Segmentation fault (core dumped) about torchnet

[chsasank]

This segfault might be because hdf5 doesn't support multi threading.

[lvdmaaten]

Ah yeah that makes sense. See for more details here: https://github.com/deepmind/torch-hdf5/blob/master/doc/usage.md#thread-safety

Closing this because this is not due to an issue in torchnet.

[lvdmaaten]

Can you please post the error message you're seeing?

[mhjabreel]

Segmentation fault (core dumped)

[lvdmaaten]

Is there no stack trace / other information dumped as part of the error message?

[mhjabreel]

@chsasank , Exactly, I have fixed the issue by using DatasetIterator instead of ParallelDatasetIterator.
Thanks.

[lvdmaaten]

@mhjabreel You may be able to build a thread-safe version of HDF5. See https://github.com/deepmind/torch-hdf5/blob/master/doc/usage.md#thread-safety

[mhjabreel]

@lvdmaaten Ok, I will check that. Thanks.