · microsoft/onnxruntime · GitHub

TVM

• ✅ [Bug] shape int32-int64 check error in trilu's te.compute · Issue #13029 · apache/tvm · GitHub
• ✅ [Bug] trilu not tagged with injective and thus miss reduce schedule · Issue #13030 · apache/tvm · GitHub
• ✅ [Bug] Wrong results of cast<int32>( cast<bool>(-1i64) ) · Issue #13048 · apache/tvm · GitHub
• ✅ [BugFix] resolve integer 32. ~ 64. mismatch by casting by ganler · Pull Request #9582 · apache/tvm · GitHub
• ✅ [onnx] fix onnx where broadcast by lazycal · Pull Request #10106 · apache/tvm · GitHub
• ✅ Fix broadcast InferCorrectLayout by lazycal · Pull Request #10156 · apache/tvm · GitHub
• ✅ [BUGFIX][ARITH] Fix FloorMod Simplifier by lazycal · Pull Request #10336 · apache/tvm · GitHub
• ✅ [BugFix]: select node type error in NarrowDataType pass by ganler · Pull Request #10519 · apache/tvm · GitHub
• [Bug] GetStoreRule failure at simple Conv2d + Squeeze model · Issue #10528 · apache/tvm · GitHub
• ✅ [Relay][ONNX][Fix] Flatten in OnnxConverter by ganler · Pull Request #10593 · apache/tvm · GitHub
• ✅ [NarrowDataType] [TIR] Fix Ramp int32~64 mismatch in VectorizeLoop and NarrowDataType passes by lazycal · Pull Request #10172 · apache/tvm · GitHub
• ✅ [VectorizeLoop] [TIR] Fix Ramp int32~64 mismatch in VectorizeLoop and NarrowDataType passes by lazycal · Pull Request #10172 · apache/tvm · GitHub
• ✅ [Bug][TE Schedule] Unsupported nested parallel created by Softmax TE schedule · Issue #12001 · apache/tvm · GitHub
• ✅ [fix] vec * mat in matmul in onnx converter by ganler · Pull Request #11174 · apache/tvm · GitHub
• ✅ fix vec*mat in PyTorch converter by ganler · Pull Request #11347 · apache/tvm · GitHub
• ✅ [TIR] Fix int32 vs int64 mismatch in For construct. by lazycal · Pull Request #10595 · apache/tvm · GitHub
• ✅ Add missing Slice layout fallback check of stride=1 . by lazycal · Pull Request #10690 · apache/tvm · GitHub
• ✅ Onnx squeeze enabled with auto axis handling. by ganler · Pull Request #10742 · apache/tvm · GitHub
• ✅ [Reduce] [ONNX] fix reduce crash on scalar inputs by ganler · Pull Request #10780 · apache/tvm · GitHub
• ✅ [ReduceSumSquare] [ONNX] fix reduce crash on scalar inputs by ganler · Pull Request #10780 · apache/tvm · GitHub
• ✅ [ReduceL1] [ONNX] fix reduce crash on scalar inputs by ganler · Pull Request #10780 · apache/tvm · GitHub
• ✅ [ReduceL2] [ONNX] fix reduce crash on scalar inputs by ganler · Pull Request #10780 · apache/tvm · GitHub
• ✅ [ReduceLogSum][ONNX] fix reduce crash on scalar inputs by ganler · Pull Request #10780 · apache/tvm · GitHub
• ✅ [FIX] resolve int64/32 for AttrStmtNode by ganler · Pull Request #10983 · apache/tvm · GitHub
• ✅ Fix onnx round import with float64 inputs. by lazycal · Pull Request #11685 · apache/tvm · GitHub
• ✅ Fix 1d-softmax schedule. by lazycal · Pull Request #11719 · apache/tvm · GitHub
• ✅ [Fix] int32/64 mismatch of buffer elem_offset at HandleBufferBindScope by ganler · Pull Request #11755 · apache/tvm · GitHub
• ✅ [Bug] Int64 BroadCast-ArgMax triggers assertion error at graph runtime · Issue #11794 · apache/tvm · GitHub
• ✅ [TE Schedule] Fix broken 2D softmax TE schedules when axis=0 by lazycal · Pull Request #11803 · apache/tvm · GitHub
• ✅ [Bug] concat([x], axis=1) return random results · Issue #11895 · apache/tvm · GitHub
• ✅ Fix infercorrect layout in Layoutrewrite and improve naming. by lazycal · Pull Request #12007 · apache/tvm · GitHub
• ✅ Several type mismatch fixes and checks by kparzysz-quic · Pull Request #12041 · apache/tvm · GitHub
• ✅ [FIX][ONNX][Relay] onnx converter on matmul with scalar; bring back nn.matmul check by ganler · Pull Request #13448 · apache/tvm · GitHub
• ✅ [Bug] Layout Error when Putting argmin after conv2d · Issue #9813 · apache/tvm · GitHub
• ✅ Fix LayoutRewriter by lazycal · Pull Request #10118 · apache/tvm · GitHub
• [Bug] concatenating strided slice and negative padding causes wrong buffer binding · Issue #11897 · apache/tvm
• [Bug] GPU lower_thread_allreduce is_zero(index) check false · Issue #11898 · apache/tvm
• Resize does not reject unsupported layout during AlterOpLayout · Issue #12008 · apache/tvm
• [Bug] Compilation failure for broadcast-argmax in internal type inference · Issue #13031 · apache/tvm · GitHub
• [Bug] Compiled squeeze-broadcast_to-argmin fails at graph runtime · Issue #13045 · apache/tvm · GitHub

