Skip to content
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.

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

tensor descriptors do not match for grouped convolutions (legacy) #188

Open
jvwilliams23 opened this issue Sep 18, 2024 · 0 comments
Open

Comments

@jvwilliams23
Copy link

Hi,
I was wondering, are grouped convolutions currently compatible with distconv/legacy dihydrogen?
When using a distconv-enabled grouped convolution in lbann, it gives the following error:

[0] [ERROR] setup algos forward: descriptors do not match:
 channel: Tensor descriptor: #dims=4, type=float, dims=1x256x128x128, strides=4194304x16384x128x1
 filter: Tensor descriptor: #dims=4, type=float, dims=1x256x128x128, strides=4194304x16384x128x1
 weights: Filter descriptor: format=NCHW, #dims=4, type=float, dims=256x1x3x3
****************************************************************
Caught signal 6 (SIGABRT - process abort signal) on rank 0
Stack trace:
   0: lbann::stack_trace::get[abi:cxx11]()
   1: lbann::exception::exception(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >)
   2: /home/jwilliams/lbann-builds/lbann-latest/build-full/install/lib64/liblbann.so.0.105.0(+0x28b7822) [0x7f9a39eb7822] (could not find stack frame symbol)
   3: /lib64/libc.so.6(+0x3e6f0) [0x7f99c503e6f0] (could not find stack frame symbol)
   4: /lib64/libc.so.6(+0x8b94c) [0x7f99c508b94c] (could not find stack frame symbol)
   5: raise (demangling failed)
   6: abort (demangling failed)
   7: distconv::Convolution<distconv::DNNBackend<distconv::GPUDNNBackend>, float>::ensure_tensor_descriptors_conform(cudnnTensorStruct* const&, cudnnTensorStruct* const&, cudnnFilterStruct* const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)
   8: distconv::Convolution<distconv::DNNBackend<distconv::GPUDNNBackend>, float>::setup_algorithms_fwd(void const*, void const*, void*, unsigned long)
   9: /home/jwilliams/lbann-builds/lbann-latest/build-full/install/lib64/liblbann.so.0.105.0(+0x18c0390) [0x7f9a38ec0390] (could not find stack frame symbol)
  10: lbann::base_convolution_adapter<float, (hydrogen::Device)1>::fp_compute_convolution()
  11: lbann::convolution_layer<float, (lbann::data_layout)1, (hydrogen::Device)1>::fp_compute()
  12: lbann::data_type_layer<float, float>::forward_prop()
  13: lbann::model::forward_prop(lbann::execution_mode, bool)
  14: lbann::SGDTrainingAlgorithm::train_mini_batch(lbann::SGDExecutionContext&, lbann::model&, lbann::data_coordinator&, lbann::ScopeTimer)
  15: lbann::SGDTrainingAlgorithm::train(lbann::SGDExecutionContext&, lbann::model&, lbann::data_coordinator&, lbann::SGDTerminationCriteria const&)
  16: lbann::SGDTrainingAlgorithm::apply(lbann::ExecutionContext&, lbann::model&, lbann::data_coordinator&, lbann::execution_mode)
  17: lbann::trainer::train(lbann::model*, long long, long long)
  18: /home/jwilliams/lbann-builds/lbann-latest/build-full/install/bin/lbann() [0x421b7f] (could not find stack frame symbol)
  19: /lib64/libc.so.6(+0x29590) [0x7f99c5029590] (could not find stack frame symbol)
  20: __libc_start_main (demangling failed)
  21: /home/jwilliams/lbann-builds/lbann-latest/build-full/install/bin/lbann() [0x421e55] (could not find stack frame symbol)
****************************************************************

but it works fine without distconv.

Regards,
Josh

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant