Dilated Conv3d segfaults (cpu)

[mys007]

I have encountered a segfault when running Conv3d with dilation under pytorch 0.4.0 on the cpu for some specific tensor sizes. This is the code to reproduce the behavior:

import torch
import torch.nn as nn
mod = nn.Conv3d(96, 256, 4, 1, 0, 2)     # does not crash with dilation = 1
mod(torch.zeros([1, 96, 111, 63, 111]))

And this is the stack trace from gdb:

#0  0x00007fffbfb0c3d7 in THNN_Floatvol2col(float const*, int, int, int, int, int, int, int, int, int, int, int, int, int, int, int, int, int, int, int, float*)
    () from /home/martin/miniconda3/envs/BAI/lib/python3.6/site-packages/torch/lib/libATen.so
#1  0x00007fffbfc240cb in THNN_FloatVolumetricDilatedConvolution_updateOutput ()
   from /home/martin/miniconda3/envs/BAI/lib/python3.6/site-packages/torch/lib/libATen.so
#2  0x00007fffbf42ae43 in at::CPUFloatType::thnn_conv_dilated3d_forward(at::Tensor const&, at::Tensor const&, at::ArrayRef<long>, at::Tensor const&, at::ArrayRef<long>, at::ArrayRef<long>, at::ArrayRef<long>) const () from /home/martin/miniconda3/envs/BAI/lib/python3.6/site-packages/torch/lib/libATen.so
#3  0x00007fffdeda145d in torch::autograd::VariableType::thnn_conv_dilated3d_forward (this=0x555556570a40, self=..., weight=..., kernel_size=..., bias=..., 
    stride=..., padding=..., dilation=...) at torch/csrc/autograd/generated/VariableType.cpp:17592
#4  0x00007fffbf625595 in at::Type::thnn_conv_dilated3d(at::Tensor const&, at::Tensor const&, at::ArrayRef<long>, at::Tensor const&, at::ArrayRef<long>, at::ArrayRef<long>, at::ArrayRef<long>) const () from /home/martin/miniconda3/envs/BAI/lib/python3.6/site-packages/torch/lib/libATen.so
#5  0x00007fffded3ec54 in torch::autograd::VariableType::thnn_conv_dilated3d (this=0x555556570a40, self=..., weight=..., kernel_size=..., bias=..., stride=..., 
    padding=..., dilation=...) at torch/csrc/autograd/generated/VariableType.cpp:17528
#6  0x00007fffbf313a2b in at::native::_convolution_nogroup(at::Tensor const&, at::Tensor const&, at::Tensor const&, at::ArrayRef<long>, at::ArrayRef<long>, at::ArrayRef<long>, bool, at::ArrayRef<long>) () from /home/martin/miniconda3/envs/BAI/lib/python3.6/site-packages/torch/lib/libATen.so
#7  0x00007fffbf620bae in at::Type::_convolution_nogroup(at::Tensor const&, at::Tensor const&, at::Tensor const&, at::ArrayRef<long>, at::ArrayRef<long>, at::ArrayRef<long>, bool, at::ArrayRef<long>) const () from /home/martin/miniconda3/envs/BAI/lib/python3.6/site-packages/torch/lib/libATen.so
#8  0x00007fffded40db7 in torch::autograd::VariableType::_convolution_nogroup (this=0x555556570a40, input=..., weight=..., bias=..., stride=..., padding=..., 
    dilation=..., transposed=false, output_padding=...) at torch/csrc/autograd/generated/VariableType.cpp:18403
#9  0x00007fffbf318fe8 in at::native::_convolution(at::Tensor const&, at::Tensor const&, at::Tensor const&, at::ArrayRef<long>, at::ArrayRef<long>, at::ArrayRef<long>, bool, at::ArrayRef<long>, long, bool, bool, bool) () from /home/martin/miniconda3/envs/BAI/lib/python3.6/site-packages/torch/lib/libATen.so
#10 0x00007fffbf620b64 in at::Type::_convolution(at::Tensor const&, at::Tensor const&, at::Tensor const&, at::ArrayRef<long>, at::ArrayRef<long>, at::ArrayRef<long>, bool, at::ArrayRef<long>, long, bool, bool, bool) const () from /home/martin/miniconda3/envs/BAI/lib/python3.6/site-packages/torch/lib/libATen.so