TensorRT

• ✅ Segfault on const+prelu+reduce_mean+comparison_op · Issue #1738 · NVIDIA/TensorRT · GitHub
• ✅ Gemm conversion error, seem to be caused by squeeze · Issue #824 · onnx/onnx-tensorrt · GitHub
• ✅ [Bug] crash on poolings with larger-than-317 pool sizes · Issue #2094 · NVIDIA/TensorRT
• ✅ [Bug] --loadInputs not working: input name mismatch when Flatten is the input node · Issue #1990 · NVIDIA/TensorRT · GitHub
• ✅ Cuda OutOfMemory when creating tensor with $2^{29}$ (~0.5 G) elements - TensorRT - NVIDIA Developer Forums
• ✅ Myelin error on onnx model: Assertion `i < crds_.size() < failed · Issue #1781 · NVIDIA/TensorRT · GitHub
• Segmentation fault when using TensorRT to compile a model - TensorRT - NVIDIA Developer Forums
• Internal Error: GPU error during getBestTactic: PWN(LeakyRelu_4) : misaligned address - TensorRT - NVIDIA Developer Forums
• 🔵 Duplicated reshapes triggers "[graphOptimizer.cpp::findOne::510] Error Code 2: Internal Error (Assertion it != v.end() failed. )" - TensorRT - NVIDIA Developer Forums
• 🔵 Incorrect slicing of boolean constant tensor with step size > 1 - TensorRT - NVIDIA Developer Forums

TensorFlow

• 🔵💥 Inconsistant behavior of Conv2D between eager mode and tracing · Issue #57664 · tensorflow/tensorflow · GitHub
• 🔵💥 TFLite fails to run a model with a dense layer following an Add operator · Issue #57697 · tensorflow/tensorflow · GitHub
• 🔵💥 TFLite throws an error with certain tensor value · Issue #57708 · tensorflow/tensorflow · GitHub
• 🔵🧮 TFLite's max operator has wrong broadcasting behavior · Issue #57759 · tensorflow/tensorflow · GitHub
• 🔵💥 Issues · tensorflow/tensorflow · GitHub
• 🔵🧮 pow operation gives valid output even the input is invalid · Issue #57757 · tensorflow/tensorflow · GitHub
• 🔵🧮 TFLite produce wrong results when add follows a leakyrelu · Issue #57818 · tensorflow/tensorflow · GitHub
• 🔵💥 TFLite runner crashes with XOR and squeeze in the model · Issue #57882 · tensorflow/tensorflow · GitHub
• ✅💥 Conv2D with XLA jit_compile=True fails to run · Issue #57748 · tensorflow/tensorflow · GitHub
• 🔵🧮 log operator outputs wrong results with XLA compilation · Issue #57744 · tensorflow/tensorflow · GitHub
• 🔵🧮 pow operator output nan for valid inputs · Issue #57747 · tensorflow/tensorflow · GitHub
• 🔵🧮 LRN operator outputs wrong results with jit_compile=True · Issue #57746 · tensorflow/tensorflow · GitHub
• 🔵💥 Conv2D layer fails to run with XLA on CUDA · Issue #57838 · tensorflow/tensorflow · GitHub
• 🔵🧴 tf.raw_ops.SegmentMax Behaves Differently Under CPU and GPU · Issue #58469 · tensorflow/tensorflow · GitHub

Hidet

Based on NNSmith, @soodoshll found a number of bugs for Hidet, including:

• ✅ [Bug] Use int64 in argmax · Issue #103 · hidet-org/hidet
• ✅ [Bug] broadcast_shape parameter type error · Issue #85 · hidet-org/hidet
• ✅ [Bug] Data type casting from onnx · Issue #87 · hidet-org/hidet
• ✅ [Bug] cuda code compilation error · Issue #89 · hidet-org/hidet
• ✅ [Bug] MinOp generates max code · Issue #90 · hidet-org/hidet
• ✅ [Bug] FP64 reduce · Issue #91 · hidet-org/hidet
• ✅ [Bug] Inconsistent definition of the inputs parameter of operators · Issue #93 · hidet-org/hidet
• ✅ [Bug] Slice indexing in ONNX · Issue #94 · hidet-org/hidet
• ✅ [Bug] binary arithmetic with CUDA scalar · Issue #95 · hidet-org/hidet
• ✅ [Bug] Unexpected behavior when inputs and outputs overlap · Issue #96 · hidet-org/hidet
• ✅ [Followup][Bug] Unexpected behavior when inputs and outputs overlap · Issue #96 · hidet-org/hidet
• ✅ [Bug] arguments of clip drop after fusion · Issue #97 · hidet-org/hidet
• ✅ [Bug] fusion rewrite fails · Issue #99 · hidet-org/hidet