#11 0x00007fffded40a6e in torch::autograd::VariableType::_convolution (this=0x555556570a40, input=..., weight=..., bias=..., stride=..., padding=..., 
    dilation=..., transposed=false, output_padding=..., groups=1, benchmark=false, deterministic=false, cudnn_enabled=true)
    at torch/csrc/autograd/generated/VariableType.cpp:18386
#12 0x00007fffbf313e0e in at::native::convolution(at::Tensor const&, at::Tensor const&, at::Tensor const&, at::ArrayRef<long>, at::ArrayRef<long>, at::ArrayRef<long>, bool, at::ArrayRef<long>, long) () from /home/martin/miniconda3/envs/BAI/lib/python3.6/site-packages/torch/lib/libATen.so
#13 0x00007fffbf620afe in at::Type::convolution(at::Tensor const&, at::Tensor const&, at::Tensor const&, at::ArrayRef<long>, at::ArrayRef<long>, at::ArrayRef<long>, bool, at::ArrayRef<long>, long) const () from /home/martin/miniconda3/envs/BAI/lib/python3.6/site-packages/torch/lib/libATen.so
#14 0x00007fffdece124c in torch::autograd::VariableType::convolution (this=0x555556570a40, input=..., weight=..., bias=..., stride=..., padding=..., 
    dilation=..., transposed=<optimised out>, output_padding=..., groups=1) at torch/csrc/autograd/generated/VariableType.cpp:18368
#15 0x00007fffbf3141aa in at::native::conv3d(at::Tensor const&, at::Tensor const&, at::Tensor const&, at::ArrayRef<long>, at::ArrayRef<long>, at::ArrayRef<long>, long) () from /home/martin/miniconda3/envs/BAI/lib/python3.6/site-packages/torch/lib/libATen.so
#16 0x00007fffbf620d22 in at::Type::conv3d(at::Tensor const&, at::Tensor const&, at::Tensor const&, at::ArrayRef<long>, at::ArrayRef<long>, at::ArrayRef<long>, long) const () from /home/martin/miniconda3/envs/BAI/lib/python3.6/site-packages/torch/lib/libATen.so
#17 0x00007fffdece147e in torch::autograd::VariableType::conv3d (this=0x555556570a40, input=..., weight=..., bias=..., stride=..., padding=..., dilation=..., 
    groups=1) at torch/csrc/autograd/generated/VariableType.cpp:18446
#18 0x00007fffdee94387 in at::conv3d (groups=1, dilation=..., padding=..., stride=..., bias=..., weight=..., input=...)
    at /opt/conda/conda-bld/pytorch_1524584710464/work/torch/lib/tmp_install/include/ATen/Functions.h:3044
#19 torch::autograd::dispatch_conv3d (groups=1, dilation=..., padding=..., stride=..., bias=..., weight=..., input=...)
    at torch/csrc/autograd/generated/python_torch_functions_dispatch.h:1073
#20 torch::autograd::THPVariable_conv3d (self=<optimised out>, args=<optimised out>, kwargs=<optimised out>)
    at torch/csrc/autograd/generated/python_torch_functions.cpp:1662

And this is the environment information:

Collecting environment information...
PyTorch version: 0.4.0
Is debug build: No
CUDA used to build PyTorch: 8.0.61

OS: Ubuntu 16.04.4 LTS
GCC version: (Ubuntu 5.4.0-6ubuntu1~16.04.10) 5.4.0 20160609
CMake version: Could not collect

Python version: 3.6
Is CUDA available: No
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA

Versions of relevant libraries:
[pip] numpy (1.14.5)
[pip] torch (0.4.0)
[pip] torchvision (0.2.1)
[conda] pytorch                   0.4.0           py36_cuda8.0.61_cudnn7.1.2_1    pytorch
[conda] torchvision               0.2.1                    py36_1    pytorch

Thanks for your help!

[ssnl]

repro'ed on master!

[ssnl]

oh it overflowed int range.... fix incoming

[mys007]

Well done, Tongzhou, that was really brisk!!